Producers, importers, users and distributors of chemical substances within the European Union (EU) are obligated to submit dossiers on the human and environmental health effects of their substances to the Registration, Evaluation, Authorisation and Restriction of CHemicals (REACH) Registry (EU 2006), a legislated environmental protection measure intended to improve the protection of human health and the environment, while maintaining the competitiveness and enhancing the innovative capability of the EU chemicals industry. Compliance with the REACH legislation is mandatory to maintain access to markets.
In December, 2002, the United Nations (UN) adopted the Globally Harmonized System of Classification and Labelling of Chemicals (GHS) (UN 2011) that had been developed within the framework of the Organization for Economic Cooperation and Development (OECD) (OECD 2001). The GHS and REACH have significant implications for environmental protection, transport, insurance, and market access, both within the EU and worldwide. The purpose of the UN GHS is to provide a foundation for national programs for the safe management of chemicals (UN 2011).
We have outlined the GHS aquatic hazard classification framework, scheme, and applications for metals, metal compounds, and alloys in previous publications (Skeaff et al. 2008, 2011). Essentially, comparisons of selected acute ecotoxicity reference values (ERVs) with dissolved metal concentrations released at specified intervals under standard transformation/dissolution (T/D) conditions (Skeaff et al. 2006; UN 2011) into standard aqueous media from mass loadings of 1, 10, and 100 mg/L of the test substance establish the corresponding Acute 1, 2, and 3 hazard classification levels of the scheme. There are also corresponding Chronic 1, 2, and 3 classification levels, unless there is evidence of both rapid removal or partitioning from the water column and no bioaccumulation (UN 2011). The acute ERV is the dissolved metal concentration required to cause mortality to 50% of the most sensitive test species among fish, crustacea, and algae. A Chronic 4 safety net classification applies if the soluble forms of the metal have an acute ERV ≤100 mg/L and there are no T/D data of sufficient validity for the metal or sparingly soluble metal compound to show that release of metal ions will not occur. If the metal or metal compound classifies as Acute 1-Chronic 1, no further information can change the classification. If, however, in a 28-d test at a loading of 1 mg/L, a metal or metal compound classified at some other level (i.e., 2, 3, or 4) does not exceed the long-term no observable effects concentration (NOEC; or chronic ERV), then the Chronic classification is removed (UN 2011).
The EU has modified the UN GHS to arrive at their classification, labeling, and packaging (CLP) regulation (EU 2008a). Currently, REACH calls for application of the 2nd revision of the CLP (EU 2011) in developing hazard classification proposals, with guidance available in EU 2012. Essentially, the EU CLP framework retains the Acute 1-Chronic 1 levels of the GHS, but omits the Acute 2 and Acute 3 levels, while retaining the Chronic 2, 3, and 4. The Acute and Chronic categories are applied independently (EU 2012). For metals, if the acute ERV of the dissolved metal ions is >1 mg/L, then the metal need not be considered further in the classification scheme for acute hazard (EU 2012). Readily soluble metal compounds are classified as Acute 1 if the acute ERV is ≤1 mg/L and are not classified for acute if the ERV is >1 mg/L. For metals with an acute ERV ≤1 mg/L and for sparingly soluble metal compounds, T/D data are used to establish their acute classification in the EU CLP framework. If chronic ERV data are available and >1 mg/L, then the metal substance need not be considered further for Chronic classification (EU 2012). If they are ≤1 mg/L, chronic levels are established by comparisons with T/D data. If chronic ERVs are not available, the EU CLP essentially follows the GHS scheme (EU 2012).
The EU CLP also provides for an approach known as read-across (EU 2012), in which substances sharing common structures, bioavailability or physicochemical properties can be grouped and hazard-classified if they exhibit similar chemical properties and behavior with respect to human health and the environment. Read-across for aquatic hazard classification may be possible for compounds that share a common metal, although a group of transition metal compounds, for instance, could exhibit a range of solubilities over several orders of magnitude, and distinctions would have to be made in the behavior among readily soluble subgroups, such as chlorides and nitrates, and sparingly soluble subgroups such as oxides.
Antimony speciation and toxicity
The speciation and physicochemical state of Sb are important for its behavior in the environment and availability to biota (EU 2008b). For instance, Krachler et al. (2001) and Stemmer (1976) have noted that the toxicological behavior of Sb depends on its oxidation state, with the trivalent form being more toxic than the pentavalent. An understanding of biogeochemical processes needed for environmental protection depends on a knowledge of the chemical species present in the natural environment (Filella et al. 2002a). Although Sb(V) is the thermodynamically stable form of dissolved Sb under oxic conditions, Sb(III) can also coexist with it due to kinetic factors and biological activity (Belzile et al. 2001; Filella et al. 2002b). Similar mechanisms apply to the thermodynamically unstable presence of Sb(V) under anoxic conditions where one would expect Sb(III) to predominate. The persistence of Sb(III) in oxic waters requires kinetic stabilization (Filella et al. 2002a). Nonetheless, because conclusive evidence of a significant difference in aquatic toxicity between the 2 Sb valence states was lacking, the EU decided “not to differentiate between relevant and reliable toxicity results originating from tri- or pentavalent Sb studies” in its Sb risk assessment (EU 2008b). However, if Sb substances yield the same Sb species on dissolution, read-across of environmental toxicity data of well-investigated Sb substances to those less well-characterized can be fully justified.
Moreover, the EU CLP provides for the consideration of rapid environmental transformation as a criterion in deriving chronic hazard classification levels for a metal or metal compound (EU 2012). For rapid environmental transformation, a 70% reduction in the concentration of the species of concern through processes such as speciation and precipitation over 28 days for metals and inorganic compounds would be consistent with the corresponding biodegradation criterion for synthetic organic chemicals. When applicable, this criterion would move a classification to a less severe level compared to the case in which rapid environmental transformation did not apply (EU 2012).
In terms of obtaining Sb speciation analyses in solution, Filella et al. (2002a) note that nearly all published data on inorganic Sb speciation are based on the analysis of Sb(III) and total Sb, with Sb(V) taken as the difference between total Sb and Sb(III). Nonetheless, there have been recent advances in the determination of Sb speciation with high performance liquid chromatographic separation and inductively coupled plasma mass spectrometric detection (HPLC-ICP-MS) (Krachler and Emons 2001; Zheng et al. 2001).
The general objective of this study was to examine the T/D characteristics of Sb metal and several Sb compounds with respect to total dissolved Sb. An additional objective was to determine the speciation of dissolved Sb in terms of the concentrations of Sb(III) and Sb(V) in certain T/D solutions. The data from this study have been used in the dossiers for Sb and its selected compounds submitted to the European Chemicals Agency (ECHA) under REACH. For this article, we will apply the EU CLP scheme in worked examples of classification proposals for Sb metal and the compounds we examined for T/D characteristics.
The substances examined for T/D characteristics were: sodium hexahydroxoantimonate, NaSb(OH)6; antimony metal, Sb; antimony trioxide, Sb2O3; antimony sulfide, Sb2S3; sodium antimonate, NaSbO3; antimony tris(ethylene glycolate), Sb2(C2H4O2)3; antimony trichloride, SbCl3; antimony triacetate, Sb(CH3COO)3; and antimony pentoxide, Sb2O5, at pH 6 and 8.5. As provided by the International Antimony Association and the producers, details on the Sb-bearing substances such as CAS number, molecular weight, typical particle size data, and supplier are presented in Table 1. All substances were in the powder form, and all were white except Sb metal, Sb2S3, and SbCl3 that were dark gray or black.
Table 1. Details on antimony metal and compounds for T/D examination
Two pH values; 1, 10 and 100 mg/L @ (three triplicate loadings plus one procedural blank); 1, 10 and 100 mg/L loadings run for seven days and 1 mg/L loading continues for 28-days; redox speciation analysis for 1 mg/L loadings over 28 days.
Two pH values; 1 mg/L @ (three triplicate loadings plus one procedural blank) for 28-days; redox speciation analysis for 1 mg/L loadings over 28 days.
Mass median aerodynamic diameter.
From supplier's label.
From particle size distribution. D = particle diameter.
The study scope provided for full 7- and 28-day tests at pH 6 and 8.5 on Sb metal, Sb2S3, NaSbO3, Sb2O5, and NaSb(OH)6 at triplicate loadings of 1, 10, and 100 mg/L, and 28-day tests at pH 6 and 8.5 on Sb2O3, Sb2(C2H4O2)3, SbCl3, and Sb(CH3COO)3 at triplicate 1 mg/L loadings. We sampled all T/D solutions in triplicate and analyzed them for total dissolved Sb, and, for the 1 mg/L loadings, Sb(III) and Sb(V).
We have previously presented in some detail the T/DP experimental procedure, including methods of agitation and temperature control, aqueous media compositions, pH control, pH and dissolved O2 measurement, and sampling procedure (Skeaff et al. 2008, 2011). Essentially, the experimental procedure involves agitating weighed quantities of the test substance in 1 L of aqueous media based on the OECD 203 medium for ecotoxicity testing (OECD 1992) in the temperature range 20°C to 23°C, followed by sampling and analyzing the solutions for the metals of interest.
We tested each substance at each of the 1, 10, and 100 mg/L loadings in triplicate (n = 3) and drew triplicate solution samples at each sampling interval of 0, 24, 48, 96, 168 h and for the 1 mg/L loadings, 336, 504, and 672 h. At 2 and 6 h, we drew a single sample. Moreover, we ran one procedural blank with each loading. A procedural blank is a T/D medium without a substance loading processed identically to and at the same time as those with substance loadings. We assume it to represent the background concentrations associated with such factors as glassware cleanliness, sampling, handling and T/D solution analysis. For the 7-day procedural blanks run simultaneously with the 10 and 100 mg/L loadings, n = 2. For the 28-day procedural blanks run simultaneously with the 10 and 100 mg/L loadings, n = 1. Multiple loadings and triplicate samples and determinations are intended to provide a level of confidence in the data and statistical analysis not otherwise achievable.
The compositions of the T/DP media for pH 6, 7, 8, and 8.5 (UN 2011) are based on that of the OECD 203 medium for aquatic ecotoxicity testing (OECD 1992) that contains 294.00, 123.25, 64.75, and 5.75 mg/L of CaCl2·2H2O, MgSO4·7H2O, NaHCO3, and KCl, respectively, with a hardness of 250 mg/L CaCO3. The OECD 203 medium has a calculated pH of 8 when equilibrated with air at 21.5°C (FactSage 2009), whereas the calculated pH of a medium comprising 10× (10 times) dilute OECD 203 equilibrated at 21.5°C with air containing 0.5% CO2 is 6.05 (FactSage 2009), both before substance additions. For pH 8.5, following the suggestion of Patricio Rodriguez (Centro de Investigación Minera y Metalúrgica (CIMM), Santiago, Chile, personal communication), we modified the OECD 203 by increasing the amount of NaHCO3 to increase the pH and by reducing the amount of CaCl2 × H2O to avoid CaCO3 formation and precipitation. The pH 8.5 medium thus contained 29.40, 123.25, 316.3, and 5.75 mg/L of CaCl2·2H2O, MgSO4·7H2O, NaHCO3, and KCl, respectively, with a hardness of 70 mg/L CaCO3, and a calculated pH of 8.70, although the actual levels in practice varied in the range of 8.4 to 8.6. The T/DP implies that the variability for pH 8 should be ±0.2 unit for pH 8 (UN 2011). Applying this criterion to pH 6.05 and 8.5, we find the target ranges to be 5.85 to 6.25 and 8.3 to 8.7, respectively. Nonetheless, the T/DP guidance provides that the “pH should not be adjusted during the test using an acid or alkali” (UN 2011).
Using a 15 mL syringe and a 0.8/0.2 µm Acrodisc filter, we drew, filtered and analyzed the T/D solution samples for total dissolved Sb, and, for the 1 mg/L loadings, Sb(III) and Sb(V). For the 1 mg/L loading T/D samples intended for Sb speciation analysis, we used an auto-pipetter to measure 8.0 mL of T/D solution, diluted this sample with an auto-pipetted 2.0 mL of 2 mM of sodium ethylenediaminetetraacetic acid (Na2EDTA) to maintain Sb(V) as the anion Sb(OH) and to transform Sb(III) to the anion Sb(EDTA)− in solution (Lintschinger et al. 1997), then analyzed for Sb(III) and Sb(V) within a day or 2. The speciation was unchanged over at least 4 days with Na2EDTA used to stabilize the samples. For the T/D samples intended for total dissolved Sb analysis, we added a few drops of trace metal grade HCl as a preservative.
To determine the concentrations of Sb(III) and Sb(V), we developed an adaptation of the methods of Lintschinger et al. (1997), Krachler and Emons (2001), and of Müller et al. (2009). The method is based on the existence of Sb5+ as Sb(OH) over the pH range of 2.7 to 10.4 and the reaction of dissolved Sb3+ with Na2EDTA to form Sb(EDTA)−. The stability of the SbEDTA− anion in the presence of other anions, including the acetate and ethylene glycolate (Filella and May 2005), is the basis of this analytical method for Sb(III).
Once they are separated on an anion exchange (a.i.x.) column, the concentrations of these anions in the column eluent can be measured. Thus the system for the determination of the concentrations of Sb(III) and Sb(V) comprised HPLC-ICP-MS. We injected the solutions for Sb speciation onto a Dionex a.i.x. column AS14/AG14 installed in a Dionex ICS3000 HPLC system. Passing the eluent from the a.i.x. column directly to the nebulizer of a PerkinElmer Elan 6100 DRC ICP-MS (dynamic reaction cell inductively coupled plasma-mass spectrometer) yielded approximate peak retention times for Sb(V) and Sb(III) of 4.6 and 6.6 min, respectively. We then used a Turbochrom Graphic Method Editor to convert the areas under the peaks to Sb(V) and Sb(III) concentrations.
To verify initial instrument calibration and to monitor instrument stability, we incorporated QA-QC (quality assurance-quality control) measures into each HPLC-ICP-MS batch of speciation analyses and total Sb. For the Sb(III) control, we used stock solutions of 1000 mg/L of Sb(III) as K2[Sb2(C4H2O6)2]·3H2O (Sb tartrate) and for Sb(V) 1000 mg/L of Sb(V) as KSb(OH)6 (potassium hexahydroxoantimonate) to make up a stock solution containing 100 mg/L of each of Sb(III) and Sb(V). Using an autodilutor just before analyses, we diluted the 100 mg/L solution of Sb(III)/Sb(V) with 2 mM Na2EDTA to obtain calibration standards with concentrations of 1, 10, and 100 µg/L of Sb(III)/Sb(V).
Typically, in the analysis of a batch of samples comprising 4 sets of 10 or fewer solutions for Sb speciation, we would also analyze a duplicate from each set; one set of the 1, 10, and 100 µg/L Sb(III)/Sb(V) calibration standards; 2 sets of the 1 and 10 µg/L Sb(III)/Sb(V) calibration standards; 5 QC solutions with Sb(III) and Sb(V) concentrations within the estimated ranges of Sb(III) and Sb(V); and a total of 14 blanks and rinses before and after each set of 10 sample solutions. Of 20 calibration and QC solutions containing Sb(III) and Sb(V), all would typically be within 10% or less of the target values, 19 of these within 5% and 3 within 1%. Rinses yielded below detection limit concentrations of Sb(III) and Sb(V). There was no evidence of analytical drift during the course of the analysis of a batch of solution samples. The limits of quantification (LOQs) for each of Sb(III) and Sb(V) were 1.0 µg/L.
Total dissolved Sb
For total dissolved Sb in the mg/L concentration ranges, we used a Varian Vista RL ICP-AES (atomic emission spectrometer) and, for the µg/L ranges, either a Thermo-Fisher X-series II ICP-MS or the Perkin Elmer Elan 6100 ICP–MS. In a typical set of 72 T/D solutions for total Sb determination on the Elmer Elan 6100, we would initially analyze a blank, 2 standards, 1 QC, 10 blanks, a QC, a blank, 2 standards, a blank, a rinse, 2 trace metal standards, followed by the 72 T/D solutions in 3, 6, or 9 subsets interspersed with rinses, QCs, and standards. Of 25 QCs and standards, 22 were within 2.5% or less of the target values and 3 were within 3.5%. Overall, the LOQs for the Thermo-Fisher X-series II ICP–MS and the PerkinElmer Elan 6100 ICP–MS varied in the range 0.08 µg/L to 1.5 µg/L. For the Varian Vista RL ICP-AES, a typical set for the analysis of 168 T/D sample solutions for total Sb determination would also comprise an initial blank, 2 standards of 5 and 10 ppm, 2 rinses, 5 determinations of the LOQ, followed by subsets of between 3 and 11 T/D samples separated by rinse-rinse-standard-rinse cycles, for a total of 21 standards and 49 rinses. Of the 21 5 and 10 ppm standards, 15 were within 2% of the target values, 3 within 5% and 3 within 6%. The blanks yielded either less than 0.2 or 0 ppm. Typically, the LOQs on the Varian Vista RL ICP-AES would be less than 0.2 ppm.
Calculation of net changes in concentration
Skeaff et al. (2011) recently presented the method of calculating net changes in concentration for T/D reaction kinetic data. Essentially, we calculate the net average total dissolved concentrations among all 3 test jars for each sampling time over the 7 or 28 days of each test for each loading as the difference between the average concentrations in the tests with loadings and those in the procedural blanks. We also calculate values of σ, the standard deviation of the 3-concentration data set, and of σ%, the percentage coefficient of variance, which is the quotient 100 σ/(within-vessel average). In the following, when we refer to the T/D data, we mean the net average T/D reaction kinetic data calculated as above.
Derivation of GHS hazard classification proposals
The acute and chronic ERVs we will use in the comparisons with the T/D data to derive EU CLP hazard classification proposals for dissolved Sb are the lowest relevant and reliable endpoints originating from tri- or pentavalent Sb ecotoxicity studies (EU 2008b), and are presented in Table 2 as concentrations of total dissolved Sb. These acute and chronic ERVs are used in the respective Chemical Safety Reports prepared for the REACH program.
Table 2. Lowest effect levels for aquatic species (From European Union Risk Assessment Report: Diantimony Trioxide. CAS No: 1309-64-4. EINICS No: 215-175-0. p 32.)
Sb concentration, mg/L
LC50 (96 hr) = 6.9 (measured total)
NOEC = 1.13 (measured total)
LC50 (48 h) = 12.2 (measured total)
NOEC = 1.74 (measured total)
Hydra (Chlorohydra viridissima - Hydra oligactis)
LC50 (72 h) = 1.77–1.95 (measured filtered)
Algae and aquatic plants
EC50 (72 h) > 36.6 (measured total)
NOEC (72 h) = 2.1 (measured total)
Lemna minor (measured dissolved)
EC50 (96 h) > 25.5 (measured dissolved)
NOEC (96 h) = 12.5
Although the value is for the marine species Pargus major, the EU selected the concentration of 6.9 mg/L as the acute Sb ERV for fish, because it is lower than the lowest valid value of 14.4 mg/L for the freshwater fish Pimephales promelas (EU 2008b). For the chronic ERV for fish, the EU selected 1.13 mg/L for the embryo-larva Pimephales promelas as the lowest NOEC value.
RESULTS AND DISCUSSION
During the tests, we took 1600 pH readings, 800 at each pH. For pH 6, 43 of the 800 were greater than 6.25 and less than 6.5. There were also 72 pH levels below 5.85 that represented more severe T/D conditions. For pH 8.5, there were 14 readings below the target limit and 2 above. The variability in pH appeared to be random and there was no evidence to suggest that it affected the reaction kinetics of the test substances.
To facilitate a discussion of Sb speciation, we have introduced a parameter, y, given as
in which Sb(III) and Sb(V) are the measured net concentrations in each of 3 1 mg/L loadings for each substance at any particular time at a given pH. We can then calculate an average value of y and an attendant σ%. A decline in the average value of y with time suggests that Sb(III) is being oxidized to Sb(V), whereas an increase in average y might indicate that Sb(V) is being reduced to Sb(III). To test for statistically significant differences between any 2 average y-values, we used the t test. With the degrees of freedom = 2 for each average y value and α = 0.05, t-crit was 2.1319.
The T/D data for the 1 mg/L loadings of NaSb(OH)6 at a target pH of 6 are plotted in Figure 1a. The data indicate that the NaSb(OH)6 dissolved rapidly and completely, so that by 24 h the net average total Sb concentration was 501 µg/L (σ% = 9), slightly greater than the stoichiometric value of 491 µg/L for 100% dissolution of 1 mg/L. Moreover, as might be expected because Sb is pentavalent in NaSb(OH)6, essentially all of the dissolved Sb existed as Sb(V), most likely as the Sb(OH) anion (Cotton et al. 1999). For instance, at 168 h, the net average concentration of total dissolved Sb was 522 µg/L (σ% = 12), which is close to the corresponding concentration of Sb(V) of 494 µg/L (σ% = 10). In the individual 1 mg/L loadings, the concentrations of Sb(III) were below the LOQ. At 672 h, the net average levels of total dissolved Sb and Sb(V) were in good agreement at 584 µg/L (σ% = 26) and 554 µg/L (σ% = 26), respectively.
For the 10 and 100 mg/L loadings, the net average concentrations of total dissolved Sb are plotted in Figure 1b. As with the 1 mg/L loadings, both the 10 and 100 mg/L loadings exhibited rapid dissolution of NaSb(OH)6, followed by slight but statistically significant declines in concentration. Although the 10 mg/L loadings never exceeded the maximum concentration for 100% dissolution, those for the 100 mg/L loadings did so, although slightly. For the 10 and 100 mg/L loadings, the limiting 168-h values of total Sb were 4.04 mg/L (σ% = 6) and 46.97 (σ% = 1), respectively, which correspond to approximately 82% and approximately 95% NaSb(OH)6 dissolution, respectively. Because the Sb(OH) anion is known to polymerize in aqueous solution (Baes and Mesmer 1976; Greenwood and Earnshaw 1998), the declines in total Sb concentration could be due to the slow formation of insoluble hydrolyzed antimonates. Baes and Messmer (1976) summarize the hydrolysis of Sb(V) as
Filella and May (2005) also note that both Sb(III) and Sb(V) readily hydrolyze. These observations are consistent with the experience in our analytical facilities in which we find it difficult to maintain Sb dissolved in solution for any extended period of time unless we add either EDTA or tartaric acid, or acidify the solution to pH 1–2.
For the 1 mg/L loadings of NaSb(OH)6 at a target pH of 8.5, the net average concentrations are plotted in Figure 1c. As at pH 6, NaSb(OH)6 dissolved rapidly and essentially completely, with Sb(V) accounting for 100% of the dissolved Sb species and the concentrations of Sb(III) below its LOQ. The net average total Sb concentrations achieved a maximum value of 555 µg/L (σ% = 15) at 672 h, which lies within the stoichiometric range of 533 µg/L to 567 µg/L for the actual loadings. The net average 672 h value of Sb(V) was 510 µg/L (σ% = 13), which is within approximately 8% of the total.
The T/D data for total dissolved Sb for the 10 and 100 mg/L loadings at a target pH of 8.5 are plotted in Figure 1d. The T/D data again reveal complete and rapid NaSb(OH)6 dissolution. In a typical 10 mg/L test, the measured 48- and 168-h values of total dissolved Sb were 4.75 mg/L (σ% = 0), and 4.96 mg/L (σ% = 2), respectively, which are in good agreement with the stoichiometric value of 4.99 mg/L. Again, in a typical 100 mg/L test, the corresponding values were 52.3 mg/L (σ% = 0) and 50.8 mg/L (σ% = 1), respectively, against a stoichiometric level of 49.47 mg/L. However, after approximately 24 h, some of the concentrations were somewhat greater than expected compared to the weights of NaSb(OH)6 added to the jars. For instance, one test delivered 7.18 mg/L (σ% = 1) at 168 h, and we expected only 5.03 mg/L. Because the HPLC-ICP-MS method of analysis yielded good agreement between standard and measured concentrations, there is no immediate explanation for the discrepancies. However, the T/DP was designed and intended to measure the rate and extent of release of metals into aqueous media from metals and sparingly soluble inorganic metal compounds, and we are dealing here with a readily soluble metal compound.
Sb(V) accounting for all dissolved Sb in the 1 mg/L loadings at both pHs is consistent with the +5 valence state of Sb in NaSb(OH)6, the lack of reduction of Sb(V) to Sb(III), and the 0 values of y for both pHs.
The T/D data for the 1 mg/L loadings of Sb metal at pH 6 are plotted in Figure 2a and present rapid increases in Sb total and Sb(III) concentrations in the first 24 h to maxima of 849 µg/L (σ% = 9) and 860 µg/L (σ% = 8), respectively, followed by steady and statistically significant decreases to 570 µg/L (σ% = 9) and 560 µg/L (σ% = 7), respectively, at 672 h. The apparent increase in Sb(V) concentration from 46 µg/L (σ% = 36) at 24 h to 54 µg/L (σ% = 37) is not statistically significant. As expected in view of the presence of unreacted zero-valent Sb metal, the predominant species in solution is Sb(III). The net average 24- and 672-h values of total dissolved Sb correspond to 85% and 57% dissolution of the Sb metal, respectively. The decreases in total dissolved Sb and in Sb(III) could be due to the formation of insoluble hydrolyzed antimonites. Baes and Messmer (1976) summarize the hydrolysis of Sb(III) as
The slight but statistically significant decline in the average value of y from 0.95 (σ% = 1.6) at 24 h to 0.91 (σ% = 1.9) at 672 h reflected the relatively constant concentration of Sb(V) in the range of 43 µg/L to 54 µg/L and the decline in Sb(III) noted above, rather than any oxidation of the latter to the former. Thus Sb(III) did not oxidize to any significant extent over time, the continuous flow of air into the aqueous medium having little apparent effect on the unreacted Sb metal, which in turn prevents the oxidation of Sb(III). From considerations of thermodynamics, all Sb metal should be oxidized to Sb(III) before the oxidation of Sb(III) to Sb(V) can occur. Nonetheless, kinetic factors may have a role in the existence of relatively small amounts of Sb(V) in the T/D solutions. In general, there is good agreement between Sb total and [Sb(III) + Sb(V)].
For the 10 and 100 mg/L loadings at pH 6, the T/D data for total dissolved Sb are plotted in Figure 2b, and reveal initial rapid increases over the first 24 h, followed, in the case of the 10 mg/L loadings, by an approach to a limiting concentration, and in the case of the 100 mg/L loadings a slight but statistically significant decline in concentration over the next 144 h, with a tendency toward limiting net average values of 8.25 mg/L (σ% = 17) and 27.2 mg/L (σ% = 7) around 168 h. For the 10 mg/L loadings, the 168-h concentrations correspond to approximately 80% dissolution of the Sb metal, and for the 100 mg/L loadings, we observe approximately 30% dissolution.
For the 1 mg/L loadings of Sb metal at a target pH of 8.5, the T/D data of Figure 2c again reveal relatively rapid increases in net concentrations of total dissolved Sb and Sb(III) to maxima of 727 µg/L (σ% = 4) and 716 µg/L (σ% = 5), respectively, at 96 h and then slow decreases to 613 µg/L (σ% = 6) and 483 µg/L (σ% = 7), respectively, by 672 h. The maximum Sb(V) concentration of 109 µg/L (σ% = 6) occurs at 336 h, which is essentially its value by 672 h. [Sb(III) + Sb(V)] was within 7% of Sb total over the 672 h, with the exception of the 336-h values for which the variability was 19%. The declines in Sb total and Sb(III) may again be due to the formation of insoluble hydrolyzed antimonites. The modest increase in Sb(V) is likely due to the oxidation of Sb(III), which, as revealed in the statistically significant increase in Sb(V) from 47 µg/L (σ% = 8) at 24 h to 105 µg/L (σ% = 4) and the equally significant decline in average y from 0.93 (σ% = 0.1) at 24 h to 0.82 (σ% = 1.4) at 672 h, was slight in extent in what might be expected with the continuous passage of air through the test medium. The maximum concentration of Sb total corresponds to approximately 73% dissolution of the Sb metal.
For the 10 mg/L loadings of Sb (Figure 2d), the concentrations of total dissolved Sb reveal variable reaction kinetics which are rapid over the first 24 h and then slower over the following 144 h. The 168-h total Sb concentration of 7.33 mg/L (σ% = 17) represents dissolution of approximately 73% of the Sb metal. The 100 mg/L loadings (Figure 2d) also exhibit rapid reaction kinetics over the first 24 h, which then slow to the point at which a maximum of approximately 23 mg/L (σ% = 8) of total dissolved Sb is attained at 96 h. This is followed by a decline to approximately 18 mg/L (σ% = 6) at 168 h, likely due again to the formation of insoluble hydrolyzed antimonites.
For the 1 mg/L loadings at pH 6 (Figure 3a), the T/D data for total Sb exhibit a rapid increase in concentration over the first 24 h, thereafter increasing more slowly to a value of approximately 70 µg/L (σ% = 3) at 336 h, which corresponds to approximately 10% dissolution of the Sb in Sb2S3. Over the next 336 h, the levels of total dissolved Sb are essentially unchanged: 66 µg/L (σ% = 7) by 672 h. The concentrations of Sb(III) remain in the range 27–35 µg/L over the 672 h, whereas those for Sb(V) arrive at 12 µg/L (σ% = 7) by 24 h and increase to approximately 40 µg/L (σ% = 18) by 672 h, more or less in parallel with the total Sb concentrations. It is thus apparent that Sb is released primarily as Sb(III) because it is trivalent in Sb2S3, but redox conditions are likely such that the T/D solution can sustain only a certain maximum Sb(III) concentration of approximately 30 µg/L, and that Sb(III) released above this amount is oxidized to Sb(V), as suggested by the statistically significant decline in average y from 0.71 (σ% = 0.5) at 24 h to 0.46 (σ% = 11.8) at 672 h. From 24 h onward, [Sb(III) + Sb(V)] was always within approximately 25% of total dissolved Sb, an agreement that we consider to be satisfactory.
For the 10 and 100 mg/L loadings at pH 6, we have plotted the T/D data in Figure 3b. As with the 1 mg/L loadings, both the 10 and 100 mg/L loadings of the Sb2S3 exhibited rapid initial reactivity, followed in the case of the 10 mg/L loadings by a slight decline in concentration to 583 µg/L (σ% = 4) at 168 h, and in the case of the 100 mg/L a gradual approach to a limiting concentration of approximately 5700 µg/L (σ% = 7) by 168 h. The 168 h concentrations of total Sb correspond to approximately 8% dissolution of the Sb in the Sb2S3.
For the 1 mg/L loadings of Sb2S3 at a target pH of 8.5, the T/D data are plotted in Figure 3c. The behavior of Sb2S3 at pH 8.5 was somewhat different than at pH 6 inasmuch that the concentrations of Sb(III) presented a considerable decline from a maximum of approximately 41 µg/L (σ% = 22) at 96 h to approximately 16 µg/L (σ% = 36) by 672 h, apparently due to its oxidation to Sb(V), which increased from approximately 18 µg/L (σ% = 3) at 24 h to 58 µg/L (σ% = 12) at 672 h and accounted for approximately 80% of the total dissolved Sb. However, the behavior of total Sb was similar at both pHs, the 71 µg/L (σ% = 19) concentration at 672 h corresponding to approximately 10% dissolution of the Sb in the Sb2S3. The relative decline in average y, from 0.66 (σ% = 2.1) at 48 h to 0.22 (σ% = 25.0) at 672 h, was significantly greater at pH 8.5 than at pH 6, 0.71 (at 24 h) and 0.46, respectively, indicating the greater susceptibility of Sb(III) to oxidation at the higher pH. The reaction involving Sb(OH) and Sb(OH) presented by Diemar et al. (2009) may explain this effect, and may be written as:
It is possible to show that, when the concentrations of Sb(OH) and Sb(OH) are equal, Reaction 3 becomes favorable above pH approximately 6.15 so that at pH 8.5, E = 0.14. Moreover, if [Sb(OH)] > [Sb(OH)], then for a given pH, Reaction 3 becomes more favorable as reflected in an increased value of E.
The T/D data in terms of total dissolved Sb for the 10 and 100 mg/L loadings at pH 8.5 are plotted in Figure 3d. Both loadings presented initial rapid increases in total dissolved Sb, followed by more gradual increases. The 100 mg/L loading exhibited classic reaction kinetic behavior, suggesting an approach to a limiting concentration of perhaps 6800 µg/L by 2000 h. The 168 h concentrations of total Sb correspond to approximately 12% and 7% dissolution of the Sb in the Sb2S3 at 10 and 100 mg/L, respectively.
The percentages of dissolved Sb lay in the range 7% to 12% and were not notably affected by pH, so we would consider Sb2S3 to be moderately soluble.
We have plotted the T/D data for the 1 mg/L loadings of NaSbO3 at pH 6 in Figure 4a. At this loading and pH, NaSbO3 was sparingly soluble, with the maximum total dissolved Sb concentration of approximately 21 µg/L (σ% = 5) at 24 h corresponding to approximately 3% of the Sb in the compound. The data also reveal that all Sb dissolved as Sb(V), as expected because Sb is pentavalent in NaSbO3. The net concentrations of Sb(III) were below detection limits, so that y was always zero over the 672 h. The slight decline in total Sb to approximately 16 µg/L (σ% = 14) at 672 h is statistically significant and could be due to the hydrolysis of Sb(V).
For the 10 and 100 mg/L loadings, the T/D data for pH 6 are plotted in Figure 4b. For both loadings, the levels of total Sb increased quite rapidly over the first 2 h, to 127 µg/L (σ% = 53) and 1833 µg/L (σ% = 2), respectively. Thereafter, both loadings presented slower dissolution rates, and approaches to 168-h limiting concentrations of 172 µg/L (σ% = 52) and 2183 µg/L (σ% = 1), respectively.
The T/D data for the 1 mg/L loadings of NaSbO3 at pH 8.5, plotted in Figure 4c, reveal that NaSbO3 was even less soluble at this pH. The maximum total Sb concentration of approximately 18 µg/L occurred at 6 h, and thereafter declined to approximately 12 µg/L (σ% = 8) at 672 h, due again most likely to Sb(V) hydrolysis. As at pH 6, the net concentrations of Sb(III) were close to or less than the detection limits, so that y was always zero over the 672 h.
The total Sb concentrations for the 10 and 100 mg/L loadings at pH 8.5 (Figure 4d), reveal rapid initial increases in total Sb to maxima of approximately 190 µg/L (σ% = 3) and 1910 µg/L (σ% = 2), respectively, at 24 h, followed by decreases to approximately 172 µg/L (σ% = 2) and 1670 µg/L (σ% = 2), respectively, at 168 h, due most likely to Sb hydrolysis. The 168-h concentrations represent approximately 2.7% of the Sb in NaSbO3.
The T/D data for the 1 mg/L loadings of Sb2O3 at pH 6 are plotted in Figure 5a. The concentrations of total dissolved Sb and Sb(III) follow a nearly curvilinear path in which there is a slow decline in the rates of concentration increase. The 672-h levels of total dissolved Sb, Sb(III) and Sb(V) were 74 µg/L (σ% = 30), 76 µg/L (σ% = 23), and 12 µg/L (σ% = 66), respectively, the former representing dissolution of approximately 8% of the 1 mg/L loading. As expected from the trivalent state of Sb in Sb2O3, and seen in the random variability of the average values of y between 0.80 (at 384 h; σ% = 10) and 0.92 (at 168 hr; σ% = 2.2), Sb(III) is the predominant species throughout the 672 h and resisted oxidation, notwithstanding the flow of air through the T/D medium. Also, there is generally good agreement between total dissolved Sb and [Sb(III) + Sb(V)].
As per the T/D data of Figure 5b, the 1 mg/L loadings of Sb2O3 were more reactive at pH 8.5 than at pH 6, attaining 672-h concentrations of total dissolved Sb, Sb(III), and Sb(V) of 176 µg/L (σ% = 7), 145 µg/L (σ% = 10), and 47 µg/L (σ% = 13), respectively. However, the T/D data followed the similar nearly curvilinear path observed at pH 6. Again, there is generally good agreement between total dissolved Sb and [Sb(III) + Sb(V)]. The statistically significant decline in average y from 0.86 (σ% = 3.8) at 24 h to 0.76 (σ% = 3.5) at 672 h suggests some oxidation of Sb(III) to Sb(V). Moreover, with the statistically significant lower value of y at pH 8.5 than at pH 6 (0.76 and σ% = 1.9 vs 0.87 and σ% = 3.5, respectively), Sb(V) comprised a greater proportion of [Sb(III) + Sb(V)] than at pH 6, reflecting the greater favorability of reaction 3 at the higher pH. The 176 µg/L at 672 h represents approximately 22% dissolution of the Sb in the Sb2O3.
The EU (EU 2008b) has reported 7-day Sb2O3 solubility measurements at pH 7.9 for 1, 10, and 100 mg/L loadings of 58, 370, and 2760 µg/L, respectively, of total dissolved Sb.
For the 1 mg/L loadings, the EU (EU 2008b) and the present study 7-day concentrations of 58 and 101 µg/L, respectively, are consistent with anticipated Sb concentration trends, given the differences in pH, 7.9 versus 8.55, respectively, and the known amphoteric behavior of Sb2O3 (Baes and Mesmer 1976). The 28-day EU and present study concentrations of 118 and 176 µg/L, respectively, follow a similar pH trend: 7.9 and 8.54, respectively.
As per Figure 5c, the 1 mg/L loadings of Sb2(C2H4O2)3 exhibited a fairly linear release in total Sb at pH 6, attaining levels of 10 µg/L (σ% = 20), 45 µg/L (σ% = 12) and 156 µg/L (σ% = 10) at 24, 168, and 672 h, respectively, with little indication of reaching limiting concentrations. These concentrations correspond to approximately 2%, 8%, and 27% of the Sb in the Sb2(C2H4O2)3, respectively, so that this compound was considerably more soluble than either Sb2S3 or NaSbO3 at this pH. The attendant 24-, 168-, and 672-h values of Sb(III) and Sb(V) were 7 µg/L (σ% = 20) and 2 µg/L (σ% = 23), 46 µg/L (σ% = 19) and 14 µg/L (σ% = 40), and 144 µg/L (σ% = 15) and 29 µg/L (σ% = 14), respectively, so that, consistent with its trivalent state in the compound, Sb dissolved primarily as Sb(III). The apparent increase in the values of y from 0.74 (σ% = 5.9) at 24 h to 0.83 (σ% = 4.7) at 672 h is not statistically significant, although it is evident that the ethylene glycolate ligand stabilized Sb(III), preventing the oxidation to Sb(V) that we observed with Sb2S3.
The T/D data for the 1 mg/L loadings at pH 8.5 plotted in Figure 5d reveal that Sb2(C2H4O2)3 dissolved more rapidly than at pH 6, and attained a limiting total Sb concentration of approximately 330 µg/L (σ% = 6) by 504 h, which corresponds to approximately 50% dissolution of the Sb in the compound. As at pH 6, most of the Sb dissolved as Sb(III). For instance, at 672 h, Sb(III) and Sb(V) were 202 µg/L (σ% = 0) and 89 µg/L (σ% = 6), respectively, and the average values of y varied in the range 0.63 (at 24 h; σ% = 7.7) to 0.84 (at 336 h; σ% = 8.9), with an average of 0.73, again suggesting the stabilizing effect of the ethylene glycolate ligand on Sb(III).
The T/D data for the 1 mg/L loadings of SbCl3 at pH 6, which are plotted in Figure 6a, present a sharp increase in total Sb to approximately 17 µg/L at 2 h that held at 17 µg/L (σ% = 47) by 24 h, most of which, 18.5 µg/L (σ% = 47), was Sb(III). These values are followed by the oxidation of Sb(III) to Sb(V) as indicated by an increase in the latter from 0.6 µg/L (σ% = 56) at 24 h to 8.4 µg/L (σ% = 38) at 672 h, and the statistically significant decline in average y from 0.97 (σ% = 0.4) at 24 h to 0.39 (σ% = 40) at 672 h. The values of 17 µg/L (σ% = 47) and 13 µg/L (σ% = 52) for total dissolved Sb at 24 and 672 h, respectively, were not statistically different. The maximum total Sb value of 17 µg/L represents approximately 3% dissolution of the Sb in the compound, so we would consider SbCl3 to be sparingly soluble at pH 6.
As per Figure 6b, the 1 mg/L loadings of SbCl3 were considerably more soluble at pH 8.5 than at pH 6, presumably as per Reaction 3. The concentrations of total Sb increased rapidly to approximately 115 µg/L at 2 h, and this was followed by more gradual, if uneven increases over the next 334 h to a maximum of approximately 200 µg/L (σ% = 35), after which they decreased, most likely due to hydrolysis, to 164 µg/L (σ% = 6) at 672 h. Initially, approximately 75% of the Sb had dissolved as Sb(III): approximately 94 µg/L (σ% = 26) of Sb(III) versus approximately 31 µg/L (σ% = 22) of Sb(V) at 24 h. Then oxidation set in and by 672 h the concentrations of Sb(III) and Sb(V) were about equal: 81 µg/L (σ% = 17) and 76 µg/L (σ% = 13) for Sb(III) and Sb(V), respectively. The oxidation of Sb(III) was reflected in the statistically significant decline in average y from 0.73 (σ% = 2.2) at 24 h to 0.51 (σ% = 13.7) at 672 h. The maximum amount of Sb in the SbCl3 that dissolved was approximately 37%, which would rank the compound as quite soluble.
From Figure 6c, the concentrations of total Sb at pH 6 exhibit classic dissolution kinetics, with rapid initial, followed by slower rates of dissolution. The maximum total Sb value of 329 µg/L (σ% = 6), comprised primarily of 286 µg/L (σ% = 9) of Sb(III) with 32 µg/L (σ% = 9) of Sb(V), occurred at 672 h, and represented approximately 82% of the Sb in Sb(CH3COO)3. With the statistically significant decline in y from 0.98 (σ% = 0.4) at 24 h to 0.89 (σ% = 3.8) at 672 h, only a slight amount of oxidation of Sb(III) to Sb(V) occurred, so that the acetate ligand appeared to stabilize Sb(III) under the conditions of the T/DP.
The dissolution behavior of Sb(CH3COO)3 at pH 8.5 was similar to that at pH 6, inasmuch that at both pHs the initial rapid reaction rates slowed after approximately 24 h, as per Figure 6d. The maximum total Sb concentration of 377 µg/L (σ% = 17) occurred at 504 h and amounted to approximately 93% of the Sb in the 1 mg/L loading. The 24-h value of y was somewhat less at pH 8.5, 0.74 (σ% = 0.5), compared to the corresponding value of 0.98 (σ% = 0.4) at pH 6. Nonetheless, the y-values of 0.74 (σ% = 0.5) and 0.69 (σ% = 5.1) at 24 h and 672 h, respectively, are not statistically different, and this would suggest that the acetate ligand stabilized Sb(III), preventing incremental Sb(III) oxidation over the 672 h. With approximately 80% Sb dissolution, Sb(CH3COO)3 would rank as quite soluble.
The T/D data for the 1 mg/L loadings of Sb2O5 at pH 6, plotted in Figure 7a, present rapid and statistically significant although relatively small increases in total Sb to approximately 3.6 µg/L (σ% = 10) at 24 h and then more gradual increases over the next 648 h to 7.3 µg/L (σ% = 17), which represents a bit less than 1% of the Sb in the compound. Because Sb is pentavalent in Sb2O5, Sb(V) was the only detected species of dissolved Sb.
For the 10 and 100 mg/L loadings at pH 6, Figure 7b, the values of total Sb exhibited essentially 2 stages of total Sb dissolution, comprising an initial nearly linear rapid increase in concentration and then a slower, also nearly linear, concentration increase. For the 100 mg/L loadings, the pH tracked the 2 stages of total Sb dissolution, abruptly declining in the first 6 h from 5.9 to 4.2 and thereafter maintaining nearly constant values from 4.2 to 3.9 by 168 h. This decline in pH could be explained by the reactions presented by Baes and Messmer (1976):
For the pH 6 T/D medium, having a relatively low ionic strength of 9 × 10−4 (FactSage 2009), we assume that the equilibrium constants for Reactions 4 and 5 (Baes and Messmer 1976) can be expressed as 10−3.7 (35°C) and 10−2.47, respectively. Assuming that these reactions apply to the pH 6 medium, we have
for which the equilibrium constant K6 becomes 10−6.17 and
In the case of the 100 mg/L loading at pH 6, the 168-h concentration of total dissolved Sb was approximately 1350 µg/L or 1.11 × 10−5 mol/L, which when entered into Equation 2 results in a pH of approximately 1.2. Although not an exact quantitative accounting of the decline in pH, this pH is qualitatively consistent with our observations.
The maximum amounts of Sb dissolved from Sb2O5 at the 1, 10, and 100 mg/L loadings were approximately 1%, 1.2%, and 1.8%, respectively. Thus we would characterize Sb2O5 as sparingly soluble when the initial pH is 6.
Sb2O5 was less soluble at pH 8.5 than at pH 6. The pH 8.5 T/D data for the 1 mg/L loadings of Sb2O5 (Figure 7c) reveal a maximum total Sb concentration of 3.9 µg/L (σ% = 12) occurring at 504 h and a moderate but statistically significant decrease to 2.2 µg/L (σ% = 44) by 672 h. As at pH 6, Sb(V) was the only soluble Sb species detected above the LOQ.
The net average concentrations of total dissolved Sb for the 10 and 100 mg/L loadings are plotted in Figure 7d. The maximum total Sb concentrations occurred at 168 h and were 30 µg/L (σ% = 13) and 456 (σ% = 34) for the 10 and 100 mg/L loadings, respectively. Again, we would characterize Sb2O5 as sparingly soluble at pH 8.5.
Comparing total dissolved Sb to [Sb(III) + Sb(V)]
To examine the agreement between the concentrations of total dissolved Sb and [Sb(III) + Sb(V)], which should be the same, we can calculate the quotient [Sb(III) + Sb(V)]/[Sb total]. For the 24-h, and 7- and 28-day T/D data, but with the exception of that for Sb2O5, which released total Sb in the low µg/L range and no measurable amounts of Sb(III), the quotient [Sb(III) + Sb(V)]/[Sb total] varied between 0.72 and 1.34. Excluding those for Sb2O5, we found that of the 48 values of [Sb(III) + Sb(V)]/[Sb total], 5 had a value of 1.00, which was perfect, 14 were within 5% of 1.00, 18 were within 10% of 1.00, 5 were within 15%, and 6 were within 34%. This agreement provides confidence in, and validation of, the analytical methodology for total Sb and Sb(III) and Sb(V) speciation.
EU CLP classification outcomes
We have summarized the worked examples of the EU CLP outcomes for Sb metal and the selected Sb compounds in Table 3 and present the underlying reasoning below.
Table 3. Summary of worked examples of EU CLP outcomes for antimony and selected antimony compounds
7 days, 1 mg/L loading:
Acute ERV, mg Sb/L
Acute EU CLP outcome, yes/no
28 days, 1 mg/L loading:
Chronic ERV, mg Sb/L
Chronic EU CLP outcome, yes/no
Net concentration, µg/L
Net concentration, µg/L
antimony tris(ethylene glycolate)
The results of the T/D tests on NaSb(OH)6 revealed that the 1, 10, and 100 mg/L loadings dissolved completely at both pHs by 168 h. As a readily soluble metal compound, the standard EU CLP classification criteria are used without the need to follow the approach for metals and sparingly soluble metal compounds. Because the acute ERV is greater than 1 mg/L, NaSb(OH) 6 does not classify for acute aquatic hazard under the EU CLP scheme, nor does it classify for chronic, because the chronic ERV is greater than 1 mg/L.
For Sb metal, the 1 mg/L loadings yielded 168-h total Sb concentrations of 709 and 682 µg/L at pH 6 and 8.5, respectively, both of which are less than the acute ERV of 6.9 mg/L, so it would not classify as EU CLP Acute. Moreover, the 672-h total Sb concentrations of 570 and 613 µg/L at pH 6 and 8.5, respectively, would not classify Sb metal as chronic, because these concentrations are less than the chronic ERV of 1.13 mg/L.
The total dissolved Sb concentrations for the 1 mg/L loadings of Sb2O3 were 36 and 101 µg/L at pH 6 and 8.5, respectively, at 168 h, which are less than the acute ERV of 6.9 mg/L, so that Sb2O3 would not classify for acute under the EU CLP scheme. Again, because the 1 mg/L loadings reached total dissolved Sb concentrations of 74 and 176 µg/L at pH 6 and 8.5, respectively, at 672 h, which are less than the chronic ERV of 1.13 mg/L, it would not classify for chronic under the EU CLP scheme.
The 1 mg/L loadings of Sb2S3 attained 168-h total Sb concentrations of 49 and 53 µg/L at pH 6 and 8.5, respectively, which are both less than the 6.9 mg/L acute ERV, so that Sb2S3 would not classify at the acute EU CLP level. Again, the 672-h concentrations of total Sb at the 1 mg/L loading were 66 and 71 µg/L at pH 6 and 8.5, respectively, so there would be no chronic classification for Sb2S3.
The 168-h concentrations of total Sb for the 1 mg/L loadings of NaSbO3 at pH 6 and 8.5 were 17 and 10 µg/L, respectively, neither of which would classify it as EU CLP Acute. And the 1 mg/L loading 672-h concentrations of 16 and 8.7 µg/L at pH 6 and 8.5, respectively, would keep NaSbO3 from chronic classification.
The 1 mg/L loadings of Sb2(C2H4O2)3 delivered 168-h Sb concentrations of approximately 45 µg/L and 210 µg/L at pH 6 and 8.5, respectively, so it would not classify as EU CLP Acute. Because the 1 mg/L loadings delivered 672-h Sb concentrations of 156 and 325 µg/L at pH 6 and 8.5, respectively, and these are considerably lower than the chronic ERV, there would be no EU CLP chronic classification.
At 168 h, the 1 mg/L loadings of SbCl3 attained Sb concentrations of 14 and 144 µg/L at pH 6 and 8.5, respectively, considerably less than the 6.9 mg/L acute ERV, so it would not classify as EU CLP Acute. Again, because the 1 mg/L loadings delivered 672-h Sb concentrations of 13 and 164 µg/L at pH 6 and 8.5, respectively, and these are considerably lower than the chronic ERV, SbCl3 would not classify.
For Sb2(CH3COO)3, the 1 mg/L loadings resulted in 168-h Sb concentrations of approximately 250 and 280 µg/L at pH 6 and 8.5, respectively, so it would not classify as EU CLP Acute. Moreover, because the 1 mg/L loadings delivered 672-h Sb concentrations of approximately 330 at each pH, and these are considerably lower than the chronic ERV, there would be no EU CLP chronic classification for Sb2(CH3COO)3.
The 1 mg/L loadings of Sb2O5 delivered 168-h concentrations of 4.6 and 2.2 µg/L to the pH 6 and 8.5 media, respectively, considerably less than the 6.9 acute ERV, so it would not classify as EU CLP Acute. The 28-day 1 mg/L loading values were 7.3 and 2.2 µg/L at pH 6 and 8.5, respectively, so there would be no chronic classification for Sb2O5.
We have applied the OECD T/D Protocol (UN 2011) to examine the T/D behavior of NaSb(OH)6, Sb metal, Sb2O3, Sb2S3, NaSbO3, Sb2(C2H4O2)3, SbCl3, Sb(CH3COO)3, and Sb2O5 at pH 6 and 8.5.
We found that, in terms of percentage of compound dissolved as calculated from the concentration of total dissolved Sb from the 1 mg/L loadings at 28 days, NaSb(OH)6 was the most soluble at 100% and both pHs. This was followed by Sb(CH3COO)3 at 81% and both pHs. With 57% and 61% at pH 6 and 8.5, respectively, Sb metal was the next most reactive. We then had Sb2(C2H4O2)3 with 27% and 56% at pH 6 and 8.5, respectively, and SbCl3 and Sb2O3 with 31% and 21%, respectively, at pH 8.5. Although somewhat arbitrary, we would characterize these compounds as “quite soluble,” or “reactive.” Next would be the “moderately soluble” Sb2S3 with 9% to 10% dissolution at both pHs and Sb2O3 with 4% at pH 6. Finally, we had the “sparingly soluble” Sb2O5 with 1.0% and 0.3% at pH 6 and 8.5, respectively; NaSbO3 with 2.5% and 1.4% dissolution at pH 6 and 8.5, respectively; and SbCl3 with 2.4% at pH 6. Thus in order of decreasing percentage of the compound dissolved or metal reacted, we would have at pH 6: NaSb(OH)6 > Sb(CH3COO)3 > Sb metal > Sb2(C2H4O2)3 > Sb2S3 > Sb2O3 > NaSbO3 ≈ SbCl3 > Sb2O5. For pH 8.5 the order would be: NaSb(OH)6 > Sb(CH3COO)3 > Sb metal > Sb2(C2H4O2)3 > SbCl3 > Sb2O3 > Sb2S3 > NaSbO3 > Sb2O5. Only NaSb(OH)6 exhibited ready solubility, in the sense of rapid 100% dissolution.
For the 1 mg/L T/D data, we introduced the parameter “y” (defined in Eqn. 1), which is a measure of the extent to which Sb exists in solution as Sb(III) in comparison with the sum of the concentrations of the 2 Sb species. For y = 1, Sb exists entirely as the trivalent species, and when y = 0, the pentavalent is unique in solution.
With NaSb(OH)6, NaSbO3, and Sb2O5, the T/D data revealed, within detection limits, that Sb dissolved entirely as Sb(V), which is its valence in these compounds. With Sb metal, Sb dissolution was primarily as Sb(III), again as expected because metallic Sb was always present in the 1 L jars to equilibrate with Sb(III) in solution. Nonetheless, the decline in the values of y from 0.95 to 0.91 at pH 6 and from 0.93 to 0.82 at pH 8.5 over the 672 h suggested some oxidation of Sb(III) to Sb(V).
Antimony dissolved mostly as Sb(III) from Sb2O3 at both pHs, a reflection of the kinetic stability of dissolved Sb(OH) 3 even in the presence of air.
For Sb2S3, the declining values of y at both pHs suggested that Sb(III) oxidized to Sb(V) over the course of the 28-day tests, at pH 6 from 0.71 at 24 h to 0.46 at 672 h, and at pH 8.5 from 0.62 to 0.22. The oxidation of Sb(III) is thermodynamically more favorable at the higher pH. Sb(III) released from SbCl3 was also readily oxidized to Sb(V) at both pHs, with y declining at pH 6 from 0.97 at 24 h to 0.38 at 672 h, and from 0.74 to 0.51 at pH 8.5.
For Sb2(C2H4O2)3, the C2H4O ligand appeared to stabilize Sb(III) in solution, because y varied in the range 0.63 to 0.85 at both pHs. Similar comments apply to the CH3COO− ligand in Sb(CH3COO)3, particularly at pH 6, when y varied only between 0.98 and 0.90 over 28 days, and again between 0.74 and 0.69 at pH 8.5. This stabilizing effect of the organic ligands is in contrast to the oxidation observed for Sb2S3 and SbCl3.
Read-across was used to fill data gaps for 2 types of Sb powder and 1 type of Sb massive with data from Sb trioxide, both yielding primarily Sb(III) in solution, considered to have similar physicochemical environmental fate and/or ecotoxicological properties (International Antimony Association 2010). Moreover, the speciation data presented here are available for future read-across hazard classification proposal derivations for other Sb compounds.
Comparing the T/D data for each substance with the acute and chronic ERVs of 6.9 and 1.13 mg/L, respectively, we found that none of the substances would classify as either acute or chronic under the EU CLP scheme.
We are grateful to the International Antimony Association (i2a) for funding this research. We also gratefully acknowledge many helpful discussions with P. Huntsman-Mapila, and the comments of ML Desforges in reviewing this report. The data in this study cannot be freely used to comply with regulatory requirements such as REACH without the formal agreement of i2a.