Determination of settled dust sources by analytical techniques and chemical mass balance receptor model

The identification of sources that produce particulate atmospheric matter (PM) can be of paramount importance for the reduction of air pollution and the development of environmental policies. In order to identify the environmental impact resulting from industrial metallurgical activities in the Metropolitan Region of Vitória, ES, Brazil, it was investigated the contribution to PM that result from industrial activities and from local natural sources. For this purpose, analytical techniques were used to identify the apportionment of sources that contribute to the formation of insoluble settled dust collected at two points near the city of Vitória. Samples of soil, iron ore, limestone, coal, iron ore pellets, sinter, coke, slag, environmental samples of settled dust, and samples representative of the actual flows of materials used in an integrated steel mill were analyzed. Physicochemical characterizations, based on X-ray diffraction and 57Fe Mössbauer spectroscopy of ferruginous compounds found in sources and receptor samples, revealed the presence of highly crystallized hematite and low crystallized hematite. The latter is primarily found in soil samples, while well-crystallized hematite is found in natural samples from iron ores or after thermo-chemical processes applied during the industrial transformation of raw materials, as it happens during the production of pellets. Ferrous crystallographic forms α-FeOOH and Fe5HO8·4H2O, observed in environmental and soil samples, were also found in samples from industrial sources. Source apportionment of carbon based on the IMPROVE_A protocol for thermal/optical carbon analysis showed the participation of the elementary carbon fractions, separating contributions originated from coke and coal sources in the environmental samples. These results allowed a significant reduction of collinearity between source profiles in the application of the chemical mass balance receptor model “EPA-CMB8.2” receptor model. Consequently, it was possible to distinguish sources that process mainly ferrous and carbonaceous materials, identifying the contribution of different sources to the settled dust collected.


Introduction
The identification of sources that contribute to the chemical composition of particulate matter (PM) in environmental samples in urban areas close to industrial regions can be difficult. This is mainly due to the diversity of contributors, each having a particular range of influence, the different aerodynamic diameters of the particles they produce, and the possible similarities in the chemical compositions of the PM they produce. Such PM becomes settled dust that directly impacts the quality of life of the population that resides at distances less than ~ 10 km from the source where it is produced. Although there are studies described in the literature using chemical composition to characterize atmospheric pollutants (Feng et al. 2009), particulate materials from industrial sources are least documented (Santos et al. 2017;Galvao et al. 2018 and, limiting the understanding of their environmental impact. In the study region, there are two industrial sources which use iron ore and high carbon containing materials as the main inputs, which make it difficult to identify the contribution of each of them to the environmental PM. The integrated steel industry processes and pelletizing plants generally have a series of activities and steps that range from the mining process to the production of steel with PM emissions with specific physicochemical characteristics for each step. For example, in studies carried out on industrial complexes (Sylvestre et al. 2017, Beddows Roy and Harrison 2018, Jia et al. 2018, compositional analyses of dust emitted by steelmaking processes show the existence of organic carbon, elemental carbon, and metals such as iron, aluminum, manganese, zinc, and calcium. An application that can be used to identify the contribution of sources to particulate matter in industrial and urban environments is the association of experimental results with receptor models (Taiwo et al. 2014;Guo et al. 2009), such as chemical mass balance (CMB), positive matrix factoring (PMF), and/or principal component analysis (PCA) models. These receptor models have been mainly applied to fine particles. Some studies have combined two or more of these receptors' models, which provide useful information about the uncertainties associated with different models (Lee et al. 2008). Solci and Queiroz (2014)sed the CMB and the PMF to evaluate the chemical compositions of settled dust samples collected in the same place during 104 months. Their results show consistency between the PMF and CMB methods, indicating that the proposed methodology is useful for separation of sources and possible collinearities.
Following these concepts, the chemical mass balance receptor model (USEPA 2004) was chosen to evaluate the contribution of different sources in the environmental samples collected at the two points of study of this work due to the restrictions in the number of samples obtained. However, the need to collect enough material for all analyses and the monthly monitoring base established, it is not possible to apply PCA or other statistical forms of evaluation of the behavior of the chemical mass balance, given the resulting small number of samples. In order to reduce the collinearity of the source profiles and obtain the linear system solution that describes the receptor model, it was required to study the structural, magnetic, hyperfine, and compositional properties of soil samples, industrial inputs (ore and coal), industrial products (pellets, coke, sinter, and slag), and a representative sample of the real flows of materials used in an integrated steel industry were investigated, as well as environmental samples of settled dust using 57 Fe Mössbauer spectroscopy (MS), X-ray diffraction (XRD), proton beam-induced X-ray emission (PIXE), and thermal optical transmittance (TOT) under the Interagency Protected Visual Environment Monitoring (IMPROVE) Thermal Evolution Protocol.
Iron is the element that brings complexity to the analysis, as it is naturally present in soils, it was found that it is possible to distinguish particles containing iron oxides originated from iron ore mining activities, from those of natural origin (soil dust particles) based on Mössbauer spectroscopy measurements (Tavares et al. 2017). The analysis of the properties of ferruginous compounds associated with the techniques used proves to be viable for evaluating the compounds present in the settled dust collected in the region using the CMB, resulting in a tool capable of identifying the contribution of the sources in this pollutant. Thus, it can be an additional tool for the selection and adoption of actions that aim to reduce the perception of discomfort by the population closest to the industrial region.

Sampling
The following types of samples were used in this work, which have been labeled as shown in Table 1: • Industrial inputs and products-samples collected in the handling area of the integrated steel industry composed of iron ore, pellets, slag, coke, coal, ferroalloys, sinter, and limestone. • Soil-surface samples from exposed soils, located within the region of interest. This type of sources is characterized by the potential for resuspension by wind action and possible subsequent deposition. • Sea-samples collected in the bay around point 2 (IB) (Fig. 1), obtained by spraying seawater on the surface in pure quartz filters. • Environmental samples-samples called settled dust was collected at points 1 and 2.

Collection points of environmental samples
The collection points selected for this study are located in two neighbor cities in the Metropolitan Region of Greater Vitória (RMGV), Serra and Vitoria, respectively. They are predominantly urban areas, with a total population estimated at 893 thousand inhabitants, which represents ~ 45% of the total population of the Espírito Santo region (IBGE 2020). The main potentially polluting activities in Espírito Santo are concentrated in these two municipalities. They involve industrial processes, civil construction, airport and port operations, waste landfills, traffic routes, among others (IEMA 2020). Also, the two biggest industries of the region, dedicated to the transformation of iron ore, are inserted in these two cities. Both of them demand significant movement of iron-containing materials, typically over 70 million tons per year (ANTAQ 2020). Additionally, three factors led to the selection of the locations of this study: (i) the concentration of industrial complexes; (ii) the presence of a significant number of individuals exposed to atmospheric pollution, especially to particulate matter; (iii) the existence of representative meteorological data. For these reasons, environmental samples were collected at point 1 (PI, northeast) and point 2 (IB, southwest) as indicated in Fig. 1, having the geographic coordinates described in Table 2. The samples were labeled to identify the collection point according to the code described below. For example, the PI1Q18 code indicates that the sample was collected at point 1 during the first quarter of 2018.

Settled dust collector
Since this study aims to determine sources of non-soluble settled dust, samples were collected using a fiberglass pot collector containing a glass bottle at the bottom and a support to retain the coarse particles. This passive collector, shown in Fig. 1, was specifically designed to retain this type of particles from the flow of PM in any climatic conditions. The interception of PM occurs at 1.1 m above the closest surface considered as base, regardless the collector position, following the ASTM D1739-98 standard test method for collection. The materials deposited in the collector are used to obtain the settled dust samples. This requires the removal of the material from the support using wet method a nonaggressive for the composition of the materials that allows to remove all coarse organic materials such as pieces of leaves, branches, and insects are removed. Each sample was formed by the sum of the monthly collected samples, forming quarterly composite samples. The period of collection was from January 2018 to December 2019. This procedure was necessary in order to enough mass of material to perform all the analytical techniques used in this work.

Chemical profile of the integrated steel plant
In order to apply the chemical mass balance receptor model EPA-CMB8.2 (USEPA 2004), it is required to create specific chemical profiles of the most relevant emission sources in the region studied. Hence, it was created a chemical profile of integrated steel plant, one of the important emission sources in the region. This chemical profile represents all the atmospheric emissions of coarse particles arising from all the extensive, diffusive, and fugitive sources involved in each of the manufacturing processes of the steel industry. For this purpose, it was required to consider all materials Settled dust Environmental sample collected in the northeast, point 1 CP-AMT Chemical profile integrated steel Specific chemical profile of materials handled in the integrated steel industry composed of iron ore, limestone, coal, coke, and slag ID Industrial dust Fines collected from the dedusting systems of the integrated steel industry composed of iron ore, limestone, coal, and coke Sinter Iron ore sinter Mixtures of fine ores, residues, and additives processed in the sintering process Coke Metallurgical coke Produced from coal through dry distillation in a coke oven and has better physical and chemical characteristics than coal Coal Mineral coal Coal is a combustible black or brownish sedimentary rock, formed as strata of rock called coal seams Pellet Iron ore pellet Pellets are formed from < 0.05 mm fine ore and additives, into 9-16 mm spheres using very high temperatures in the pelletizing process Iron ore Iron ore Iron ore is the raw material used for the manufacture of pig iron, which is one of the main raw materials for the manufacture of steel Fe-Cr-Mn Iron alloy Cr-Mn An iron alloy containing a percentage of chromium and manganese to form a steel alloy ESAF AF slag Slag produced in the iron reduction process in the blast furnace ESKR KR slag Slag produced in the KR desulfurization process of liquid pig iron ESLD LD slag Slag from the process of converting pig iron into steel in the steel basic oxygen furnace Sludge Steel sludge Fine solids from the steel gas wet cleaning process Vent2 Dust from the dedusting system Sample of secondary dedusting systems in steel basic oxygen furnaces IS2 Industrial sources 2 Blending iron ore and iron ore pellets from other industrial sources Sea Sea Spraying sea water on the surface of pure quartz filters manipulated during a typical year and the participation of each material, in each process. The representative samples obtained from the flows of materials used in the steel production process was labeled "CP-AMT." The characterization and analysis of a sample having mass composition representative of all materials involved in the steel industry processes allowed to obtain such representative profile. The mass composition of such sample is presented in Table 3. As well as CP-AMT, there are other industries in the surroundings of the region of study that also use iron, pellets, limestone, and coal. Those industries were identified as source IS2. Following the EPA-CMB8.2 receptor model, the chemical profiles of the environmental samples (receptors) are   written as linear combinations of the chemical profile of the main sources (CP-AMT, soils, sea, iron ore, pellet, coke, coal, ESLD, ESKR). This procedure allows to estimate the source contribution to the environmental sample by solving a linear system of equations (Belis et al. 2013) which expresses each environmental chemical profile as a linear sum of the chemical abundances within each source and its apportionment to the total mass of the environmental profile. Note that samples of marine aerosols used to obtain the chemical profile labeled as "Sea" source were obtained by spraying sea water on the surface of pure quartz filters. Marine water samples were collected in the bay around point 2 (IB) (Fig. 1). In order to identify the possible sources that would contribute to each environmental sample, several characterization techniques were used. XRD was used to identify the crystallographic phases (Dinnebier et al. 2008;Klug and Alexander 1974;Rompalski et al. 2019). MS were applied to characterize the iron phases. Trace elements were studied by PIXE (Ishii 2019), being considered in the balance those in concentrations greater than 10 ppm. The amounts of organic (OC) and elemental (EC) carbon contained in the samples were obtained by TOT, following the thermal protocol IMPROVE_A (DRI Model 2001). These analytical techniques allow to differentiate specific characteristics of each source that aid to eliminate possible strong collinearities between them, and therefore differentiate the origins, either natural or industrial, of the PM samples studied.

Results and discussion
The results obtained were based on the qualitative analysis of samples obtained from settled material at short distances. It was not the aim of this study to quantify the amount of settled dust deposited during period of study. However, the local environmental agency, IEMA, is responsible for this task, having several collection stations in the region of study. Data available of the deposition rate of settled dust in the region located close to points 1 and 2 in this study are named as "PMPS -RGV2" and "PMPS -RGV10" respectively by IEMA are shown in Table 4 (IEMA 2018 and 2019).
Data shown in Table 4 shows values ~ 50% below the current standard of 14 g/m 2 ·30 days for this pollutant according to State Decree No. 3.463R/2013. The largest variations observed are due to change in wind direction (IEMA 2019). Note that it is not the object of this study to evaluate the amount of matter deposited in the region, but to carry out an evaluation of the contributions of the sources in the non-soluble settled particles.
In order to refine the analysis of hematite with low crystallinity (α-Fe 2 O 3 *) in the outputs of the receptor model, the Vent2 source was added because expressive amounts of α-Fe 2 O 3 * were observed in its composition. As this phase is also present in soil samples, for the purpose of identifying sources, it is more advantageous to isolate Vent2 from the CP-AMT profile (Tables 6 and 9).

Proton beam-induced X-ray emission
PIXE measurements of PI and IB environmental samples show that Fe, Si, Al, and Ca are the predominant elements present in their composition (Fig. 2). Secondary elements such as P, S, Cl, K, Ti, Na, Cu, Mn, and Mg were observed in concentration above 0.1% and below 1.0%. Minority elements (Sc, Cr, Cu, Zn, Ga, As, Br, Rb, Sr, and Zr) were found in concentrations between 10 and 900 ppm. It is also noticeable that there is a variation along the year both in IB and PI, which can be related to seasonal climate variations, such as wind velocity and direction, pluviosity, and temperature.
IB samples contain larger percentage of Fe and Si, and less Ca than PI samples. During the first quarter of the year, IB samples exhibit the largest content of Fe, while PI samples exhibit the lowest. This fact may be related to the wind being predominantly in the northeast direction during this period of the year, together with low pluviosity and increased solar radiation.
Most of the elements observed in Fig. 2 are associated with oxides. Therefore, the amount of oxygen in the samples is high. There is also a participation of nitrogen and hydrogen in the compounds that are present in the major elements.  (Chow et al. 2007). The carbon fraction uncertainty is 5%. The greatest contribution from coal is evidenced by OC3, EC1, and EC2 fractions. The coke source is more represented by EC2 and EC3. CP-AMT source is represented by the EC2 and EC3 fractions (from coke production), but the possible combination with OC3, EC1, and EC2, from the coal input, must be considered (Figs. 3 and 4). Figure 3 presents the most important results of OC and EC obtained by TOT. Only those fractions greater than 0.1 wt% are shown. The mass percentage OC and EC carbon content for all main source samples considered in this study is shown as Table 5. Note that source VENT2 results from secondary dedusting systems used in the basic oxygen furnaces at the steel factory of the integrated steel industry. This source was not included as a component of the chemical profile shown in Table 3.
The environmental samples studied show a greater participation of fractions EC1 and EC2, being larger for those collected in point (PI) than for those collected in point (IB). Along 2018 and 2019, the observed trends for each quarter are similar (Fig. 4). When they are compared to carbon apportionment of the sources, there is a participation of coal, coke, and CP-AMT sources. Table 6 summarizes the main results obtained by PIXE and TOT corresponding to the wt% element composition of all sources considered. The wt% of organic carbon and elemental carbon (total OC and total EC) correspond to the total fractions obtained by TOT for each sample, independent of subfractions evolved during the analyses. Such information is used to build a chemical profile for each source, together with XRD and MS characterization, which are used to enhance the chemical profiles for sources and environmental samples.
Similar crystalline phases and proportions were observed for 2018 and 2019 samples. Given that XRD is not able to   57 Fe Mössbauer spectroscopy is a highly selective technique, which can detect very subtle changes in the chemical environment of the 57 Fe atom, allowing to obtain information regarding the various Fe compounds contained in the sample, non-equivalent Fe crystallographic sites, oxidation states, chemical bonds, and magnetic properties, among others. Besides, the high energy ~ 10 6 eV and the narrow line width of gamma radiation energy allow an extremely high energy resolution, of few parts in 10 11 (Gonser 1975).
These features make Mössbauer spectroscopy an essential non-destructive technique, since it allows the distinction between the iron oxides present in soil, iron ore, and PM samples, enabling Fe to be used as a tracer element to assess the contribution of each source (Tavares et al. 2017). The Mössbauer spectra were obtained in transmission mode at room temperature and at 25 K, using a conventional spectrometer, with constant acceleration and 57 Co source embedded in a Rh matrix. Isomer shift values quoted are relative to metallic α-Fe. The experimental data, represented by dots and black line, and fitted subspectra are shown in Figs. 6 and 7. The hyperfine parameters used to fit the data are shown in Tables 7 and 8. Mössbauer spectra measured at room temperature (RT) were fitted with two sets of subspectra: (i) two doublets to account for the central lines in the spectrum and (ii) one sextet that account for the remaining six lines. The first set of subspectra was attributed to iron silicates (Si-Fe) and iron oxides corresponding to very small particles, having mean diameter in the nanometer range, that exhibit superparamagnetic behavior at RT, and the sextet subspectrum was attributed to magnetic iron oxides which may exhibit antiferromagnetic (AF) or weekly ferromagnetic (WF) behavior (Murad and Schwertmann 1986;Greenwood and Gibb 1971). In order to avoid superparamagnetic behavior of the particles at room temperature and improve the resolution of the hyperfine magnetic parameters, low temperature (25 K) measurements were required. As the temperature is reduced, the superparamagnetic particles are blocked, they are less susceptible to thermal disorder, and tend to order magnetically. This allows to obtain information about the different phases existing in the material studied. The spectra measured at 25 K were fitted with a doublet subspectra, associated to iron silicate minerals ([(Ca,Fe 2+ ) 3 (Al,Fe 3+ ) 2 (SiO 4 ) 3 ]) and a set of four subspectra associated to soil hematite (α-Fe 2 O 3 *), iron ore (α-Fe 2 O 3 ) (Tavares et al. 2017), goethite (α-FeOOH), and ferrihydrite (Fe 5 HO 8 ·4H 2 O) (Tavares et al. 2017;Murad and Schwertmann 1986;Greenwood andGibb 1971, Murad andSchwertmann 1980).
The identification of soil hematite and iron ore hematite is based in the Morin transition (TM) that occurs at TM ≈ 250 K for hematite (Morin 1950;Özdemir et al. 2008). Below TM, spins lie along the trigonal or c-axis of the rhombohedral lattice, forming an AF spin structure. This AF ordering changes to weakly ferromagnetic (WF) on heating above TM, as the spins flip 90° to the basal c plane and arrange out of exact antiparallelism, by a fraction of a degree (Özdemir et al. 2008). Such canted arrangement results in WF behavior. When hematite experiences such transition upon cooling below TM, there is a change of the algebraic sign of the quadrupolar displacement, from negative value at RT, i.e., above TM, to a positive value below TM (e.g., at 25 K). Therefore, at room temperature, quadrupolar displacement values (2ξq) are typically negative for well-crystallized hematite, as found in iron ore. When the hematite does not undergo the Morin transition, the quadrupolar shift remains negative below TM. This is typically observed in soil hematite (Tavares et al. 2017). Hence, in the case of iron-rich samples, the quadrupolar displacement below TM can be taken as a primary signature pointing to the origin of the hematite contained in the sample studied, and therefore to a possible source of PM contained in that sample.
The relative area under the curve for each subspectrum to the total spectrum area is proportional to the apportionment of the corresponding material. Following this, it can be inferred from Fig. 8 the presence of both types of hematite in the environmental samples. This is a very important result, since it reveals that the sources of emission contributing to the PM collected in those samples are both from natural origin (e.g., soils, rocks, sands) as well as of anthropic origins (e.g., industries). Another interesting observation is that the hematite content has significant variations depending on the quarter/year and the region where the samples were collected.
PI samples collected in 2018 show a significant increase in the spectral area of Fe 3+ , iron silicate mineral (Ca,Fe 2+ ) 3 (Al,Fe 3+ ) 2 (SiO 4 ) 3 phases. Regarding 2019 samples, no major variations in the relative spectral area were observed compared to 2018. However, it was observed an increase in the goethite (α-FeOOH) spectral area for IB samples, while in the case of PI samples, it was found a ~ 40% reduction of LC-hematite (α-Fe 2 O 3 *). These results can be explained considering the variations in wind speed and direction, temperature increase, and the incidence of typical rains from October to March in the studied region, affecting the contribution from each source to the PI and IB samples. In addition, there is a growing trend of environmental controls of industrial activities located in the region, leading to lower emissions of PM.

Assessment of anthropogenic and natural contributions to the hematite content in the samples
Based on PIXE quantification of iron apportionment and the proportions of anthropogenic iron oxides (having HC-hematite) and natural soil (LC-hematite content) obtained by MS, an estimate of the anthropogenic and natural contributions to the hematite contained in environmental PM samples is calculated using Eq. 1.
where wt% total corresponds to the mass content of iron oxides in the sample either of natural or industrial origin; wt% Fe corresponds to the Fe mass in the sample determined by PIXE; and % area corresponds to the relative spectral area of the Fe phases (HC and LC hematite) determined by MS. Figure 9 shows the contribution of anthropogenic hematite (α-Fe 2 O 3 ) and soil hematite (α-Fe 2 O 3 *) in the environmental samples studied. Both PI and IB samples exhibit hematite content from both origins being the contribution (1) wt% total = wt%Fe × %area from iron ore larger than that from soil. IB samples exhibit a much larger contribution from anthropogenic origin (HChematite from iron ore) than PI samples. It can be observed that the contribution from soil hematite varies along 2018 and 2019 following similar trends in both collection points, although PI samples exhibits lower apportionments of both iron ore and soil hematite than IB samples.

Application of chemical mass balance receptor model
The main results of the analytical techniques presented above were used as input data for the EPA-CMB8.2 receptor model (data shown in Fig. 2 and Tables 5 and 9). Some of the elements are present in several sources, such as Fe, Al, Ca, OC, and EC. However, the quantification of EC, OC, highly crystallized, and low crystallized hematite allow to significantly reduce the collinearity between sources.
Emissions from VENT2 source were found to contain a significant apportionment of α-Fe 2 O 3 *, which initially could be related only to soils samples. For this reason, VENT2 source was inserted in the CMB model as part of the CP-AMT. α-Fe 2 O 3 * was also found in the ESAF and ESKR sources, most likely due to an isomorphic substitution of Fe 3+ in hematite and goethite by another cation, mainly Al, under high process temperatures, reducing or suppressing the Morin transition temperature (Murad and Schwertmann 1986;Özdemir et al. 2008).
The presence of Fe-Ca-Si in the environmental PM samples, identified in the MS spectra as (Ca, Fe 2+ )3(Al, Fe 3+ )2(SiO 4 )3, was not balanced by the chemical profile of industrial sources (CP-AMT and IS2), when the EPA-CMB8.2 receptor model was used. This could be explained by a possible variation of this phase in the composition Fig. 8 Relative spectral area of Fe phases obtained from Mössbauer spectra measured at 25 K for the environmental samples studied. Note: LC, low crystallinity; HC, high crystallinity of CP-AMT source, or by the existence of an additional unknown source. The modeling did not accept a Fe-Si-Ca virtual source as an artificial input to balance the phase, indicating that a mixture of Fe-Ca-Si with other phases and chemical elements would be required to improve the solutions.
The results obtained from CMB calculations yield good statistical values assuring that the solutions obtained are reasonably accurate (Tables 10 and 11). T-statistic values were used to determine the statistical significance of the calculation (USEPA 2004The lowest T values obtained, "T min ," are shown in Table 8. Notice that T > 5 means that the source Fig. 9 Weight percentage of HC-hematite ((α-Fe 2 O 3 ) gray bars) related to iron ore, and LC-hematite (α-Fe 2 O 3 *) related to natural soil contribution estimate (SCE) has a relative uncertainty of less than 20%. The contribution of marine aerosol in the region is subject to the effects of the sea breeze in the early hours of afternoons and effects due to natural convection in the central area of Vitória Island. The microcirculation in the region is induced by turbulence generated predominantly by buoyancy. Also, as the contributions are relative shares, for most of the year, under the predominance of the northeast wind, industrial sources affect with lesser intensity the PI sampling point leading to more participation of sea source. Particulate contributions from combustion processes are predominantly fine (less than 2.5 µm) and do not sediment at such short distances, making their contribution to settled dust unlikely. Emissions from different sources with different carbon compositions (e.g., coal and coke) reach sampling points differently throughout the year due to changing wind directions and relative distances. Figure 10 shows variations in the balance of the contributions of dust sources present in the region when environmental samples collected in 2018 and 2019 are compared. It is noticeable that the evolution of the apportionment of industrial sources can be tracked down to increases or reductions in their activities. For instance, a major change that occurred in the industrial source (IS2) and observed as a reduction of its contribution in the IB region between the first and third quarter of 2019 (IB1Q19 and IB3Q19) correlated with the reduction of the iron ore supply chain during this period. Figure 10 also reveals an asymmetry in the contributions of CP-AMT and IS2 to IB and PI samples, reflecting the fact that IB is closer to IS2 than to integrated steel industry, and the opposite for PI. During the third quarter of 2019, it is observed a striking contribution of integrated steel industry to the source apportionment of that sample. The contribution of CP-AMT to the settled dust collected in PI during that period has been found to be due to activities involving the relocation of granular materials that took place in CP-AMT yards nearby.
These results show the validity of the application of the EPA-CMB8.2 receptor model, and the effective reduction of source collinearity obtained by Mössbauer spectroscopy, PIXE, and TOT. Such experimental procedure, together with an extensive characterization of the sources to obtain detailed chemical profiles, should help each industrial company to develop effective actions into the appropriate processing activities in order to achieve their environmental objectives. Moreover, it may be used as a systematic monitoring method applied to the prevention of pollution, as well as aid to develop policies towards those goals. As a result of this study, integrated steel industry has taken actions to improve environmental control methods and procedures involving granular materials handling operations (e.g., slag, coal, coke), aiming to reduce the impact of this activity on the air quality.

Conclusions
The establishment of reproducible quantitative methods to obtain the source contributions to atmospheric PM can be of great importance for the reduction of air pollution and the development of environmental policies. The application of the chemical mass balance receptor model EPA-CMB8.2 to the study of deposited atmospheric PM can be challenging in regions in the domain of influence of industries that transform ferruginous materials, due to the predominance of Fe in its composition. It is therefore necessary to identify the Fe phases present in sources and receptors, in order to reduce uncertainties in their chemical profiles and address the origins of PM deposited as settled dust. Physicochemical characterizations employed in this work, based on X-ray diffraction and 57 Fe Mössbauer spectroscopy of ferruginous compounds found in sources and receptor samples, revealed the presence of HC-hematite (α-Fe 2 O 3 ) and LC-hematite (α-Fe 2 O 3 *). The latter is primarily found in soil samples, while well-crystallized hematite is found in natural samples from iron ores and after thermo-chemical processes applied during the industrial transformation of raw materials, as it happens during the production of pellets. Nevertheless, ferrous crystallographic forms α-FeOOH and Fe 5 HO 8 ·4H 2 O, observed in environmental and soil samples, were also found in samples from industrial sources. Similarly, source apportionment of carbon based on the IMPROVE_A protocol for thermal/optical carbon analysis showed the participation of the elementary carbon fractions, separating the contribution of the coke source from the contribution of the coal source in the environmental samples. This approach was applied to determine the origins and the apportionment of the sources in environmental PM samples collected in IB (Vitória) and PI (Serra) sites, both in the surroundings of the city of Vitória (ES, Brazil). The combination of all analysis techniques allowed to significantly reduce the collinearity of the source profiles in the application of CMB. Potentially collinear sources were separated by well resolved split profiles based on the participation of iron in its various phases. Also, the splitting of carbon participation into several organic and elementary mass fractions generated a substantial improvement in the application of mass balance calculations. The use of these expanded compositions allowed the separation of sources that mainly process ferrous and carbonaceous materials. It also generated enough sensitivity in the model to distinguish variations in emission rates due to production situations altered by changes in the supply of raw material in one of the largest production units in the region and by weather conditions. This analytical approach can be used to obtain well resolved split profiles of sources and environmental samples that can be used to feed models such as chemical mass balance (CMB), positive matrix factorization (PMF), and principal component analysis (PCA), allowing to identify sources that may contribute to environmental PM in regions where there is a predominance of steel making or similar processes.