Laboratory Study of Simulated Atmospheric Transformations of Chromium in Ultrafine Combustion Aerosol Particles

While atmospheric particles can have adverse health effects, the reasons for this toxicity are largely unclear. One possible reason is that the particles can contain toxic metals such as chromium. Chromium exists in the environment in two major oxidation states: III, which is an essential nutrient, and VI, which is highly toxic and carcinogenic. Currently little is known about the speciation of chromium in airborne particles or how this speciation is altered by atmospheric reactions. To investigate the potential impacts of atmospheric aging on the speciation and toxicity of chromium-containing particles, we collected chromium and chromium-iron combustion ultrafine particles on Teflon filters and exposed the particles to a combination of light, ozone, water vapor, and, in some cases, basic or acidic conditions. After the aging process, the aged and not-aged samples were analyzed for Cr oxidation state using X-ray Absorption Near Edge Spectroscopy (XANES). We found that the aging process reduced Cr(VI) by as much as 20% in chromium particles that had high initial Cr(VI)/Cr(total) ratios. This reduction of Cr(VI) to Cr(III) appears to be due to reactions primarily with light and hydroperoxyl radical (HO ₂ ) in the chamber. Particles that had low initial Cr(VI)/Cr(total) ratios experienced no significant change in Cr oxidation states after aging. Compared to particles containing only Cr, the addition of Fe to the flame decreased the Cr(VI)/Cr(total) ratio in fresh Cr-Fe particles by ∼60%. Aging of these Cr-Fe particles had no additional effects on the Cr(VI)/Cr(total) ratio.

INTRODUCTION

Understanding the concentrations and composition of particulate matter (PM) in the atmosphere has become a focus for research because fine particles can be inhaled deeply into the lungs and cause a variety of negative health affects, such as bronchial irritation, reduced lung function, respiratory disease, cancer, and mortality (Pope et al. 1995; Schlesinger 2000). On a mass basis, ultrafine PM (particles with aerodynamic equivalent diameters ≤ 0.1 μ m) can have more adverse health effects than larger particles of the same composition (Ferin et al. 1990; Oberdorster et al. 1992). Ferin et al. propose that this is because the smaller particles can penetrate the interstitial space of the lungs and because the large number of particles being deposited can overwhelm lung defenses, such as removal by alveolar macrophages.

While ultrafine particles can be present at very high number concentrations, they make up only a small portion of the total particle mass in the atmosphere (Anastasio and Martin 2001). Though their high number concentrations might influence the toxicity of ultrafine particles, their composition, specifically their transition metal content, can also influence toxicity. For example, it has been observed that the transition metal component of particles is associated with acute lung injury (Dreher et al. 1997). The initial composition of “fresh” ultrafine particles is determined by the PM sources, which include fossil fuel combustion, incineration, and other human activities (Anastasio and Martin 2001). In addition, the composition of ultrafine PM will be altered by atmospheric aging involving oxidation-reduction reactions as well as the condensation of secondary species such as organics, nitrate, and sulfate (Finlayson-Pitts and Pitts 2000). Because ultrafine particles can have relatively long atmospheric lifetimes (between 1–10 days for particles between 10–100 nm (Brasseur et al. 1999); they can be transported over long distances and undergo extensive atmospheric reactions.

Chromium is a transition metal of particular interest in PM because its toxicity varies based on oxidation state. Chromium exists in two major oxidation states: +3, which is an essential nutrient, and +6, which is highly toxic and carcinogenic (Cieslak-Golonka 1996). It has been proposed that Cr(VI) is toxic because it can penetrate cells, be reduced to Cr(III), and then generate reactive oxygen species, such as hydroxyl radical, that can cause DNA damage (Cohen et al. 1993). The highest exposures to Cr(VI)-containing particulate matter likely occurs to workers during welding, chrome plating, spray painting, and chrome pigment productions (Cohen et al. 1993). However, a study in Los Angeles showed that ambient fine particles can also contain significant amounts of chromium and that the amount of Cr (relative to the total mass of metals) is greater in the ultrafine particles than that in the fine particles (Hughes et al. 1998). In that study chromium accounted for up to ∼ 10% of the transition metal mass in ultrafine particles (diameters < 0.097 μm). Chromium is also among the top five most abundant metals in particle emissions from diesel fuel combustion (Wang et al. 2003).

Sampling of emissions of particles from specific applications, such as welding fumes and fly ash from coal burning (Minni et al. 1984; Shoji et al. 2002), have shown that Cr(VI) accounts for approximately 40–90% of the total PM mass in metal arc welding fumes and 0–30% in fly ash (depending on the origin of the coal). However, very few studies have examined how the speciation of chromium, and therefore its toxicity, changes during atmospheric transport and aging.

Computer modeling results suggest that under typical atmospheric conditions Cr(VI) is reduced to Cr(III) by the reductants As(III), Fe(II), V(II), V(III), VO²⁺, and SO₂ (Seigneur and Constantinou 1995). However, this model did not include reactions of chromium species with atmospheric oxidants such as ozone (O₃), hydrogen peroxide (H₂O₂), hydroxyl radical (OH), or hydroperoxyl radical (HO₂). These reactive species are responsible for the oxidation of most trace pollutants in the atmosphere (Finlayson-Pitts and Pitts 2000) and it is possible that they also affect the speciation of Cr on atmospheric particles. Zhang and Bartlett (1999), for example, have found that aqueous Cr(III) is oxidized to Cr(VI) in the presence of light under acidic conditions, and hypothesized that this oxidation is mediated by OH generated during the photolysis of Fe(OH)²⁺. Further studies showed that the presence of organic acids, such as acetate, could have significant impacts on the light- and Fe(III)-induced oxidation of Cr(III) (Zhang 2000; Zhang and Bartlett 1999).

There has also been only one previous study of particulate chromium transformation during simulated atmospheric reactions, however, they did not include illumination during their experiments (Grohse et al. 1988). In this study, particles were generated using an aerosol nebulizer containing an aqueous solution of Cr(VI), dried, and collected on Teflon or PVC filters. These filters were then exposed to pollutants such as HNO₃, transition metals, organic acids, and ozone via a nebulizer in a plexiglass or aluminum chamber. From their chamber studies they concluded that: (1) under acidic conditions Cr(VI) reacts with oxidizable species, such as unsaturated hydrocarbons and transition metals, to form Cr(III); (2) there is an overall reduction of Cr(VI) in the reaction chamber, as well as in separate solution studies; (3) at ambient Cr levels (typically 20–100 ng) particulate Cr(VI) has an average half-life of ∼ 13 hours, though the range is broad; (4) the role of relative humidity on Cr chemistry is unclear; and (5) there is no indication of Cr(III) oxidation to Cr(VI). Cr(VI) in the experiments of Grohse et al. (1988) was determined by digestion of the filter followed by column separation with ion chromatography detection.

An alternative analytical method to determine Cr oxidation states is X-ray Absorption Near Edge Spectroscopy (XANES), which is non-destructive and requires no sample extraction. Another advantage of using XANES to measure Cr oxidation state is that the octahedrally coordinated Cr(VI) has a distinct pre-edge peak in the Cr spectrum that is nearly independent of the presence of the tetrahedrally coordinated Cr(III). Although XANES has not been previously used to determine metal oxidation states in atmospheric particles, it has been used to quantify Cr(VI) in coal (Huggins et al. 1999), on rice husks (Hu et al. 2004), and on treated wood (Nico et al. 2004).

The overall goal of this project was to determine how atmospheric aging affects the oxidation state of Cr in combustion-generated ultrafine particles in order to understand how atmospheric reactions might alter the chromium-associated toxicity of PM. In this work, we generate particles containing chromium (or a mixture of chromium and iron) in a diffusion flame, age the collected PM in a Teflon-coated reaction chamber with light, ozone, and water vapor, and determine Cr oxidation states using XANES on pairs of aged and not-aged portions of our samples.

EXPERIMENTAL

Sample Preparation

Chromium and chromium-iron aerosols were generated by a hydrogen diffusion flame that was stabilized on a nozzle located on the axis of a wind tunnel; the tunnel working section was 300 mm by 300 mm with a working length of ∼ 1.5 m. Filtered, particle-free air flowed through the wind tunnel to form a uniform laminar flow in the working section. Air velocity in the wind tunnel was kept constant for all experiments. Chromium hexacarbonyl (Cr(CO)₆) and iron pentacarbonyl (Fe(CO)₅), organometallic compounds with high saturation vapor pressures and moderate decomposition temperatures, were used as the precursors for chromium and iron. To deliver chromium to the flame, a slow stream of fuel gas (H₂) flowed through a Cr(CO)₆ cartridge (built from a commercial 47-mm filter holder) to saturate the flow with Cr(CO)₆ vapor. In the case of iron, liquid Fe(CO)₅ was contained in an air-tight bubbler, which was slowly purged with a separate stream of H₂ gas to produce saturated Fe(CO)₅ vapors. The metal-laden H₂ streams were then mixed with clean fuel gas and/or the diluent gas before they were delivered to the flame. The concentrations of Cr(CO)₆ and Fe(CO)₅ in the fuel mixture were estimated from the saturation vapor pressures of the precursors and the mixing ratio of the gases. Ambient temperature was kept constant at 23 ± 0.5°C, minimizing fluctuations in the precursor vapor pressures. We diluted the H₂ fuel gas with Ar to vary the flame temperature.

A sampling probe was positioned 520 mm above the nozzle on-axis, where the temperature and concentrations were so low that no significant chemical reaction and aerosol development was expected to take place. The post-flame aerosols were drawn through the sampling probe by vacuum and the particles were collected onto 47 mm Zefluor filters with nominal pore size of 0.2 μ m (Pall Gelman Laboratory). Based on analysis by SMPS particles were typically between 10–100 nm in size. More details on the apparatus for synthesizing the Cr and Cr-Fe aerosols can be found elsewhere (Guo and Kennedy 2004).

Aging Procedure

The atmospheric aging system is illustrated in Figure 1. Particle-free air was generated using a 737-R Aadco Pure Air Generator or introduced from a zero air cylinder (99.8% purity). This air was passed through a charcoal trap, a 47 mm Teflo filter (2.0 μm pore; Pall Gelman Labs), then two glass impingers filled with Milli-Q water (≥18.2 MΩ cm) to bring the relative humidity of the air to approximately 100%. Ozone was generated by passing oxygen (medical grade, Puritan Bennett; 0.25 sLpm) through a GE 021 quartz tube (7″ long, 5 mm ID, 7 mm OD) that ran parallel to an ozone-generating Hg lamp (P/N 81-1127-1; BHK). The amount of ozone produced in the ozonator was varied by moving an adjustable steel pipe to cover different amounts of the UV lamp. Prior to each experiment, ozone mixing ratios in the gas stream from the ozonator were measured using a Dasibi 1003-PC analyzer after mixing with 1.75 sLpm of purified, humidified air to obtain sufficient flow. During experiments the humidified air flow was adjusted to 0.75 sLpm and was combined with the ozone stream to make a total flow into the chamber of 1.00 sLpm. The final ozone mixing ratio in the aging chamber was ∼ 1 ppmv.

FIG. 1 Schematic of the aging chamber and experimental set up. All tubing is PTFE while fittings are either stainless steel or Teflon.

The aging chamber is a Teflon-coated, air-tight cylinder with a diameter of 14 cm, height of 6.5 cm, and internal volume of ∼ 1000 cm³. The air-tight Teflon lid contains a circular quartz window with a diameter of 7 cm so that samples can be irradiated during aging. The inlet and exhaust ports on the chamber are located 2.5 cm from the base of the reaction chamber and opposite one another. A charcoal trap was attached to the exhaust and vented appropriately. Samples in the chamber are held on a Teflon stand (6 cm height and 12.5 cm diameter base) with a 47 mm dia recessed edge in order to center the filter.

The illumination system (Spectral Energy Corp., model LH 151N) contains a 1000 W xenon bulb (Osram Sylvania Inc., model XBO 1000W/HS OFR) powered by an Oriel Spectral Physics power supply (model 69920) run at an average current of 43 A and voltage of 22.5 V. We use a dichroic cold mirror in the lamp housing to transmit wavelengths between 300–500 nm to the sample but remove more energetic UV as well as IR wavelengths to reduce heating of the sample and chamber.

Before each aging experiment, sampled filters were cut into 8 equal pie-shaped wedges using a custom-made cutting guide. This guide was a 2-inch high, 47 mm diameter polyvinylchloride cylinder with a 1/8 pie wedge removed (as a cutting template) and with a 2-mm high rim on the bottom to be able to hold down the perimeter of the filter without disturbing the sampled particles. For each sample, four pieces of filter were placed in a Petri dish and stored in the dark in an airtight, N₂-purged box (Pelican, model #1050, with a Swagelock fitting and septum added for purging). The other four pieces of filter were placed in the aging chamber described above.

Samples were aged in the chamber for 4–24 hours. They were exposed to simulated sunlight, ∼ 75% relative humidity, and 0.8–1.0 ppm ozone. In several experiments we aged samples without ozone or illumination. In a few cases, we exposed samples to acidic or basic vapors prior to aging by placing the filters in air-tight jars on a platform above either concentrated H₂SO₄ or NH₄OH, respectively, for 24 h. After aging, both aged and not aged pieces of sample were placed in Petri dishes and stored in an air-tight, nitrogen-purged box in a cool, dark place until analysis.

Sample Analysis by XANES

Bulk X-Ray Absorption Near Edge Spectroscopy (XANES) analysis was performed at three beamlines: (1) the majority of our data was taken at GSECARS beamline 13-BM at the Advanced Photon Source (APS) of Argonne National Laboratory using a 16 element Ge detector, a Si (111) monochromator, a beam size of ∼ 1.5 × 6.0 mm, and a relatively constant beam current of ∼100 mA; (2) beamline 4–3 at Stanford Synchrotron Radiation Laboratory (SSRL) using a Lytle type ionization detector, a Si (220) monochromator, a beam size of ∼ 20 × 4 mm and a beam current ranging from ∼ 50–100 mA; and (3) Beamline 10.3.2 at the Advanced Light Source (ALS) of Lawrence Berkeley National Lab using a seven element Ge detector, a Si(111) monochromator, a beam size of ∼ 16 × 7 μ m, and a beam current between 200–400 mA.

Pie-shaped wedges of sample filters were mounted with Kapton tape onto slide mounts then placed in the beam. Samples were run in fluorescence mode at a 45° angle to the incident beam; the fluorescent intensity was normalized to the incident X-ray intensity (I_o), which was measured using in-line ionization chambers. The energy scan was typically 5990–6150 eV in 0.5 eV steps to encompass the K-edge of Cr. Because the concentrations of Cr in our samples were relatively high, only one or two scans were taken of each filter. The monochromator energy was calibrated to a suite of chromium standards, with the Cr(VI) pre-edge peak being assigned a value of 5994.5 eV. Figure 2 shows spectra of the four standards (K₂CrO₄, Cr₂O₃, Cr(OH)₃, and chromite (FeCr₂O₄, a Cr-Fe containing spinel) used for calibration and species identification. The chromite sample was obtained from Ward's Natural Science as Chromite Research Mineral. Standard spectra of Cr(OH)₃, Cr₂O₃, and K₂CrO₄, were collected at the SSRL, the ALS, and the APS. No significant differences in spectral features were observed between standards collected at the various beam lines as a result of monochromator energy resolution, energy steps used in data collection, or count rate and shaping times of the Ge detectors. However, some standard spectra did appear to be somewhat over absorbed. The spectra in Figure 2 are the best representatives obtained for each standard species independent of where they were collected. The chromite spectra were collected solely at the APS because it was added to the set of standards after we observed it in our sample data. As shown in Figure 2a, the “pre-edge” of the spectrum, which includes the pre-edge peak refers to the normalized fluorescence below ∼ 6000 eV, the “edge” is the region between 6000–6020 (depending on the coordination chemistry of the Cr atoms), and the “post-edge” region is between approximately 6020–6150 eV.

FIG. 2 XANES spectra of standards of (a) K₂CrO₄ and chromite (Cr-Fe spinel) and (b) Cr(OH)₃ and Cr₂O₃. The insets in both figures show the pre-edge peaks at 5994.5 eV in more detail. Spectral features are described in the text.

Collected data were processed using the SixPack program (Webb 2002). Spectra taken at different times and on different beamlines were calibrated to standards such that the Cr(VI) pre-edge peak was at 5994.5 eV. Multiple scans of the same filter were checked for consistency and averaged together. Post-edges of all spectra, including standards and not-aged and aged pairs of the same sample were normalized to unity. In addition, the baseline for each pre-edge peak of Cr(VI) was subtracted to more accurately determine Cr(VI) amounts, as shown in Figure 3. The baseline subtracted height of the pre-edge peak in a given spectrum with the post-edge normalized to unity is equal to the fraction of total chromium that is present as Cr(VI), that is, the value for Cr(VI)/Cr(total) in each sample is determined directly as the baseline-subtracted height of the normalized pre-edge peak. This procedure has been previously shown to be accurate within a few percent (Patterson et al. 1997; Peterson et al. 1997). Because of the small differences between many of our not-aged and aged pairs, it was important to obtain a good estimate of our method precision. We calculated standard errors for Cr(VI)/Cr(total) in the not-aged samples based on results from replicate samples prepared on the same day under the same conditions. Because we typically did not have replicates of the aged samples, we determined uncertainties in these cases by assuming that the relative standard error of Cr(VI)/Cr(total) for the aged samples was the same as for the not-aged samples that were prepared on the same day under the same flame conditions. In order to determine whether aging had any effect on the amount of Cr(VI) in our particles we calculated p-values from a standard two-tailed Student t-test to compare the values of Cr(VI)/Cr(total) for the not-aged and aged samples within each pair.

FIG. 3 Example of how the heights of the pre-edge peaks were determined. The values for y_n, the I/I₀ value for the not-aged sample, and y_a, the I/I₀ value for the aged sample, were determined by measuring the peak height from the baselines (dotted lines) of the normalized spectra.

Metal Analysis by ICP-MS

Metals were analyzed using ICP-MS in order to determine the Cr:Fe ratio in the Cr-Fe samples as well as the ranges of concentrations of the metals in our particles. To extract the particles, we placed one eighth of a filter in a 7 mL PTFE bottle, added 6 mL of a 10% HNO₃ solution (Optima, Fisher) and sonicated for approximately 4 h. After sonication, we quantitatively transferred 5.0 mL of extract to a test tube, spiked the sample with 5.0 mL of an internal standard (100 ppb germanium in a 10% HNO₃ solution) and analyzed the extract for Cr, Fe, V, Mn, and Cu by ICP-MS (Agilent 7500a). The concentrations of V, Mn, and Cu were not found to be significant.

RESULTS AND DISCUSSION

Analysis of Standards

In order to calibrate the X-ray beam to the Cr-edge, several known Cr containing compounds were analyzed using XANES (Figure 2). Potassium chromate was used as the Cr(VI) standard and our spectrum is similar to that previously reported (Balasubramanian and Melendres 1999). The background was subtracted from the K₂CrO₄ Cr(VI) standard and the pre-edge peak was adjusted so that the standard had a Cr(VI)/Cr(total) ratio equal to 1. The spectra of Cr(OH)₃ and Cr₂O₃, which were used as representative Cr(III) standards, are shown in Figure 2; note that our Cr₂O₃ standard is similar to that reported previously (Wei et al. 2002).

The major spectral difference between Cr(VI) and Cr(III) is the presence of the pre-edge peak at 5994.5 eV caused by Cr(VI). As discussed above, the height of this peak is proportional to the fraction of Cr(VI) in a sample. In addition, the post-edges in the spectra of the various Cr species also vary, as a result of differences in the atoms surrounding the Cr atoms. For example, the chromite sample (Cr-Fe spinel) has a distinctive post-edge shape that is quite different from the other three Cr standards (Figure 2). In addition to these differences, the position of the main absorption edge shifts to higher energies as the average oxidation state of Cr in the sample increases. For example, as shown in Figure 2, the edge for both Cr(III) standards occurs between 5995–6005 eV, while the edge for K₂CrO₄ occurs between 6005 and 6015 eV.

Effects of Flame Temperature and Composition on the Cr(VI)/Cr(total) Ratio

The first variable we explored was the effect of flame temperature on the Cr(VI)/Cr(total) ratio in the particles. The two dominant fuel conditions we used were a hotter, 100% H₂ flame with a measured temperature of approximately 2480 K and a cooler, 33% H₂/67% Ar flame with a calculated temperature of approximately 1840 K (Reynolds 1986). We also prepared a series of samples using a cooler 33% C₃H₈/67% Ar flame, which produced particles containing soot as well as metal. Based on ICP-MS results from a subset of samples, the amount of Cr collected on each filter was typically 10–30 μ g for particles from a 33% H₂ flame and 140–340 μ g per filter for particles from the 100% H₂ flame.

As shown in Figure 4, the fuel composition, and thus the temperature of the flame, has a large effect on the oxidation state of the chromium in the resulting particles. The inset of Figure 4 shows that particles from the 100% H₂ flame seeded with Cr have a baseline-subtracted pre-edge peak height (i.e., Cr(VI)/Cr(total) ratio) of 0.43, indicating that 43% of the chromium is present as Cr(VI). In contrast, the sample from the cooler 33% H₂ flame seeded with Cr has a Cr(VI)/Cr(total) ratio of 0.051, almost an order of magnitude less than that from the 100% H₂ sample. Samples from a propane flame, by far the coolest of three flame conditions, had no detectable pre-edge peak, indicating there is little, if any, Cr(VI) in these samples (Figure 4).

FIG. 4 XANES spectra of particles collected from flames fed with 100% H₂, 33% H₂, and 33% C₃H₈ and containing Cr(CO)₆ or both Cr(CO)₆ and Fe(CO)₅. The laboratory where each sample was analyzed is listed in parentheses.

Figure 4 also shows that as the Cr(VI)/Cr(total) ratio increases, the edge is shifted to higher beam energies, consistent with the theoretically expected trend observed with our standards (Figure 2). The edge for the 100% H₂ Cr only sample begins at approximately 6000 eV, which coincides with the more tightly bound electrons around the Cr(VI) atoms in K₂CrO₄. In addition, the post-edge oscillations of Cr particles from the 100% H₂ flame resemble those of K₂CrO₄ (Figure 2a), while the post-edge oscillations of both the 33% H₂ and C₃H₈ flames resemble those of Cr₂O₃ (Figure 2b).

We also explored how the presence of Fe affects the oxidation state of Cr in ultrafine combustion particles. Iron is a common constituent in combustion particles as well as in ambient aerosols (Puxbaum et al. 2004; Wang et al. 2003). For example, ratios of Cr:Fe in ambient particulate matter collected in and around Vienna range from 1:2 to 1:10 in urban and rural settings respectively, depending on the types of industry in the area (Puxbaum et al. 2004). The Cr:Fe ratios in our samples, which were determined by ICP-MS, were between 1:2 to 1:3, with a typical iron concentration of 190–460 μ g per filter.

As shown in Figure 4, this sample from a 100% H₂ flame seeded with both Cr and Fe has a Cr(VI)/Cr(total) ratio of 0.12, compared to a value of 0.43 in the particles from the 100% H₂ flame that contain Cr but no Fe. This indicates that adding Fe decreases, but does not eliminate, the presence of Cr(VI) in combustion particles. Note that the edge of the Cr-Fe sample also lies between the edges for the Cr-only particles created from the 100% H₂ and 33% H₂ flames, consistent with the intermediate Cr(VI)/Cr(total) ratio in the presence of iron.

Interestingly, in addition to decreasing the Cr(VI)/Cr(total) ratio, the presence of Fe in the flame appears to lead to the formation of a new solid phase material. As seen in Figure 4, except for the presence of a small amount of Cr(VI), the edge and post-edge sections of the spectra of the Cr/Fe samples strongly resemble that of the Cr-Fe mixed phase (chromite) standard spectra shown in Figure 2. The similarity between these spectra implies that the major form of Cr in these mixed phase samples is that of a Cr-Fe spinel with a structure very similar to natural chromite.

Effects of Atmospheric Aging on Cr Only Samples

Though it is important to understand the composition of particles from combustion processes, it is equally important to know what happens to the composition during particle transport. We performed simulated aging experiments on particles containing only Cr as well as those containing both Cr and Fe in order to see how atmospheric chemistry affects the oxidation state of Cr. Figure 5 shows one example of the effects of simulated atmospheric aging on Cr speciation in particles created from a 100% H₂ flame seeded with Cr(CO)₆. Prior to aging, the Cr(VI)/Cr(total) ratio is 0.44, while after 24 hours of aging (with ozone, humidified air, and illumination) the Cr(VI)/Cr(total) ratio decreased to 0.34, a ∼ 23% reduction in the initial amount of Cr(VI). The post-edge oscillations in the aged sample are also somewhat different from the not-aged sample, indicating that the bonding environment around the Cr atoms has changed after exposure, consistent with a change from Cr(VI) to Cr(III).

FIG. 5 XANES spectra of not-aged and aged chromium ultrafine particles generated using a Cr-seeded 100% H₂ flame. The inset is a close-up of the pre-edge peaks, which have heights of 0.44 and 0.34, for the not-aged and aged samples respectively. This corresponds to a 23% reduction in the peak height and thus in the amount of Cr(VI) during aging.

Figure 6 shows the Cr(VI)/Cr(total) ratios for all the aged and not-aged samples created from Cr(CO)₆ seeding of a flame with either 33% H₂ (samples 1–5) or 100% H₂ (samples 6–14). The average (±1 σ) initial Cr(VI)/Cr(total) ratio in samples from the cooler 33% H₂ flame (0.043 ± 0.0059) is approximately ten-fold smaller than that in samples made with the hotter 100% H₂ flame (0.49 ± 0.060). The large variation in the initial Cr(VI)/Cr(total) ratio of samples collected in the 100% H₂ flames is due to variations in the combustion conditions between different collection periods. In order to remove this variation in initial Cr(VI)/Cr(total) ratios and more sensitively determine the effects of aging, we used different portions of the same filter to make the not aged and aged pair for a given sample.

FIG. 6 Comparison of Cr(VI) amounts in not-aged and aged particles from Cr-seeded flames. The dotted line separates the samples made with 33% H₂ and 100% H₂ flames. The information listed below each pair of samples corresponds to: the laboratory where the analysis was completed, the sample identification number, how long the sample was aged, and additional information. The standard aging conditions were ∼ 1 ppmv O₃, 75% RH, and illumination. Samples 7 and 8 were placed in glass jars and exposed to NH₃ and H₂SO₄ vapors, respectively. A single * represents a pair of samples where the difference in the amount of Cr(VI) is statistically different at the p < 0.10 level, while a **designates a pair with a difference at p < 0.005. The error bars represent ± one standard deviation.

As shown in the left panel of Figure 6, aging of ultrafine particles from the 33% H₂ flame had no real impact on the oxidation state of Cr. While aging did lead to a statistically significant (p < 0.10) increase in Cr(VI)/Cr(total) in samples 4 and 5, these represent very small increases (0.0055 ± 2.1 × 10^{− 4}). In addition, samples 1 and 2 in this series show the opposite trend, that is, minor (and statistically insignificant) reductions of Cr(VI) during aging. Thus it appears that aging has, at most, only a minor impact, on particles with low initial Cr(VI) levels, suggesting that at these low Cr(VI)/Cr(total) ratios there might be a steady state between Cr(III) and Cr(VI).

In the case of Cr ultrafine particles formed in the hotter, 100% H₂ flames (with higher initial Cr(VI)/Cr(total) values), the aging reactions had a more pronounced impact on chromium oxidation states (Figure 6). For example, as shown by sample 6, 24 hours of aging decreased the Cr(VI)/Cr(total) ratio from 0.44 to 0.34, a reduction of 23% (p = 0.0014). We found similar reductions in the amount of Cr(VI) for the same particles that were exposed to either basic (sample 7) or acidic (sample 8) vapors prior to aging. Values of Cr(VI)/Cr(total) in these samples were decreased by 28% and 15%, respectively, by aging (p < 0.004). The fact that we observe conversion of Cr(VI) to Cr(III) under basic conditions, and in particles with no pH adjustment, is in contrast to the conclusions of Groshe et. al. (1988), where acidic conditions were a major factor for Cr(VI) reduction. However, the low pH conditions of Groshe et. al. were probably more acidic than ours since they used a permeation source to deliver a continuous flow of HNO₃ through their filters while we merely exposed the samples to H₂SO₄ vapors in a chamber before aging.

In addition to these exposures for 24 hours, we also examined changes in Cr oxidation states after aging for 4, 8, 12, and 24 hours (samples 9–12; Figure 6). After 4 hours of aging there was a small (1%) and statistically insignificant (p = 0.26) reduction in Cr(VI)/Cr(tot) (Figure 6). Larger, and statistically significant (p < 0.09) reductions in Cr(VI) occurred after aging for 8, 12, and 24 hours. After 24 hours (sample 12), the Cr(VI)/Cr(total) ratio was decreased by 4% (p = 0.015). Based on the results from these four experiments, which are consistent with a first order decay of Cr(VI), we calculated the half life of Cr(VI) to be approximately one year. In contrast, the range of Cr(VI) half-lives calculated for samples 6–8 is much shorter, approximately 50–100 hours. However, it should be noted that these values were calculated based on only two points for each sample, assuming a first order decay.

As illustrated by these different half-lives, the extent of Cr(VI) reduction seen in sample 12 is much smaller than that in samples 6–8, which were prepared on a different day and analyzed using a different beamline. While samples 6–8 were aged using air from an Aadco Pure Air generator, while samples 9–14 were aged using air from a zero air cylinder. After the SSRL samples (6–8) were aged and analyzed, we found that the Aadco methane reactor, which removes hydrocarbons from the air stream, was faulty. While we do not know when this component failed, it is possible that it did so prior to aging of samples 6–8, which would mean that the air stream would have had levels of organics similar to that in our laboratory ambient air (which was used as the feed for the pure air generator). This could be the reason why the reduction of Cr(VI) was more rapid in samples 6–8. If this is the case, it suggests that this more rapid reduction, as well as the shorter Cr(VI) half life, is more environmentally relevant because of the presence of organics in ambient air.

As for the other samples in this series of experiments, sample 13, which was aged with no ozone and consequently no radicals, had a significant 7% reduction in the amount of Cr(VI) (p < 0.001). While these conditions would need to be more carefully explored to test this result, it suggests that the combination of illumination, water vapor, and whatever trace pollutants are in the air stream (e.g., hydrocarbons) can reduce Cr(VI) to Cr(III). The dark control, that is, the sample exposed to H₂O(g) and O₃ with no light (sample 14), showed no significant change in the Cr(VI)/Cr(total) ratio (p = 0.37), suggesting that light and the subsequently formed reactive oxygen species from ozone photolysis play a major role in the reduction of Cr(VI) in our system. It also illustrates that ozone itself (in the presence of water vapor) is not an important reactant for Cr transformations.

Effects of Aging on Cr-Fe Samples

We also investigated the effects of aging on the Cr-Fe ultrafine particles using experimental procedures identical to those used with the Cr only samples. Figure 7 shows one example of the effect of aging on Cr-Fe particles, which in this case leads to a reduction of Cr(VI). The not-aged Cr-Fe sample has a Cr(VI)/Cr(total) ratio of 0.12, which decreases by 20% to 0.096 after aging.

FIG. 7 XANES spectra of not-aged and aged particles containing both Cr and Fe from a 100% H₂flame. The inset shows the pre-edge peak heights of the not-aged and aged peaks, which have baseline-subtracted values of 0.12 and 0.096, respectively.

Figure 8 shows the results of aging for all of the Cr-Fe samples, both for the 33% H₂ flame (sample 15) and the hotter 100% H₂ flame (samples 16–24). Sample 15 shows that the 33% H₂ flame Cr-Fe particles have a very low initial Cr(VI)/Cr(total) ratio and that aging causes no change in the Cr speciation. For the 100% H₂ flame samples (16–24) there are small changes in Cr(VI)/Cr(tot) ratios between the aged and not-aged samples but only Sample 19, which was aged for 8 hours, shows a significant difference (p = 0.10). However, averaged over samples 16–22, aging caused no significant change in the Cr(VI)/Cr(tot) ratio (average ± σ = 0.0102 ± 0.019). Furthermore, samples 18, 20, and 21, which were aged for 4–24 hours, did not show any significant changes, suggesting that exposure time does not affect the results. Furthermore, exposing sample 24 to light and humidified air did not significantly change the Cr(VI)/Cr(tot) ratio, in contrast to the case of the Cr-only sample (Figure 6). Sample #24, which was not exposed to light, showed no significant change. Overall, aging had no significant effect on the Cr-Fe samples.

FIG. 8 Comparison of Cr oxidation state in not-aged and aged Cr(VI)/Cr(total) for all particles containing both Cr and Fe at a ratio of approximately 1:2–1:3. A single * represents a pair of samples where the difference in the amount of Cr(VI) is statistically different at the p < 0.10 level, while a **designates a pair with a difference at p < 0.005. The error bars represent ± one standard deviation.

Conceptual Model of Cr Chemistry During Aging

Our goal in this section is to examine the thermodynamics during sample aging in order to determine the likely mechanisms responsible for the observed Cr(VI) reduction in some of our samples. As shown in Figure 9, photolysis of ozone in the presence of water vapor forms a series of secondary photo-oxidants that include hydroxyl radical (OH), hydroperoxyl radical (HO₂), and hydrogen peroxide (HOOH), all of which can partition to the particles and potentially alter Cr speciation. The first step in examining this chemistry is to estimate the composition of the particles during aging. Based on TEM images of Cr particles from the diffusion flame (Guo and Kennedy 2004) we expect that at 75% RH our Cr ultrafine particles contain a solid core of Cr(III) (with some Cr(VI)) surrounded by a liquid layer containing primarily Cr(VI). We estimate that this liquid layer has a pH of ∼ 5 based on rough estimates made using pH paper on very slightly moistened PM samples. From the pKa values for H₂CrO₄ and HCrO₄ ⁻ (0.75 and 6.49, respectively (Perrin 1982); most Cr(VI) in the liquid layer will be present as HCrO₄ ⁻. Assuming that Cr(III) solubility is controlled by equilibrium with solid Cr(OH)₃, we estimate that the liquid layer also contains 6 μ M Cr(H₂O)₆ ³⁺ and 90 μ M CrOH(H₂O)₅ ²⁺. Finally, based on estimated concentrations of the gas phase oxidants and their Henry's law constants (Finlayson-Pitts and Pitts 2000) we estimate that the concentrations of oxidants in the particle liquid layer are: OH (1 × 10⁻¹² M), HO₂ (8 × 10^{− 9} M), H₂O₂ (8 × 10^{− 5} M), O₃ (1 × 10^{− 8} M), HO₃ (assumed to be 1× 10^{− 15} M), and O₂ (2.6 × 10^{− 4} M). Although only approximations, these oxidant and chromium concentrations are used to adjust standard state Δ G⁰ values to more realistic thermodynamic values, Δ G, as discussed below.

FIG. 9 Schematic of the proposed chemistry occurring in the chamber and on the particles during the aging experiments.

The five main photo-oxidants in our system vary significantly in strength from OH (the strongest oxidant) to O₂ (the weakest) as can be seen from their standard state reduction half-reactions (Wardman 1989):

Similarly, the chromium species have different stabilities as shown by the standard state reduction half-reactions of bichromate to form different Cr(III) species (Woods and Garre1987):

By combining the oxidant couples (half-reactions (R1–R5) with the chromium half-reactions (R6–R9), and adjusting the standard state values using the assumed reaction concentrations above, we can estimate whether a given reaction should occur spontaneously in our model system (i.e., Δ G (or Δ G⁰) < 0). As a first step at understanding the chemistry, we focus here on the two main Cr(III) species, Cr₂O₃(s) and Cr(OH)₃(s). For example, the combination of OH (R1) with Cr₂O₃ (R6) is non-spontaneous:

Indeed, solid Cr₂O₃ does not react spontaneously with any of the oxidants to form Cr(VI), indicating that Cr₂O₃ in our aging chamber is essentially inert to oxidation. The situation is similar, although more complicated, for solid Cr(OH)₃. While oxidation of Cr(OH)₃ by OH and HO₂ (R11 and R12) is thermodynamically favored, oxidation by the other photo-oxidants is not spontaneous (R13–R15).

In fact, these calculations show that HO₂ can actually serve to reduce Cr(VI) to Cr(III), with a very strong thermodynamic driving force (i.e., ΔG = −377 kJ/mol for the reverse of R15). This potential role of HO₂ as a reductant is interesting because simple kinetic photochemical modeling of our aging chamber conditions indicates that the gas phase concentration of HO₂ is approximately 20 times higher than that of OH. When these observations are combined with the known kinetic stability of Cr(III) species (Nico and Zasoski 2000), it seems reasonable that we observed no significant oxidation of Cr(III) species, but rather saw Cr(VI) reduction. Based on the abundance of HO₂ and our thermodynamic calculations, this reduction might have been due to HO₂, but it is unclear whether this reaction is rapid enough to account for our observations. Overall, the low solubility and kinetic stability of the Cr(III) compounds strongly limits their tendency to be oxidized. On the other hand, HCrO₄ ⁻ is more available for reaction because of its solubility and it is generally thermodynamically poised for reduction. It should also be mentioned that since a number of the Δ G values are close to zero (within ± 50 kJ/mol) there is a possibility that a pseudo-equilibrium condition could exist between Cr(OH)₃ and HCrO₄ ⁻ under atmospheric conditions. Similarly, since both oxidation and reduction reactions are thermodynamically allowed under the current experimental conditions, it is possible that a steady state between Cr(III) and Cr(IV) may exist under some circumstances. The presence of a such a steady state could explain the observation that the distribution of Cr oxidation states remains unchanged during aging of particles with low initial Cr(VI), possibly because the rates of Cr(III) oxidation by OH and Cr(VI) reduction by HO₂ are approximately equal under these conditions.

ENVIRONMENTAL IMPLICATIONS AND CONCLUSIONS

Our goal was to examine how atmospheric aging affects Cr speciation on ultrafine particulate matter from combustion sources. We simulated a simple atmosphere with humidified air, ozone, and light to determine the fate of Cr on ultrafines generated from a laboratory flame. Our results show that the amount of Cr(VI) present in the particles (prior to aging) decreases with decreasing flame temperature and that the addition of Fe to the fuel also reduces the amount of Cr(VI). Our aging experiments show that after 24 hours of aging there is up to a 20% reduction in the Cr(VI)/Cr(total) ratio for particles produced from the hottest flame (100% H₂). The half life for Cr(VI) in these experiments was typically 50–100 hours. Groshe and co-workers reported an average half-life of 13 ± 5.8 hours for experiments that included organics. However, the half life of Cr(VI) in their experiments also ranged as high as ∼200 hours under conditions where there were no organics (Grohse et al. 1988). Our observed reduction of ultra fine Cr(VI) to Cr(III) during simulated atmospheric aging is supported by thermodynamic calculations, which suggest that Cr(VI) reduction is more important than Cr(III) oxidation during aging. This net reduction of Cr(VI) is consistent with previous model predictions (Seigneur and Constantinou 1995), although our mechanism of reduction is very different.

The fact that Cr(VI) is generally reduced, both in our experiments as well as those of Grohse and co-workers (1988), suggests that the Cr-associated toxicity in particulate matter decreases over time. However, there are a number of important caveats to this statement. First, we find that particles where Cr(VI)/Cr(total) is relatively low (≤15%), the Cr(VI) is not significantly reduced by aging (e.g., in our Cr PM from the 33% H₂ flame or in the Cr-Fe particles from the 100% H₂ flame). Grohse and co-workers observed a wide range in the reduction and half life of Cr(VI) depending on the experimental conditions under which their particles were aged (Grohse et al. 1988). Second, both our particles and aging conditions are simple compared to ambient PM and atmospheric conditions, and thus our experiments could be missing significant reactions that affect Cr oxidation states in the atmosphere. Finally, even if Cr(VI) is reduced to Cr(III) on a timescale of tens of hours in the atmosphere, there can still be significant exposure for populations that are relatively near Cr(VI) emission sources.

Acknowledgments

We thank Dr. Matthew Newville of the GSECARS at the Advanced Photon Source (APS), Argonne National Laboratory, for his assistance. The majority to the X-ray data presented above were collected at GeoSoilEnviroCARS (Sector 13), which is supported by the National Science Foundation—Earth Sciences (EAR-0217473), Department of Energy—Geosciences (DE-FG02-94ER14466) and the State of Illinois. Use of the APS was supported by the U.S. Department of Energy under Contract No. W-31-109-Eng-38. We would also like to thank Dr. Matthew Marcus of beamline 10.3.2 of the Advanced Light Source (Lawrence Berkeley National Laboratory), which is supported by the US Department of Energy (DE-AC02-05CH11231). Portions of this research were also carried out at the Stanford Synchrotron Radiation Laboratory, operated by Stanford University on behalf of the U.S. Department of Energy. Finally, we thank David Paige (Paige Instruments) for constructing the ozone generator, John Newberg and Mike Jimenez-Cruz for building the ozone dilution and delivery system, and Michelle Gras of the UC Davis Interdisciplinary Center for Plasma Mass Spectrometry for conducting the ICP-MS analyses. This work was supported by grant number 5 P42 ES04699 from the National Institute of Environmental Health Sciences (NIEHS) of the NIH, and by a fellowship from the UC Davis NEAT-IGERT program to Michelle Werner (NSF IGERT Grant # 9972741). Partial support was also provided by the U.S. Department of Energy under contract number DE-AC02-05CH11231. The contents of this paper are solely the responsibility of the authors and do not necessarily represent the official views of the NIEHS, NIH.