Aqueous solubility of Cr ( VI ) compounds in ferrochrome bag filter dust and the implications thereof

Since its discovery in 1798, chromite has remained the only commercially viable source of new chromium (Cr) units (Murthy et al. 2011; Riekkola-Vanhanen, 1999). Approximately 90% to 95% of mined chromite is used in the metallurgical industry for the production of various grades of FeCr, which is produced through the carbo-thermic reduction of chromite ore (Riekkola-Vanhanen, 1999). About 80% to 90% of ferrochrome (FeCr) is consumed by the stainless-steel industry, primarily as high-carbon or charge-grade FeCr (Murthy et al. 2011). Stainless steel is a vital alloy in various applications (Gasik, 2013), with the Cr content of stainless steel increasing its corrosion resistance. South Africa holds the largest chromite deposits and is the second-largest FeCr producing country (ICDA, 2012), with 14 FeCr smelters (Beukes et al., 2012). Cr is present as Cr(III) in chromite ore, while Cr(0) occurs in the produced FeCr. Although completely unintended, small amounts of Cr(VI) are formed during FeCr production, which may be present in waste materials (Beukes et al., 2010). Cr(VI) is generally regarded as carcinogenic (IARC, 2012), with specifically airborne exposure to Cr(VI) being associated with cancer of the respiratory system (Proctor et al., 2002). In contrast to Cr(VI), Cr(III) and Cr(0) are not classified as carcinogenic. Cr(III) is in fact used a dietary supplement for certain human health abnormalities (Hininger et al., 2007). The main types of waste generated during FeCr production are slag, sludge from wet venturi scrubbers and bag filter dust (BFD) (Van Staden et al., 2014). Slag-to-FeCr generation ratios of 1.1:1 up to 1.9:1 are common in this industry, with ratios varying according to different production technologies employed (Beukes et al., 2010; Niemelä and Kauppi, 2007). Although the volumes of slag generated are large, the Cr(VI) content of the slag is usually very low (Beukes et al., 2010; Niemelä and Kauppi, 2007). Additionally, FeCr slag can be used in various commercial applications (Niemelä and Kauppi, 2007; Zelić, 2005; Lind et al., 2001; Riekkola-Vanhanen, 1999). Sludge, which is generated during the wet venturi scrubbing of closed furnace off-gas (Beukes et al., 2010; Niemelä et al., 2004) does not usually contain significant Cr(VI) concentrations (Gericke, 1998). However, in contrast to slag and sludge, BFD generated during the cleaning of off-gas from open/semi-closed FeCr furnaces contains significant levels of Cr(VI) (Beukes et al., 2012; Maine et al., 2005; Gericke, 1998), which cannot be disposed without proper Cr(VI) treatment (Beukes et al., 2012). After the BFD is treated to reduce Cr(VI), it is usually disposed in fit-for-purpose waste facilities such as slimes dams. Treatment of BFD (and indeed any Cr(VI) containing FeCr waste) usually involve the aqueous reduction of Cr(VI) with an appropriate inorganic reducing agent such as Fe(II) (Beukes et al., 2012; Seaman et al., 1999; Buerge and Hug, 1997) and S(IV) (Beukes et al., 2001 and 1999), or bacterial reduction (Dhal et al., 2013; Molokwane et al., 2008). Although Cr(VI) can also be reduced with numerous organic compounds (Kassem, 2010; March, 1992), this is usually avoided due to the potential solubilisation of Cr(III) (Beukes et al., 2012; Apte et al., 2006). In this paper, the leaching potential of Cr(VI) from BFD is investigated and it is proven that this can be a major limitation in the effectiveness of the conventional Cr(VI) treatment used by industry. *To whom all correspondence should be addressed. Tel.: +27 (0) 18 299 2337; fax: +27 18 (0) 299 2350; e-mail: paul.beukes@nwu.ac.za Received 7 September 2016; accepted in revised form 27 March 2017 ABSTRACT The production of ferrochrome (FeCr) is a reducing process. However, it is impossible to completely exclude oxygen from all of the high-temperature production process steps, which may lead to unintentional formation of small amounts of Cr(VI). The majority of Cr(VI) is associated with particles found in the off-gas of the high-temperature processes, which are cleaned by means of venturi scrubbers or bag filter dust (BFD) systems. BFD contains the highest concentration of Cr(VI) of all FeCr wastes. In this study, the solubility of Cr(VI) present in BFD was determined by evaluating four different BFD samples. The results indicate that the currently applied Cr(VI) treatment strategies of the FeCr producer (with process water pH ≤ 9) only effectively extract and treat the water-soluble Cr(VI) compounds, which merely represented approximately 31% of the total Cr(VI) present in the BFD samples evaluated. Extended extraction time, within the afore-mentioned pH range, proved futile in extracting sparingly-soluble and water-insoluble Cr(VI) species, which represented approximately 34% and 35% of the total Cr(VI), respectively. Due to the deficiencies of the current treatment strategies, it is highly likely that sparingly water-soluble Cr(VI) compounds will leach from waste storage facilities (e.g. slimes dams) over time. Therefore, it is critical that improved Cr(VI) treatment strategies be formulated, which should be an important future perspective for FeCr producers and researchers alike.


INTRODUCTION
Since its discovery in 1798, chromite has remained the only commercially viable source of new chromium (Cr) units (Murthy et al. 2011;Riekkola-Vanhanen, 1999).Approximately 90% to 95% of mined chromite is used in the metallurgical industry for the production of various grades of FeCr, which is produced through the carbo-thermic reduction of chromite ore (Riekkola-Vanhanen, 1999).About 80% to 90% of ferrochrome (FeCr) is consumed by the stainless-steel industry, primarily as high-carbon or charge-grade FeCr (Murthy et al. 2011).Stainless steel is a vital alloy in various applications (Gasik, 2013), with the Cr content of stainless steel increasing its corrosion resistance.South Africa holds the largest chromite deposits and is the second-largest FeCr producing country (ICDA, 2012), with 14 FeCr smelters (Beukes et al., 2012).
Cr is present as Cr(III) in chromite ore, while Cr(0) occurs in the produced FeCr.Although completely unintended, small amounts of Cr(VI) are formed during FeCr production, which may be present in waste materials (Beukes et al., 2010).Cr(VI) is generally regarded as carcinogenic (IARC, 2012), with specifically airborne exposure to Cr(VI) being associated with cancer of the respiratory system (Proctor et al., 2002).In contrast to Cr(VI), Cr(III) and Cr(0) are not classified as carcinogenic.Cr(III) is in fact used a dietary supplement for certain human health abnormalities (Hininger et al., 2007).
The main types of waste generated during FeCr production are slag, sludge from wet venturi scrubbers and bag filter dust (BFD) (Van Staden et al., 2014).Slag-to-FeCr generation ratios of 1.1:1 up to 1.9:1 are common in this industry, with ratios varying according to different production technologies employed (Beukes et al., 2010;Niemelä and Kauppi, 2007).Although the volumes of slag generated are large, the Cr(VI) content of the slag is usually very low (Beukes et al., 2010;Niemelä and Kauppi, 2007).Additionally, FeCr slag can be used in various commercial applications (Niemelä and Kauppi, 2007;Zelić, 2005;Lind et al., 2001;Riekkola-Vanhanen, 1999).Sludge, which is generated during the wet venturi scrubbing of closed furnace off-gas (Beukes et al., 2010;Niemelä et al., 2004) does not usually contain significant Cr(VI) concentrations (Gericke, 1998).However, in contrast to slag and sludge, BFD generated during the cleaning of off-gas from open/semi-closed FeCr furnaces contains significant levels of Cr(VI) (Beukes et al., 2012;Maine et al., 2005;Gericke, 1998), which cannot be disposed without proper Cr(VI) treatment (Beukes et al., 2012).After the BFD is treated to reduce Cr(VI), it is usually disposed in fit-for-purpose waste facilities such as slimes dams.
Treatment of BFD (and indeed any Cr(VI) containing FeCr waste) usually involve the aqueous reduction of Cr(VI) with an appropriate inorganic reducing agent such as Fe(II) (Beukes et al., 2012;Seaman et al., 1999;Buerge and Hug, 1997) and S(IV) (Beukes et al., 2001 and1999), or bacterial reduction (Dhal et al., 2013;Molokwane et al., 2008).Although Cr(VI) can also be reduced with numerous organic compounds (Kassem, 2010;March, 1992), this is usually avoided due to the potential solubilisation of Cr(III) (Beukes et al., 2012;Apte et al., 2006).In this paper, the leaching potential of Cr(VI) from BFD is investigated and it is proven that this can be a major limitation in the effectiveness of the conventional Cr(VI) treatment used by industry.

Materials
Numerous factors can influence the composition of BFD, including the production technology employed (e.g.open or semi-closed furnace technology, type of filters used in the bag filter plant), physical separation in the bag filter plant itself (e.g.some compartments containing finer material than others), metallurgical operating conditions (e.g.basic or acid slag operating conditions) and composition of feed materials (e.g. chemical and physical differences of ores).Therefore, four different BFD samples were obtained from FeCr producers in South Africa.These producers preferred to remain anonymous and therefore the samples are merely referred to as BFD samples A, B, C and D.
All chemicals used were analytical grade reagents.Standard Cr(VI) solutions were prepared from a 1 009 ± 5 mg•mL -1 aqueous chromate (CrO 4 2-) analytical solution (Spectrascan, distributed by Teknolab AB, Sweden), which were used for calibration and verification of the analytical technique employed.The post-column reagent that was used during Cr(VI) analysis was prepared using 1,5-diphenylcarbazide (DPC) (FLUKA), 98% sulfuric acid (Rochelle Chemicals) and HPLC grade methanol (Ace).Solutions of sodium hydroxide (Merck) and perchloric acid (Merck) were used to adjust the pH of aqueous solutions/ mixtures.According to solubility data, Cr(VI) chemicals can be classified as water soluble, sparingly water soluble and water insoluble (IARC, 2012;Ashley et al., 2003).Ammonium sulphate with 99% purity (Merck SA) and 25% ammonia solution (Associated chemical enterprises) were used to prepare a 0.05 M (NH 4 ) 2 SO 4 -0.05M NH 4 OH extraction buffer that is capable of leaching all water-soluble and sparingly water-soluble Cr(VI) compounds (Ashley et al., 2003).Sodium hydroxide (Promark chemicals) and anhydrous sodium carbonate (Merck SA) were used to prepare a 3 wt% Na 2 CO 3 -2 wt% NaOH extraction buffer, to quantitatively extract all Cr(VI) in the BFD samples, including the water-insoluble fraction (Ashley et al., 2003).Ultra-pure deionized water (resistivity, 18.2 MΩcm -1 ), produced by a Milli-Q water purification system, was used for all dilutions and aqueous extractions.99.999% pure nitrogen gas (N 2 ) (AFROX) was used to provide an inert environment during leaching.Hydrophilic PVDF 0.45 µm filters (Millipore Millex, USA) were used for the filtration of solutions.

BFD sample characterisation
A Malvern Mastersizer 2000 was used to determine the particle size distribution of the BFD samples.A diluted suspension of material was ultra-sonicated for 1 min prior to the particle size measurement, in order to disperse the individual particles and to prevent the use of a chemical dispersant.Laser obscuration was kept at 10% to 15%, while mechanical stirring was set to 2 000 r•min -1 .Ten measurements were made for each sample and the average taken.
Scanning electron microscopy with energy dispersive x-ray detectors (SEM-EDS) was used to perform morphological and chemical characterisation of the BFD particles.FEI Quanta 250 FEG ESEM incorporating Oxford Inca X-Max 20 EDS system with a 15 kV electron beam at a 10 mm working distance was used.BFD samples were prepared with two different procedures for SEM analysis.Firstly, samples were mounted onto a specimen stub with carbon-coated tape and subsequently gold coated in order to determine the general BFD particle characteristics, e.g., size and shape.In order to determine the chemical compositions of the BFD samples by SEM-EDS, the samples were set in resin and polished before being analysed.The SEM-EDS analysis was performed by scanning a wide rectangular area of the sample surface in a raster-like pattern for 60 s.Spot analyses were also performed to determine the composition of individual particles present in the BFD.
X-ray diffraction (XRD) analyses were performed on BFD samples ground to < 45 µm using a Rigaku D/MAX 2500 rotating-anode powder diffractometer with Cu Kα radiation at 50 kV, 260 mA, a step-scan of 0.02°, and a scan rate at 1 min -1 in 2 h from 5° to 70°.Phase identifications were made using JADE v.9.3 coupled with the ICSD and ICDD diffraction databases.

Trace metal analysis
Inductively coupled plasma mass spectroscopy (ICP-MS) (Agilent 7500CE ICP-MS with Octopole Reaction System) was used to determine the total metal concentrations, irrespective of oxidation state, present in aqueous leach solutions.These solutions were obtained as described in the next section.

Cr(VI) extraction and analysis
For each BFD sample, 0.50 g was transferred into a 100 mL glass beaker, which contained 50 mL of deionized water.The initial pH of the deionized water had been adjusted to a predetermined value.Leaching solutions were purged with N 2 for 10 min prior to Cr(VI) extraction to remove dissolved oxygen from the solution, which may cause unwanted Cr(III) oxidation (Ashley et al., 2003).Each beaker and its contents were placed on a magnetic stirrer and agitated for a period of 2 h.A steady stream of N 2 was continually used to purge the suspension, ensuring an inert environment during leaching.After Cr(VI) extraction, the leaching solutions were filtered through hydrophilic Millipore 0.45 µm PVDF filters to remove particulate matter, followed by another pH measurement to ensure that no pH drift occurred.Thereafter the filtrates were transferred to 100 mL A-grade volumetric flasks and diluted to a final volume of 100 mL with pH-adjusted deionized water.50 mL of each solution was used for aqueous trace metal analyses, while the remaining 50 mL was used for Cr(VI) analysis.All pH measurements were conducted with a HANNA -HI 991001 pH meter with a temperature probe.
In addition to Cr(VI) extraction at specific pH values, Cr(VI) extractions were also performed using hot plate digestion.A buffer of 0.05 M (NH 4 ) 2 SO 4 and 0.05 M NH 4 OH was used to leach water-soluble and sparingly water-soluble Cr(VI) compounds, while a buffer of 3 wt% Na 2 CO 3 and 2 wt% NaOH was used to leach total Cr(VI) (Ashley et al., 2003;NIOSH 7605, 2003).The afore-mentioned buffer (Na 2 CO 3 -NaOH) enhances the dissolution of insoluble Cr(VI) compounds at higher pH levels by the addition of carbonate (CO 3 2-), as indicated in Eq. 1.
In the presence of a large excess of CO 3 2-the equilibrium shifts quantitatively to the right of Eq. 1 and insoluble Cr(VI) becomes soluble (Ashley et al., 2003).
with M = cation associated with chromate The Cr(VI) concentrations of aliquots were determined with a Thermo Scientific Dionex ICS 3000 ion chromatograph (IC) according to the procedure described by Loock et al. ( 2014 and a Dionex IonPac AS7 4 x 250 mm analytical separation columns.A 1 000 µL injection coil and a 2 x 375 µL knitted reaction coil coupled to a UV-visible absorbance detector set at a wavelength of 530 nm were used.250 mM ammonium sulphate was used as an eluent and an isocratic pump carried the eluent and the injected sample through the system. 2 mM DPC solution was used for post column coloration, which was delivered into the system by an AXP pump.The flow rates of the eluent and colorant were 1.0 mL•min -1 and 0.5 mL•min -1 , respectively.A six-point calibration was performed for Cr(VI), ranging from 5 to 75 µg•L -1 .The correlation coefficient of this linear calibration line was > 99.96%.The detection limit of this method at a confidence level of 98.3%, which was calculated according to the method specified by Skoog et al. ( 2013), was experimentally determined as 0.9 µg•L -1 by Loock-Hattingh (2016).The accuracy of the specific analytical method (and the specific instrument used) was recently indicated (Venter et al., 2016) by analysis of a Community Bureau of Reference (former reference materials program of the European Commission) BCR N o 545 welding dust filter certified reference material (individual identification N o B7-36, purchased on 25 January 2013), for which a value of 39.7 mg•g -1 Cr(VI) was obtained, which was within the uncertainty of the reference material that was defined as the half width of the 95% confidence interval of the mean of 40.2 mg•g -1 Cr(VI).

X-ray absorption spectroscopy (XAS)
XAS experiments were carried out at PNC-CAT's bending magnet beam-line 20-BM of the Advanced Photon Source (APS), Argonne, IL, USA.Finely ground and homogenized BFD samples were placed as monolayers on tapes and loaded into Plexiglass sample holders.Experiments were carried out at room temperature both in transmission and fluorescence modes using ion chambers filled with nitrogen and helium gases, and a Canberra 13-element detector.A Cr foil was used for energy calibration.Between 4 and 6 scans were collected from each sample and the reference materials.Data reduction and analysis were performed with ATHENA (Ravel and Newville, 2005).The least squares fitting of the XANES spectra were made with LSFitXAFS (Paktunc, 2004).XAS, including both X-ray absorption near-edge spectroscopy (XANES) and extended X-ray absorption fine structure (EXAFS) regimes, was employed to determine the sources and quantitative speciation of Cr in the BFD samples and possible cation association of CrO 4 2-that could be leached and precipitated from the BFD samples.Precipitates were prepared by contacting 50 g of BFD with 1 L of deionized water for 4 consecutive 24 h contacting periods.After 24 h contact, the water was separated from the BFD by means of filtration, after which the BFD were re-contacted with fresh water and the procedure repeated.All four leachates were combined and then allowed to evaporate at room temperature in a fume cupboard circulated with filtered air to prevent contamination, producing a solid residue that was analysed with XAS spectroscopy techniques.

Particle size analysis
The d 90 , d 50 and d 10 of the particle size distribution of the different BFD samples are presented in Table 1.The d 90 is defined as the equivalent particle size for which 90% of the particles are smaller, with definitions of d 50 and d 10 derived similarly.The particle size analyses indicated that the d 90 of all four BFD samples ranged between 32.8 and 152.2 μm, with the A and B samples being coarser than the C and D samples.The particle sizes reported here are similar to BFD particle sizes reported previously (Van Staden et al., 2014;Beukes et al., 2010).al., 2014;Beukes et al., 2010;Beukes et al., 1999).However, to confrim that unreacted/partially-reacted chromite particles were present in the BFD samples, one of the unevenly-shaped grey particles, marked with a white box in Fig. 1(a) (BFD sample A), was analysed with SEM-EDS, which revealed Cr and Fe contents of 30.7 and 20.8 wt%, respectively.These correspond to a Cr/ Fe ratio of 1.48 which is comparable to typical South African chromite ore (Cramer et al., 2004).The SEM micrographs also revealed that the BFD samples A and B (Figs. 1(a) and 1(b)) contained more unreacted/partially-reacted larger raw material particles when compared to C and D (Figs. 1(c) and 1(d)).The SEM observations confirmed the particle size analysis results (Table 1), which indicated that the A and B samples were coarser than C and D. As previously indicated, the presence of a greater number of larger raw material particles in the BFD could be ascribed to the differences in the technology applied or the feed materials used at different production facilities.For instance, some FeCr smelters utilize a cyclone to remove larger and denser particles before the off-gas is cleaned in the bag filter plant, while others merely use a drop-out box that is less effective.SEM-EDS chemical analyses of the BFD samples are presented in Table 2.These results should be considered as semi-quantitative, with O content determined by difference.Carbon (C) was not analysed, since a carbon-based resin was used to set the BFD samples prior to polishing.Cr content of the samples varied between 3.28% (Sample C) and 7.62% (Sample B).As expected, the elements with lower elemental mass that  SEM-EDS chemical analyses of the BFD samples are presented in Table 2.These results should be considered as semi-quantitative, with O content determined by difference.Carbon (C) was not analysed, since a carbon-based resin was used to set the BFD samples prior to polishing.Cr content of the samples varied between 3.28% (Sample C) and 7.62% (Sample B).As expected, the elements with lower elemental mass that are typically associated with actual off-gas particles, such as magnesium (Mg), aluminium (Al), silicon (Si) and zinc (Zn), were present in significant concentrations in all of the BFD samples.Of particular interest was that no noteworthy amounts of Na were detected in any of the BFD samples.It is well known that alkali roasting of chromite is an industrial process for the production of sodium chromate, the formation efficiency of which depends on the amount of O 2 and alkali present, as well as the temperature during the production process (Antony et al., 2001).Maine et al. (2005) proposed that Cr(VI) could mainly be associated with alkali metals in FeCr BFD, but did not present any evidence to support this statement.However, the lack of Na in  are typically associated with actual off-gas particles, such as magnesium (Mg), aluminium (Al), silicon (Si) and zinc (Zn), were present in significant concentrations in all of the BFD samples.Of particular interest was that no noteworthy amounts of Na were detected in any of the BFD samples.It is well known that alkali roasting of chromite is an industrial process for the production of sodium chromate, the formation efficiency of which depends on the amount of O 2 and alkali present, as well as the temperature during the production process (Antony et al., 2001).Maine et al. (2005) proposed that Cr(VI) could mainly be associated with alkali metals in FeCr BFD, but did not present any evidence to support this statement.However, the lack of Na in the BFD samples cannot be considered as evidence that anionic Cr(VI) (CrO 4 2- ) is not associated with significant amounts of Na since the SEM-EDS technique has limitations with respect to Na analysis.

XAS characterisation
As is evident from Fig. 2(a

Trace metal leachability
It is generally accepted that higher concentrations of most heavy and transitional metals will leach out of a solid matrix (e.g.waste and soil) at low pH values (Ding et al., 2014;Lee et al., 2012).This was also observed experimentally, as indicated in Figs. 3  and 4, which present the log concentrations of alkali and alkali earth (Fig. 3), as well as transition and heavy (Fig. 4) metals as a function of leach solution pH.The higher solubility at lower pH was especially evident at pH ≤ 3. Elements occurring in low concentrations in the BDF samples (< 10 mg•kg -1 BFD) were excluded from all figures to prevent congestion of the graphs, and Na contents were excluded from Figs 3(a) and (b) at pH > 7, since NaOH solution was used to increase the pH of the leach solutions. 10 It is generally accepted that higher concentrations of most heavy and transitional metals will leach out of a solid matrix (e.g.waste and soil) at low pH values (Ding et al., 2014;Lee et al., 2012).This was also observed experimentally, as indicated in Figs. 3 and 4, which present the log concentrations of alkali and alkali earth (Fig. 3), as well as transition and heavy (Fig. 4) metals as a function of leach solution pH.The higher solubility at lower pH was especially evident at pH ≤ 3. Elements occurring in low concentrations in the BDF samples (< 10 mg•kg -1 BFD) were excluded from all figures to prevent congestion of the graphs, and Na contents were excluded from When comparing metal concentrations presented in Fig. 3 to concentrations presented in Fig. 4, it is evident that most alkali and alkali earth metals, excluding Ba, were present in much higher concentrations than transition and heavy metals, except for Zn that had concentrations comparable When comparing metal concentrations presented in Fig. 3 to concentrations presented in Fig. 4, it is evident that most alkali and alkali earth metals, excluding Ba, were present in much higher concentrations than transition and heavy metals, except for Zn that had concentrations comparable to alkali and alkali earth metals, especially at very low pH.The results presented in Figs 3 and 4 will be contextualized further in the following sections.

Cr(VI) leachability
Figure 5(a) presents the leachable Cr(VI) concentrations as a function of leach solution pH, as well as the two solubility fractions, i.e., combined water-soluble and sparingly watersoluble Cr(VI), as well as total Cr(VI) determined.Figure 5(b) indicates the same data normalized to percentage, based on the total extractable Cr(VI) of each sample.From these results, it is evident that the amount and percentage of leachable Cr(VI) increased with an increase in leach solution pH at pH ≥ 9.This is in contrast to the results obtained for all the other metals (Figs 3 and 4) for which the leachability increased as the leach solution pH decreased.
From the data presented in Figs 3, 4 and 5, it is impossible to shed any light on the possible association of anionic CrO 4 2with cations, to indicate the identity of the Cr(VI) compounds that are present in the BFD samples.However, the observed increase of leachable Cr(VI) with an increase in pH (Fig. 5(b)) is an indication that not only water-soluble (e.g.Na 2 CrO 4, K 2 CrO 4 ), but also sparingly water-soluble (e.g.CaCrO 4 , SrCrO 4 ) and waterinsoluble (e.g.BaCrO 4 , PbCrO 4 ) Cr(VI) compounds occur in the samples evaluated.This data can also be used to estimate the average water-soluble, sparingly water-insoluble and    2005) attempted to explain the slow release of Cr(VI), beyond the initial fast solubilisation of a certain fraction.These authors postulated that most of the Cr(VI) condenses onto the surface of the rounded fly ash type BFD particles, which can dissolve very rapidly in water.They further suggested that a small fraction of Cr(VI) might be trapped by the glassy surface of the BFD particles, which resulted in the slow release of the remaining Cr(VI).However, the results presented in this paper prove that the slow release, following the rapid initial dissolution of a fraction of the Cr(VI), can rather be attributed to the presence of substantial fractions of sparingly water-soluble and water-insoluble Cr(VI) compounds.
According to a survey conducted by Beukes et al. (2012) at various FeCr producers, the pH of FeCr process waters usually varies between 6.2 and 9.0.As indicated in Fig. 5(b), as well the discussions thereafter, approximately 31% of the total Cr(VI) is leached at pH ≤ 9, which correlates with the water-soluble Cr(VI) fraction.In order to confirm that the currently applied Cr(VI) treatment strategies extract only the water-soluble Cr(VI) fraction, even if extended extraction times are applied, the solid residue obtained from the evaporation of 4 successive 24 h extractions of the BDF samples at pH 7 were characterized with XAS and presented in Fig. 7.The pH level of 7 was chosen to avoid potential complications from pH adjustments (i.e.addition of CO 3 2-and/or OH -altering cation association of CrO 4 2-at higher pH levels).
As is evident from Fig. 7, the XANES spectra of the solid residue of all four BFD samples investigated (A, B, C and D) indicated that the Cr present in the samples is dominated by Cr(VI) species.The spectra display a strong similarity to the spectrum of Na 2 CrO 4 , which is a water-soluble Cr(VI) species.This does not necessarily imply that Na 2 CrO 4 was the sole source of Cr(VI) in the leachate precipitates of the BFD samples.Due to the low Cr concentrations, EXAFS spectra were not optimal and the results should therefore be regarded as qualitative, rather than quantitative.These results also did not permit modelling of the local coordination environment of the Cr(VI) species.Notwithstanding the afore-mentioned limitations, the presence of Cr(VI) in all BFD leachate precipitates is unequivocal; additionally, the Cr(VI) species were similar to Na 2 CrO 4 .

CONCLUSIONS
The results presented in this paper indicate that the currently applied Cr(VI) treatment strategies of FeCr producer (with process water pH ≤9) only effectively extract and treat the water-soluble Cr(VI) compounds from the BFD, which represent approximately 31% of the total Cr(VI) present in the BFD samples evaluated.Extended extraction times of 4 consecutive 24 h extractions, proved futile in extracting sparingly watersoluble and water-insoluble Cr(VI) species within the aforementioned pH range, as confirmed with XAFS.
Considering the above-mentioned, it is critical that improved Cr(VI) treatment strategies be formulated that would extract all of the Cr(VI) from BFDs.The results presented in this paper suggest that extraction at pH 13 would be required to solubilise the sparingly water-soluble Cr(VI) species from BFD.This will ensure that sparingly water-soluble Cr(VI) compounds do not leach from the waste storage facilities (e.g.slimes dams).Otherwise, there is the potential of Cr(VI) releases over time, which has been proven to results in environmental Cr(VI) pollution (Loock-Hatting, 2016;Loock-Hatting et al., 2015;Loock et al., 2014).However, it is highly unlikely that such a high process-water pH would be economically feasible.Additionally, it is well known from literature that Fe(II), currently the most commonly employed chemical reductant of Cr(VI), oxidizes very quickly with dissolved oxygen at such high pH levels.Therefore, the effectiveness of this and other potential reductants at elevated pH levels needs to be investigated.Furthermore, the need to extract the water-insoluble Cr(VI) species prior to chemical reduction should be investigated in greater detail, in order to assess the likelihood of such species being mobilised over extended time periods relevant to waste disposal facilities.
Figure 1 SEM micrographs for BFDs A (a), B (b), C (c) and D (d), at 2 000 (larger image) and 100 (smaller image) times magnification.The larger particle with an uneven morphology, marked by the white box in (a), is an example of an un-reacted chromite particle.

Figure 1
Figure 1 SEM micrographs for BFDs A (a), B (b), C (c) and D (d), at 2 000 (larger image) and 100 (smaller image) times magnification.The larger particle with an uneven morphology, marked by the white box in (a), is an example of an un-reacted chromite particle.
Figure 2 Cr K-edge XANES spectra of the BFD samples and their comparison to the UG2 chromite, slag and FeCr reference samples obtained from a South African FeCr producer (a).Least squares fitting of the derivative of the normalized BFD samples with end member compositions of chromite, slag and FeCr.Measured spectra are shown in black and the fits are indicated by the red circles (b).

Figure 2
Figure 2 Cr K-edge XANES spectra of the BFD samples and their comparison to the UG2 chromite, slag and FeCr reference samples obtained from a South African FeCr producer (a).Least squares fitting of the derivative of the normalized BFD samples with end member compositions of chromite, slag and FeCr.Measured spectra are shown in black and the fits are indicated by the red circles (b).

Figure 3
Figure 3Figure 3 Log concentration of alkali and alkali earth metals in (a) BFDs A and B, as well as (b) BFDs C and D, as a function of leach solution pH.

Figure 4
Figure 4 Log concentration of transition and heavy metals present in (a) BFD A and B, as well as (b) C and D as a function of leach solution pH Figure 5 Leachable Cr(VI) as a function of leach solution pH, combined water-soluble and sparingly watersoluble Cr(VI), as well as total Cr(VI) (a).The same data normalized to % Cr(VI), based on the total extractable Cr(VI) (b).

Figure 5
Figure 5 Leachable Cr(VI) as a function of leach solution pH, combined water-soluble and sparingly water-soluble Cr(VI), as well as total Cr(VI) (a).The same data normalized to % Cr(VI), based on the total extractable Cr(VI) (b).

TABLE 1 The average d 90 , d 50 and d 10 values of BFDs BFD Particle size (µm)
Each of these figures present combined images, with the larger images indicating crosssectional polished micrographs.The smaller images (in the bottom right-hand corners) represent micrographs of the BFDs pressed onto stubs with adhesive carbon tape.It is evident from these micrographs that the BFD samples consisted mainly of very small particulate matter (mostly < 40 µm), with some larger particles of up to approx.200µm.It was previously shown that the larger unevenly-shaped particles observed in BFD are unreacted or partially-reacted feed materials such as chromite, fluxes or reductants, while the very small spherical particles represent the actual off-gas particulates(Van Staden et