Chromium and Zinc Speciation in Airborne Particulate Matter Collected in Ulaanbaatar, Mongolia, by X-Ray Absorption Near-edge Structure Spectroscopy

This study examined the chemical speciation of atmospheric particulate matter (PM) samples collected in Ulaanbaatar, Mongolia. The health effects of atmospheric aerosol depend on the particles’ size distribution, elemental composition and chemical species as well as other factors, which vary according to the pollution sources and meteorological conditions. Employing the International Atomic Energy Agency (IAEA) supported X-ray fluorescence (XRF) beamline at Elettra Sincrotrone Trieste, Italy, we used a small incidence angle for the synchrotron emission radiation in this ultra-high-vacuum environment to analyze the fine and the coarse mode (PM2.5 and PM2.5-10, respectively) through X-ray absorption near-edge structure (XANES) spectroscopy, which was applied to the K-edges of the chromium (Cr) and zinc (Zn) in PM samples selected via energy-dispersive XRF (EDXRF) analysis for their high concentrations of these transition metals. The spectroscopic results identified trivalent chromium sulfate [Cr2(SO4)3] and chromium oxide (Cr2O3) compounds as the major chemical forms of Cr in both the coarse and the fine PM fraction. Furthermore, both fractions contained abundances of sulfate (ZnSO4) and silicate (Zn2SiO4) compounds, but only the fine mode contained Zn oxalate (ZnC2O4), whereas only the coarse mode contained Zn chloride (ZnCl2). These Cr and Zn species seemed to originate from local anthropogenic sources, e.g., combustion products or traffic-related resuspended road dust. Our findings, which are based on the first chemical speciation analysis from Ulaanbaatar, provide insight into the physiochemical characteristics of atmospheric aerosols as well as information on the potential sources of Cr and Zn species bound to fine and coarse PM.


INTRODUCTION
Atmospheric particulate matter (PM) originating from anthropogenic and natural sources represents a common concern worldwide due to their associations with climate change, atmospheric visibility and more importantly with human health (Fittschen, 2014). The size of PM is considered significant since an epidemiological study revealed a correlation between PM mass of coarse particles with diameter of less than 10 µm, (PM10-2.5) and especially fine particles with diameter of less than 2.5 µm (PM2.5) and a variety of adverse effects on health (Cruz et al., 2015). Causes and consequences of PM depend on the size distribution, concentration level, elemental composition and especially on the type of chemical species. These features characterize the toxicity of PM, including physical and chemical properties and this might vary from one place to another, characteristics of an area and contributing sources. Therefore, chemical speciation studies are important for the assessment of the toxicity of certain elements, particularly of heavy metals (Osan et al., 2010), as well as for the identification of PM sources.
X-ray absorption near-edge structure (XANES) spectrometry is one of the most suitable techniques to characterize different oxidation and chemical states of elements in atmospheric PM samples. Other techniques, such as ultraviolet-visible (UV-Vis) spectroscopy, inductively coupled plasma mass spectrometry and X-ray emission techniques, can measure only the total concentration of the element and require either a wet extraction or separation step for the speciation analysis (Ashley et al., 2003;Unceta et al., 2010). However, the experimental conditions adopted for the extraction may significantly influence the reliability of the measurement results (Pettine and Capri, 2005). Great advantages of XANES rely on the fact that it does not require any chemical pretreatment of the sample and can distinguish between different compounds of the same metal, including different oxidation states (e.g., Cr(VI), Cr(III), and Cr(0)) as well as different compounds with the same oxidation state (e.g., Cr2O3 vs. Cr(OH)3) (Werner et al., 2007). Moreover, XANES allows nondestructive speciation of metals in solid samples, even for minor or trace concentrations (< 1%) (Osan et al., 2010).
The X-ray absorption measurements require an excitation source with high monochromaticity and energy tunability to scan with fine energy steps the selected absorption edge of the probed element and reveal the fine structure. When the probed element is available only in minor/trace amounts, the high intensity is an additional strict requirement for the X-ray source to allow short measuring times. Synchrotron radiation is optimally suitable for such studies, when using a double-crystal monochromator providing a resolving power ΔE/E better than 10 −4 .
A number of studies have been published on chromium and zinc speciation of airborne PM samples by XANES. Ohta et al. (2006) examined the variation of chemical compositions and speciation of Zn and other elements in different particle size aerosols in aeolian dust during their transportation from China to Japan. The results suggested that Zn originated primarily from crustal aerosols during the dust event, except for that in fine grains, and that the fraction of anthropogenic materials increased from 40% to 80%, concomitant with decreasing particle size. Wang et al. (2007) also applied XANES to Zn, Cr and other metals on two fractions of aerosol samples from different sites of Shanghai, China. Close to iron and steel industrial area, the main chemical components of Cr in PM were chromite and oxides of trivalent chromium. Werner et al. (2007) studied Cr speciation in three different areas of north California, USA. For every sampling station, trivalent chromium oxides, Cr-Fe and chromite-like phase were found to be the dominant species, while some reduced Cr species (i.e., Cr(0) or Cr3C2) were also identified as minor chemical components. Cr(VI) containing particles were only found at the site of the most urban character. Similarly, in Saudi Arabia, the dominant state of chromium in two-fractioned aerosol samples was the trivalent chromium oxide, while Cr(VI) was a minor contribution with range of 7-9% (Shaltout et al., 2018). Many other studies were successfully conducted using XANES analysis for elemental speciation (Huggins et al., 2000;Boman et al., 2004;Galbreath and Zygarlicke, 2004;Goodzari et al., 2008).
The aim of the present paper is to gain information about chemical speciation of two elements of interest (Cr and Zn) in atmospheric PM2.5 and PM10-2.5 samples, which were collected from Ulaanbaatar, the capital city of Mongolia. These two transition metals were selected because of their importance and abundance in the collected atmospheric samples. Acquiring knowledge on Cr and Zn speciation in local atmospheric PM is important for the assessment of toxicity as well as for better characterization of their corresponding sources.
The present study provided enhanced insights in Ulaanbaatar's size-fractioned atmospheric aerosols' physiochemical nature. It delivers further information on the potential origin and contributing pollution sources of the selected elements through their speciation for the first time.

Sample Collection
Atmospheric particles were collected on polycarbonate filters at the air quality monitoring site, 4 km east of the center of Ulaanbaatar (latitude 47°55ʹ, longitude 106°55ʹ, 1300 m above sea level). More information about the sampling site is described in Gunchin et al. (2019). In total, 52 samples were collected, twice a week (every Monday and Thursday) throughout the period June 2016-January 2017. A GENT sampler (Maenhaut et al., 1993) was used, comprising a PM10 impactortype size-selective inlet and a stacked filter unit (SFU) assembly connected to a pump and an air flow meter. The SFU is made up of two filters arranged sequentially, allowing the fractionation of particles on two different stages. The top filter (pore size 8 µm) collected the coarse (PM10-2.5) particle fraction and the bottom filter (pore size 0.4 µm) collected the fine (PM2.5) particle fraction.
The sampler was operated with the average flow rate of 16.0 L min −1 in alternating 30 min onand 30 min off-periods over the course of the 24 hours as to provide a representative sample for the whole day. However, due to atmospheric conditions and the level of the pollution of that sampling day, it was not possible to collect over the entire 24 hours as high pollution resulted in clogging of the filters. Due to this fact, the flow rate of the sampler was varying between 18.0-14.0 L min −1 and averaging and calculations of the total air volume for each sample measurement were done accordingly. Gravimetric masses of PM samples were subsequently normalized by the estimated volume.
The filter cassette was mounted approximately 2.5 m above the ground. The sampling point is surrounded by buildings, traditional dwelling areas (ger khoroolol), and the combinations of asphalt and unpaved roads.

Selection of Samples for XANES Measurements
For the XANES measurements, six PM2.5 samples and three PM10-2.5 samples were selected for the speciation of two elements of interest-Cr and Zn. Selections were based on the highest relevance when these elements were found at elevated concentrations. The elemental analysis of the fine and coarse PM samples was performed by means of the PANalytical Epsilon 5 energydispersive X-ray fluorescence (EDXRF) spectrometer at the premises of the Nuclear Science and Instrumentation Laboratory of the International Atomic Energy Agency (IAEA), Seibersdorf, Austria (Gunchin et al., 2019). The spectrometer features various secondary targets, such as Al, CaF2, Fe, Ge, Zr, Mo, Ag and Al2O3 (Barkla polarization target) which are irradiated by a 600 W Sc-W anode X-ray tube. The characteristic X-ray radiation emitted by the sample is detected by an LN2-cooled LEGe detector with energy resolution of about 150 eV at Mn-Kα (5.89 keV).
Fe and Ge secondary targets were used for Cr and Zn determination, respectively. Operational conditions for both secondary targets were 75 kV and 8 mA. The live time for each secondary target was set to 300 s. The analysis was performed under vacuum conditions.
The quantification of elements' concentrations was carried out using a calibration performed by measuring thin film single element and compound reference materials. The built-in software of the PANalytical Epsilon 5 was used for spectrum deconvolution, whereas the elemental concentrations were calculated using a sensitivity curve determined from the measured set of reference materials. Thin film approximation was applied. A Standard Reference Material (SRM) of particulate matter on filter media (SRM 2783; NIST, Gaithersburg, MD, USA) was used for validation of the analytical method. The uncertainty of the results were evaluated according to the study of Gutknecht et al. (2010), where the following contributing factors were considered: the peak area statistical uncertainty, the fitting error in the calibration (sensitivity) curve, the quoted uncertainties of the elemental areal densities reported for the standards (5%), the estimated error due to the omittance of the self-attenuation effect as well as the relative standard deviation of three replicate measurements (reproducibility in operation). A more detailed description of the evaluation of the combined uncertainty of obtained result can be found in Gunchin et al. (2019).
In Fig. 1, the EDXRF results reported as time series of Zn and Cr concentrations in ng m −3 for all measured fine (a) and coarse (b) particulate samples are shown (Gunchin et al., 2019). The sample selection for XANES analysis was based on higher Zn and Cr concentration: Those samples are marked with red circles in Fig. 1. The selected six fine samples were collected on 7 Nov, 14 Nov, 5 Dec, 19 Dec 2016 and 5 Jan, 12 Jan 2017 and the selected three coarse samples have collection dates of 7 Nov 2016 and 5 Jan and 12 Jan 2017.

XANES Measurement
X-ray absorption near-edge structure (XANES) spectra were acquired at the XRF IAEA experimental end station (Lubeck et al., 2016) of Elettra Sincrotrone Trieste (Basovizza, Italy). The experimental set-up is comprised by a Si(111) double-crystal monochromator, an ultra-thin-window Bruker silicon drift detector (30 mm 2 , energy resolution 135 eV at Mn-Ka), photodiodes and other beam-monitoring devices. The samples were inserted into an ultra-high-vacuum chamber equipped with a multi-axis motorized sample manipulator, which offers the possibility to vary the orientation of the sample surface with respect to the exciting beam direction (Wrobel et al., 2017;Karydas et al., 2018).
The selected PM samples were mounted on Al holders using carbon tape and they were forced to be fixed as flat as possible. For the XANES measurements the samples have been excited under a shallow angle of incidence (~1°). The influence of the angle of incidence on detection limits was investigated by Karydas et al. (2018) using a thin reference material deposited on polycarbonate filter. The grazing incidence geometry provides a larger beam footprint resulting to an increase of the PM analyzed mass and better counting statistics. The specially designed Al holder allows to measure in this geometry without producing parasitic fluorescence and scattered radiation. The comparison of shallow angle excitation geometry on aerosol sample to a conventional 45°/45° geometry resulted in significantly increased signals (~10-fold). XANES spectra with satisfactory good statistics could be therefore collected in a shorter time with the beam irradiating the samples with shallow angle. Energy scans were performed around the K-edges of the two elements of interest (5989 eV for Cr and 9659 eV for Zn) using the Si(111) monochromator (energy resolution of about 1 eV) and recording the corresponding Kα fluorescence intensity in fluorescence mode. The used energy step size varied between 0.5 (edge region) to 2 eV. The measurement time for each energy step varied from 10 to 20 s, depending on the concentration of the element in the PM samples. Two to four repetitions were done for each sample, in order to increase the data quality. Cr and Zn metal foils were used for energy calibration of the Si(111) monochromator.
The obtained spectra were processed with the Athena Demeter software (Ravel and Newville, 2005), which uses a least-squares algorithm to refine the linear combination fit (LFC) of a given number of reference spectra to the measured spectrum. The refining procedure is the following: The sample spectra were initially fit using all of the given reference spectra. After each fitting, any references that accounted for less than 5% of the total amount of species in the sample were removed from the fitting to eliminate both minor components and erroneous fittings. After this, the sample spectrum was fit again with only the references of the remaining species. This process was repeated until the sample spectrum was described by one or more references each accounting for 5% or more of the total species. The quality of the fit can be quantified by an error term (R-factor) which indicates the proper choice of standard compounds. Averaging, background removal and normalization of XANES spectra were done by the Athena software.

Speciation of Chromium
Compositional information derived from LCF suggested that chromium sulfate and chromium oxide are likely present in both the fine and coarse fraction samples. Fig. 2 shows the normalized XANES spectra at the Cr K-edge of the measured fine (a) and coarse (b) sample sets from Ulaanbaatar, together with two reference compounds. Representative XANES spectra of Cr K-edge from the LFC analysis for fine (a) and coarse (b) sample are also illustrated in Fig. 3.   Fig. 2. Cr K-edge XANES spectra of (a) fine and (b) coarse samples; standard spectra of Cr sulfate and Cr oxide are plotted for comparison. Fig. 3. Representative LCF spectra of the (a) fine and (b) coarse samples for Cr K-edge. The blue dots are the acquired data, while the black line (smooth) is the linear combination fit to the data; the two pecked lines are reference spectra (fine: Cr2(SO4)3 is 49% and Cr2O3 is 51%; coarse: 58% and 42%, respectively) and the grey dots are the residuals.
The determined relative concentration (or abundance) percentages of chromium sulfate and chromium oxide from XANES measurements, the concentration values of Cr and S in the air from EDXRF analyses and the R-factor, which expresses the quality of the fitting for each fine and coarse sample, are reported in Table 1 and Table 2, respectively.
In both fractions it can be noticed that chromium sulfate is a bit more dominant than chromium oxide, except the sample "5 Dec 2016," which has almost equal contributions of sulfate and oxide, and the sample "12 Jan 2017," where the percentage ratio is inverted. PM2.5 samples are comprised between 49% and 70% of chromium sulfate and 30-51% of chromium oxide. In PM10-2.5 samples, chromium sulfate is contained in the range of 36-93% and Cr oxide between 7% and 64%. There was no clear correlation observed between speciation abundances and concentration values of Cr and S. The main reason for this might rely on secondary organic aerosol (SOA) and secondary inorganic aerosol (SIA) concentrations in the sample. SOA and SIA are emitted or formed by the oxidation of gas-phase species in the atmosphere (De Gouw and Jimenez, 2009).
The majority (60-70%) of chromium sources are of anthropogenic origins such as fuel combustions, metallurgical industries, refractory brick production, waste incineration and Crcontaining chemicals (Seigneur and Constantinou, 1995;Werner et al., 2007;Catrambone et al., 2013). The remaining 30% comes from natural sources like crustal matter and volcanic eruptions  (Catrambone et al., 2013). The fact that smaller grain sizes of atmospheric particles are mainly generated from anthropogenic sources, as well as the coexistence of SO4 2− , may indicate that chromium sulfate is the main Cr compound of PM2.5 samples. Our previous research has shown that the common primary emission sources of Ulaanbaatar's atmospheric PM are traffic, coal combustion and soil (Gunchin et al., 2019). Due to this fact, chromium sulfate may correspond to anthropogenic combustion sources. One of the main occurrences of Cr is the trivalent oxidation state in forms of chromium oxides, such as Cr2O3. This compound is present in both measured fractioned samples. These results are similar to those reported in the study of Wang et al. (2007), which indicated that the main chemical component of Cr in PM was chromite and oxides of trivalent chromium, close to iron and steel industrial area. However, it should be noted that due to geographical and climate differences, the results of these works cannot be compared directly. It is worth to mention that Cr(VI) content, which has high health concern, was not present in the measured samples.

Speciation of Zinc
Four Zn compounds were identified in the measured samples: zinc sulfate, zinc silicate, zinc chloride and zinc oxalate. Fig. 4 shows the normalized Zn XANES spectra of the measured fine (a) and coarse (b) samples, together with those of the reference compounds. Representative XANES spectra of Zn K-edge from the LFC analysis for fine (a) and coarse (b) sample are illustrated in Fig. 5.
The determined relative concentration (or abundance) percentages of zinc sulfate, zinc silicate and zinc oxalate from XANES measurements; the concentration values of Zn, S and Si in the air from EDXRF analyses; and the R-factor, which expresses the quality of the fitting for each fine and coarse sample, are reported in Table 3 and Table 4, respectively.
In the fine fraction, zinc sulfate and zinc silicate were identified as the major constituents of all measured samples and zinc oxalate accounted for the rest contribution. Sulfur is often used as a tracer for coal burning in source apportionment studies (Davy et al., 2011) and SO2 is produced during combustion processes (Ohta et al., 2006), leading to formation of secondary sulfate particles mainly present in the fine fraction. In Wang et al. (2007), the authors report that Zn exists mainly as ZnSO4 in PM2.5 samples from both industrial and agricultural areas of Shanghai city. Therefore, Zn in the form of sulfate can originate either directly from a local anthropogenic emission as ZnSO4 or from Zn originally emitted in other forms that condenses to secondary sulfate particles. Mattielli et al. (2009) reported that zinc is present as a sulfide (ZnS) or carbonate (ZnCO3) or as minor constituent of silicate minerals in the natural environment, so that it can be assumed that   (smooth) is the linear combination fit to the data; the two pecked lines are reference spectra (fine: ZnSO4 is 46%, Zn2SiO4 is 42% and ZnCl2 is 12%; coarse: 37%, 51%, and 12%, respectively) and the grey dots are the residuals. zinc silicate could be related to crustal sources, or dust resuspension due to traffic. According to the study of Furukawa and Takahashi (2011), zinc oxalate can be the main oxalate species considering its high concentration in aerosols and the presence of this compound particularly in finer particles. Zinc oxalate is an important component of secondary organic aerosol and its formation in the atmosphere can be obtained in several different ways. However, the formation of aqueous phase on the particle surface is dominant; therefore, Zn oxalate is found to be mainly present in the fine fractions.
In the coarse fraction, Zn is contained mostly as silicates which may be a clear indication of crustal source. The second identified constituent in the measured samples was zinc sulfate. Sulfate forms of Zn are usually pronounced and characteristic in winter season. Since all measured samples were taken in mid-winter months, this compound could also be referred to the combustion source. The residual contribution is zinc chloride. In their study, Furukawa and Takahashi (2011) also show that Zn chloride was observed only in the coarser particles, having no relation with the seasons. Nevertheless, Zn chloride (along with Zn silicate) was found as significant component of Zn in fly ash particles sampled directly from the stack of coal and municipal solid waste (Cai et al., 2015;Yu et al., 2009). Origins and formulation of this compound in atmospheric particles can be related to anthropogenic sources. There was finally no clear correlation between speciation abundances and concentration values of Zn, S, Si and Cl.

SUMMARY
Using a small incidence angle for the synchrotron radiation, we performed XANES spectroscopy on size-fractionated atmospheric particulate matter (viz., PM2.5 and PM2.5-10) collected on polycarbonate filters in the city of Ulaanbaatar, Mongolia. We characterized the chemical states of the Cr and Zn in the PM2.5 and PM2.5-10 and selected suitable PM samples for the XANES spectroscopy through EDXRF analysis. Our results revealed that Cr mainly existed as Cr2(SO4)3 and Cr2O3 in both the fine and the coarse fraction. However, the abundances of these constituents varied and showed no direct relation with the concentrations of the airborne Cr or S. Additionally, we only found Zn in its divalent chemical forms. Although both PM fractions contained abundances of sulfate (ZnSO4) and silicate (Zn2SiO4), only the fine mode contained Zn oxalate (ZnC2O4), whereas only the coarse mode contained Zn chloride (ZnCl2). As with Cr, the relative concentrations of these Zn species exhibited no correlations with the levels of the airborne Zn, S or Si. Finally, anthropogenic sources, such as combustion products and traffic-related resuspended road dust, primarily contributed the Cr and Zn species detected in our PM samples.
These findings shed light on the physiochemical characteristics and potential sources of metallic constituents in atmospheric fine-and coarse-mode PM, thereby supporting future studies on mixture toxicity and synergistic effects that, in contrast to the current efforts, are not based solely on the total elemental concentrations.

ACKNOWLEDGEMENT
This project was funded by Mongolian Foundation for Science and Technology and supported in part by the International Atomic Energy Agency (IAEA) through the Regional Cooperation Agreement Program RAS/7/023. Moreover, this work has been supported through the joint IAEA and the International Centre for Theoretical Physics (ICTP) Sandwich Training Educational Program (STEP) fellowship C6/MON/15001 granted to Gerelmaa Gunchin. Ms. Gunchin expresses her gratitude to IAEA's former officials Dr. Andreas Karydas and Dr. Iain Darby, XRF end station staffs of Elettra as well as XRF group of Atominstitut, TU Wien, for their support and guidance throughout the beamtime experiment and data evaluations.