Sulfate precipitation treatment for NOM-rich ion exchange brines

Ion exchange (IEX) resins can remove natural organic matter (NOM) from drinking water sources. However, the IEX system produces a waste brine rich of sodium, chloride, NOM and sulfate. The treatment of the waste brine aims to recover a clean solution rich of sodium chloride, that can be reused to regenerate IEX resin. Previous research showed that ceramic nanofiltration partially removes NOM from the waste brine, but sulfate removal requires additional treatment. Sulfate removal by chemical precipitation was previously studied either on brines with low NOM concentrations or water with low concentrations of NOM and salts. The current work focussed on sulfate removal from NOM-rich brines by chemical dosing of (1) BaCl 2 , resulting in precipitation of barite (BaSO 4 ), and (2) CaCl 2 , Ca(OH) 2 and NaAlO 2 , resulting in precipitation of calcium sulfate and, subsequently, ettringite (Ca 6 Al 2 (SO 4 ) 3 (OH) 12 ). Additionally, the effect of NOM on SO 42 (cid:0) removal was studied. Modelling and batch experiments were conducted with IEX and synthetic brines within the typical ion strength range of 0.1 to 1 M. With doses of 2.2 g of BaCl 2 per g of initial sulfate, BaSO 4 precipitation removed more than 83 percent of sulfate, resulting in final concentrations below 0.4 g/L even in the presence of NOM. However, NOM inhibited the precipitation of calcium sulfate and, subsequently, ettringite. With doses of 1.3 g of CaCl 2 , 0.5 – 0.7 g of Ca (OH) 2 and 0.4 – 0.6 g of NaAlO 2 per g of initial sulfate, calcium sulfate and ettringite precipitation removed between 8 and 95 percent of sulfate from NOM-rich brines, resulting in final concentrations between 0.8 and 2 g/ L. As a reference, NOM-free brines required doses of 1.3 g of CaCl 2 , 0.2 – 0.7 g of Ca(OH) 2 and 0.1 – 0.6 g of NaAlO 2 per g of initial sulfate for 89 to 99 percent of sulfate removal, resulting in final concentrations of 0.2 g/L. The inhibition might be attributed to covering of crystal sites by NOM molecules, and to NOM coagulation with aluminium.


Introduction
Anion exchange (IEX) can effectively remove negatively charged natural organic matter (NOM) during drinking water treatment [1]. In IEX processes, resins are reused after cleaning with an electrolyte regenerant solution. In IEX for NOM removal, the regenerant solution is usually NaCl [2][3][4]. The regenerant solution is then reused several times before disposal, which increases the concentrations of NOM and anions, like sulfate (SO 4 2− ) [3][4][5]. The composition of spent IEX brines depends on the quality of the water to be treated, the affinity of negatively charged components with the resin, and the specific IEX system operation. Spent IEX brines obtained by four pilot and full-scale installations had concentrations between 0.04 and 1.6 g/L of dissolved organic carbon (DOC), and concentrations of chloride (Cl − ) and SO 4 2− in a broad range of 2.6 to 19.1 g/L and 0.3 to 24.3 g/L, respectively (Supplementary information). Discharging waste streams with NOM and salts is often problematic. Therefore, water companies aim to limit waste volumes by recovering some of the spent IEX brine components, such as clean water or concentrated NOM [4,6,7]. Additionally, the recovery of clean NaCl regenerant was previously studied [2,[7][8][9]. Previous work has also shown that nanofiltration of brines can remove high levels of NOM, humic substances [2,9,10]. On the other hand, residual SO 4 2− was still present in the nanofiltration permeate, giving potential risk of SO 4 2− accumulation in the recovered regenerant. Chemical precipitation can be an option to remove SO 4 2− from spent IEX brine. SO 4 2− can precipitate with various cations to form sparingly soluble salts, for instance calcium sulfate, barite (BaSO 4 ), and ettringite (Ca 6 Al 2 (SO 4 ) 3 (OH) 12 ). Calcium sulfate exists in different phases, such as gypsum (CaSO 4 *2H 2 O) and anhydrite (CaSO 4 ). Gypsum precipitates at lower temperatures and NaCl concentrations than anhydrite. Gypsum was found to precipitate below 25 • C and with NaCl concentrations below 4 M at 25 • C [11], and is therefore potentially relevant for treatment of IEX spent regenerant, that has NaCl concentrations below 2 M [2][3][4]. A disadvantage of gypsum is its high solubility product, i.e. a log Ksp of − 4.31 at 25 • C [12], and, thus, the required low SO 4 2− concentrations cannot be reached. BaSO 4 has a much lower solubility product, i.e. a log Ksp of − 9.96 at 25 • C [12], but it requires the dosage of toxic BaCl 2 . An alternative is ettringite precipitation. Ettringite is stable at high alkaline conditions, with an optimum pH close to 12 [13], and its solubility product is low, i.e. log Ksp of − 44.91 at 25 • C [14]. SO 4 2− removal from brines by ettringite precipitation has been frequently studied, mostly subsequent to calcium sulfate precipitation [15][16][17][18][19]. Generally, very low concentrations of SO 4 2− could be obtained by chemical precipitation. However, to the authors' knowledge, brines with high NOM concentrations were not studied before. Based on other applications, some NOM interference on chemical precipitation could be expected. NOM and polymaleic acid, which is a synthetic surrogate of the fulvic fraction of humic substances [20], for instance, has been found to inhibit chemical precipitation in studies for water recycling in cooling towers [21,22]. In addition, Banz and Luthi [21] found that NOM of wastewater origin inhibited calcium sulfate precipitation, which was attributed to complexation of Ca 2+ and NOM. However, NOM and salts concentrations in cooling tower water are much lower than in spent IEX brine and different mechanisms might be involved.
In the present study we therefore studied chemical precipitation as an alternative to remove SO 4 2− from NOM-rich spent IEX brines. In particular, the focus was on the performances of BaSO 4 precipitation, and the combination of calcium sulfate and ettringite precipitation to obtain low concentrations of SO 4 2− . The impact of NOM on chemical precipitation in brines and the mechanisms involved were also investigated. We studied spent and synthetic brines with varying NOM, sodium (Na + ), Cl − and SO 4 2− concentrations within the typical ion strength range of 0.1 to 1 M, by means of laboratory experiments and modelling.

Analyses for NOM and ions' concentrations
NOM was measured as dissolved organic carbon (DOC) by a total organic carbon analyser (TOC-VCPH, Shimadzu, Japan) after filtration of the sample. For some of the analyses, the Cl − and SO 4

Preparation of synthetic brines
For the synthetic brines, Na 2 SO 4 and NaCl were weighted and dissolved in demineralized water. The pre-set anion concentrations divided the synthetic brines into two groups: (1) low concentration brines, in the range of 0.2 to 5 gCl − /L and 0.2 to 2 gSO 4 2− /L; and (2) high concentration brines, in range of 9 to 18 gCl − /L and 8 to 16 gSO 4 2− /L. In addition, for the synthetic brines with NOM, concentrated NOM (HumVi, Vitens) was added to obtain concentrations of 0.5 and 2 gDOC/ L. The NOM of HumVi has groundwater origins and was recovered from spent IEX regenerant brine. HumVi was also used and described in previous research [10,23].

Characterisation of spent IEX brine
A spent IEX brine was provided by a drinking water facility in Sweden (Sweden brine). This facility piloted suspended ion exchange (SIX®), as described by Galjaard and Koreman [24]. The NOM of the brines was characterized using liquid chromatography-organic carbon detection (LC-OCD), according to the procedure from Huber et al. [25]. LC-OCD gave the chromatographic fractionation of organic carbon (CDOC), being the sum of the concentrations of five NOM fractions. In decreasing size, the fractions are biopolymers (BP), humic substances (HS), building blocks (BB), low molecular weight acids (LMWa) and neutrals (LMWn) [25].

Precipitation experiments
Sweden brine and the synthetic brines of Table 1 were tested in duplicate for BaSO 4 precipitation. Samples for NOM and anion measurements were taken before and after precipitation. First, 150 mL of brine in a plastic container was stirred on a magnetic plate (speed 9%, Labinco, the Netherlands). The acidity (measured by Multi 3630 with SenTix 940 electrode, WTW, Germany) was adjusted to pH 8 by adding 0.1 M NaOH, to the brines. While stirring, BaCl 2 ⋅2H 2 O, dissolved in ultrapure water, was added in the Ba:SO 4 moles proportion of 1:1, considering the pre-set SO 4 2− concentration of the synthetic brines and the initial SO 4 2− concentration measured in the Sweden brine. After the BaCl 2 ⋅2H 2 O was added, the brines were mixed for 30 min, and the precipitate was allowed to settle for another 30 min. The supernatant was then filtered to collect the samples to be analysed, according to Section 2.1. Sweden brine and the synthetic brines of Table 2 were tested for calcium sulfate precipitation and subsequent ettringite precipitation, similar to the work of Almasri et al. [15]. All the experiments were in duplicate. Samples for NOM and anion measurements were taken before calcium sulfate precipitation, and before and after ettringite precipitation. For the calcium precipitation, 150 mL of brine in a plastic container was stirred on a magnetic plate (at 9% speed, Labinco, the Netherlands). While stirring, CaCl 2 was added in the Ca:SO 4 moles proportion of 1:1, considering the pre-set SO 4 2− concentration of the synthetic brines and the initial SO 4 2− concentration of the Sweden brine. The plastic container was closed directly after the addition of CaCl 2 , and the stirring continued for 2 h. The pH before the addition of CaCl 2 and after precipitation was 7.6 ± 1.2 and 8.3 ± 0.7 (average ± standard deviation), respectively. The solid content in the plastic container after calcium sulfate precipitation was separated using gravity glass fibre filters. The supernatant was then filtered to collect the samples to be analysed, according to Section 2.1. For the subsequent ettringite precipitation, 115 mL of the filtered brine was again stirred on the magnetic plate. While stirring, NaAlO 2 and Ca(OH) 2 were added as solids in the Al:SO 4 and Ca:SO 4 moles proportion of 0.67:1 and 1:1, respectively, considering the concentration of the brines after gravity filtration. Afterwards, the stirring continued for 2 h. The supernatant was then filtered to collect the samples to be analysed, according to Section 2.1. The pH before the addition of NaAlO 2 and Ca(OH) 2 and after precipitation, was 8.2 ± 0.6 and 11.8 ± 0.1 (average ± standard deviation), respectively. The final pH was in the range for ettringite formation according to Almasri et al. [15] without further adjustment, except for Sweden brine. For Sweden brine, the pH before ettringite precipitation was increased from 8.1 to 11.9 with addition of 1 M NaOH. The chemicals used for the precipitation experiments have a purity ≥ 93 percent.

PhreeqC model
The results of the precipitation experiments of the synthetic brines without NOM were compared to the results modelled with PhreeqC, a geochemical modelling software. Solutions with high salinity can be modelled using the Pitzer database, as an alternative for the default PhreeqC database [26]. The PhreeqC script for calcium and ettringite precipitation was validated using the data from Almasri et al. [15]. According to the reasoning in Chapter 1, the calcium sulfate precipitate in the model was gypsum. Our scripts of the models and their validation are presented in the Supplementary Information.

Brines characteristics
The NOM in the Sweden brine and in the synthetic brines consisted mostly of HS and BB (Table 3), because these fractions are preferentially removed by IEX from natural water [27][28][29][30].

Modelling of sulfate precipitation with BaSO4, calcium sulfate and ettringite
The scripts for the PhreeqC model, and the procedure of its validation are presented in the Supplementary Information. Model simulations are shown in Figs. 1-3, including the validation points of BaSO 4 and calcium sulfate precipitation. The model shows that, in Na 2 SO 4 solutions with an ionic strength of 0.1 to 1 M, SO 4 2− precipitation with calcium sulfate depends on the initial SO 4 2− concentration (Fig. 1). The low SO 4 2− removal at low ionic strength is explained by the relatively high solubility product of calcium sulfate, that puts a theoretical limit on the achievable minimum concentration of SO 4 2− to 1.5 g/L [31]. In the same SO 4 2− range, precipitation with BaSO 4 and ettringite, that have low solubility products, only depends on stoichiometry (Figs. 2 and 3).

SO 4 2− removal and effect of NOM during BaSO 4 precipitation
For BaSO 4 precipitation, 2.2 g of BaCl 2 was dosed per g of initial SO 4 2     the synthetic brines without NOM was above 98 percent, resulting in SO 4 2− concentrations below 0.2 g/L (Table 4). showing that the presence of NOM slightly inhibited BaSO 4 precipitation, probably attributed to the antiscalant properties of NOM [32,33]. During BaSO 4 precipitation of Sweden brine, also NOM was removed by 23 percent, which could be a potential problem in case NOM recovery is desired.
Another application issue is linked to the toxicity of barium. The toxicity is linked to its chemical form [42]. In particular, barium salts with low solubility, such as BaSO 4 , are generally considered less dangerous than free Ba 2+ and readily soluble barium salts. Therefore, residual Ba 2+ in the treated brine should be measured. The EPA drinking water standard for barium from 2002 was 2 mg/L [43]. However, ecotoxicity studies derived lower limits for environmental quality standards [44]. In the Netherlands, the maximum concentration of Ba 2+ in surface water intended for drinking water production is 200 µg/L [45]. At high initial SO 4 2− concentration (Fig. 5), calcium sulfate precipitation removed 75 percent of SO 4 2− from the synthetic brine without NOM. However, calcium sulfate precipitation was inhibited by NOM, and the average removed SO 4 2− dropped to 35 and 6 percent in the synthetic brines with 0.5 and 2 gDOC/L, respectively. Similar to the case of BaSO 4 , inhibition of calcium sulfate precipitation by NOM was attributed to the antiscalant properties of NOM [34,35]. Due to the fact that calcium sulfate precipitation only removed 3 to 4 percent of NOM from the synthetic brines, SO 4 2− removal by subsequent ettringite precipitation was inhibited by NOM as well. The dose of chemicals for ettringite precipitation was dependent on the remaining SO 4 2− concentration after calcium sulfate precipitation. Therefore, considerably more NaAlO 2 and Ca(OH) 2 were dosed in the NOM-rich synthetic brines than in the brine without NOM (Fig. 5). Nevertheless, the overall SO 4 2− removal was between 86 and 95 percent, although the final SO 4 2− concentrations for the NOM-rich synthetic brines were still above 0.8 g/L. The overall NOM removal from the NOM-rich brines was between 23 and 67 percent ( Table 5). The calcium sulfate precipitation step only removed between 3 and 11 percent of NOM. The ettringite precipitation step removed 14 percent of the residual NOM from Sweden brine with low initial SO 4 2− , and between 51 and 65 percent of the residual NOM from the two NOM-rich synthetic brines with high initial SO 4 2− , likely due to coagulation of NOM by NaAlO 2 [36,37]. Table 6 summarises the initial NOM/sulfate ratio of the brines, the   percentage of removal and chemicals dosed per initial sulfate concentration (specific dose). Calcium sulfate and subsequent ettringite precipitation is the most suitable for NOM-free brines with high ionic strengths, as indicated by the relatively low specific dose of chemicals and high SO 4 2− removal. However, brines with low ionic strength required a relatively high specific dose of Ca(OH) 2 and NaAlO 2 . When applied to NOM-rich brines, calcium sulfate and subsequent ettringite precipitation removed hardly any SO 4 2− at high initial NOM/ sulfate ratio, or required a relatively high dose of chemicals. The potential effect of the initial NOM/sulfate ratio in practice was to see in our experience with additional spent IEX brines. The percentage of sulfate removed was higher in brines with less NOM and higher initial sulfate concentration than Sweden brine (Supplementary Information), indicating that calcium sulfate and ettringite precipitation were more suitable for IEX brines with low initial NOM/sulfate ratio.
The inhibition of precipitation of Ca 2+ or SO 4 2− crystals caused by NOM and organic acids could be attributed to Ca 2+ complexation or covering of nucleation and growth crystal sites [33,35,[38][39][40]. Table 7 shows that the Ca 2+ that can potentially be consumed by complexation was negligible compared to the available Ca 2+ , i.e. below 10 percent. Therefore, similar to the experiments of Lee et al. [35], covering of crystal site by NOM molecules is suggested as precipitation inhibition mechanism during our experiments. PhreeqC models the interaction between NOM and ions is modelled as complexation and the antiscalant properties of NOM are not considered. Therefore, PhreeqC could not be used to model the NOM-rich brines of our experiments.
An application issue is linked to the purity of the chemical used for the precipitation of the sulfate salts. Natural limestone (calcium carbonate) contains magnesium in case of dolomitization [46], and therefore, Mg 2+ ions can be present as impurity in the produced CaCl 2 . In our experiments, CaCl 2 had high purity, but the presence of Mg 2+ should be checked in application. Previous studies showed that Mg 2+ can maintain SO42-in the soluble form Mg(SO 4 ) [47,48].

Conclusions
Chemical precipitation of SO 4

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.