Removal of strontium from aqueous solutions and synthetic seawater using resorcinol formaldehyde polycondensate resin

Abstract Strontium (Sr) is a valuable metal found in abundance in seawater. However, its recovery from seawater has received little attention despite its many industrial applications. Batch and column adsorption experiments were conducted on the removal of Sr by resorcinol formaldehyde (RF) resin in the presence of co-existing cations at pH 7.5–8.5, where maximum adsorption was found. Batch adsorption capacities of cations followed the decreasing order of Sr > Ca > Mg > K, the order being the same as that of reduction of negative zeta potential. The adsorption data for Sr, Ca and Mg satisfactorily fitted to the Langmuir adsorption model with maximum adsorption capacities of 2.28, 1.25 and 1.15 meq/g, respectively. Selectivity coefficients for Sr with respect to other metals showed that Sr was selectively adsorbed on RF. Column adsorption data for Sr only solution fitted well to the Thomas model. Sr adsorption capacity in the presence of seawater concentrations of Ca, Mg, K and Na was reduced in both batch and column experiments with highest effect from Ca and Mg. However, if Ca and Mg are removed prior to RF adsorption process by precipitation, the negative effect of these ions on Sr removal can be significantly reduced.


Introduction
Seawater is becoming an important source of several valuable mineral resources because of the depletion of high-grade mineral ores on land and recent problems associated with landbased industries resulting from sustainable water and energy demand and environmental issues [1]. Due to its large reserve, seawater is an attractive resource for valuable minerals such as lithium, uranium, rubidium, and strontium (Sr) despite their very low concentrations. Among these valuable minerals, Sr is one of the most abundant minerals in seawater with an average concentration of 6-7 mg/L [2]. Strontium, which is an alkaline earth metal, has many industrial applications, such as in ceramic ferrite magnets, ceramics, glass and pyrotechnics industries, fluorescent lights and fire-works, and also for drilling mud in the oil and gas industries [1,3].
To date, however, the recovery of Sr from seawater has received little attention despite its many industrial applications.
(24 ± 1° C) for 24 h. To increase the ion exchange capacity, the H + form of the resin was conditioned separately with alkaline solutions of 1 M NaOH, KOH and Ca(OH)2 by shaking 5 g of the resin in 500 mL solution of each of the metal hydroxides separately at 150 rpm for 24 h. Following that, the resin was thoroughly washed with MQ water to remove excess NaOH, KOH and Ca(OH)2 and then dried at room temperature (24 ± 1º C) for 24 h and kept stored in air tight containers. The reactions involved in the conditioning of RF resin with the different metal hydroxides are given in Eq. (1,2).

R -OH + AOH => R -O -A + H2O
(1) where R is the resorcinol formaldehyde polymer and A is Na + or K + and B is Ca ++ .
Based on earlier research on the adsorption of Cs by RF resin, it can be said that the phenolic OH group of the resin is ionised to phenolate ion under alkaline conditions [23,26,30] and then it participates in ion exchange reaction of Sr. The uptake of Sr ions by the resin conditioned with Na + , K + , Ca ++ is shown by similar ion exchange reactions in Eq. (3,4).
The virgin and the three RF resins conditioned with NaOH, KOH and Ca(OH)2 were used in the adsorption experiments.

RF resin characterisation
Zeta potential is an important parameter for understanding the mechanism of adsorption as it is the electrical potential close to a particle surface where adsorption of ions from solution phase occurs and it is positively related to the surface charge. The zeta potential values were measured on 0.5 g/L of RF resin suspensions in the presence of 10 -3 M of NaCl, KCl, CaCl2, MgCl2, or SrCl2 in the pH range of 2.5-10.0 using a Zetasizer nano instrument (Nano ZS Zen3600, Malvern, UK). Triplicate measurements were made to minimise undesirable biases (with differences between replicates always being less than 5%) and the average value was used for data analyses.
RF resin was also characterised using X-ray diffraction (XRD) and Fourier transform infrared (FTIR) spectroscopy. The FTIR pattern was obtained using a Miracle-10 Shimadzu FTIR spectrometer. The spectra were obtained at 4 cm -1 resolution by measuring the absorbance from 400 to 4000 cm -1 using a combined 40 scans. XRD was carried out using a PANalytical Empyrean instrument operated at 60-kV with Cu-Ka1 radiation on powdered samples of RF resin.
Ion exchange capacity of the RF resin was measured using two methods [23,28]. In both methods, the RF was equilibrated with known concentrations of NaOH and NaCl and the amounts of NaOH consumed in the ion exchange reaction were determined by titrating the remaining NaOH with 0.1 M HCl using phenolphthalein indicator. The measurements were made in triplicate and the average values were recorded.

Sr adsorption by different RF resins
The first batch experiment was conducted with virgin RF resin and RF resin conditioned separately with NaOH, KOH, and Ca(OH)2 at adsorbent doses of 0.1-1.0 g/L and fixed Sr concentration of 10 mg/L to select the adsorbent having the highest Sr adsorption capacity. The amount of Sr adsorbed at equilibrium, qe (mg/g), was calculated using Eq. (5).

 
M V C C q e 0 e   (5) where C0 = initial concentration of Sr (mg/L); Ce = equilibrium concentration of Sr (mg/L); V = volume of the solution (L); and M = mass of adsorbent (g).
As RF resin conditioned with NaOH had the highest Sr adsorption capacity among the four resins (Fig. S1), the subsequent adsorption experiments were conducted with this resin only.

Sr adsorption by different commercial adsorbents
Equilibrium adsorption experiments were conducted in a set of glass flasks with 100 mL solutions of pH 7.5-8.5 containing Sr at a concentration of 10 mg/L that is approximately the same as the Sr concentration in seawater. Four commercially available adsorbents (CsTreat, AMP, Zeolite and Amberlite FPC 3500) at doses of 0.1 -10 g/L were added to the solutions and agitated as before in a flat shaker to determine their relative Sr adsorption capacities. At the end of the shaking period of 24 h, the samples were filtered and the filtrates were analysed for Sr concentration. The amount of Sr adsorption was calculated using Eq. (5).

K, Ca, and Mg adsorption
The adsorption of K, Ca, and Mg on RF was investigated in experiments similar to that described previously for Sr adsorption with initial cations concentration of 10 mg/L at RF doses of 0.1 -10 g/L. The adsorption data for these metals and Sr were fitted to the Langmuir model (Eq. 6). (6) where, Qe = amount of metal adsorption (meq/g) at the equilibrium metal concentration Ce (meq/L), qmax = maximum amount of the metal adsorption (meq/g), KL = Langmuir adsorption constant (L/meq).

Effect of pH on Sr adsorption
The effect of pH on Sr adsorption was investigated by adding 0.05 g RF resin to 100 mL solution containing 10 mg Sr/L and the suspension was shaken in a flat shaker with a shaking speed of 120 rpm at room temperature (24 ± 1 °C) for 24 h. The solutions' pH values were initially set at a range of 2.5 -10.0 using 0.1 M HCl and 0.1 M NaOH. They were adjusted back to their initial values after a shaking period lasting 4 h. The final pHs at the end of the shaking period were measured. A HQ40d portable pH Meter was used for all pH measurements.

Effect of co-ions in seawater medium on Sr adsorption
Three adsorption experiments were conducted to determine the effect of co-ions on Sr adsorption. The first experiment was conducted in synthetic seawater containing Na (20 g/L), Ca (1 g/L), Mg (2 g/L), K (0.85 g/L) and Sr (10 mg/L) with RF doses of 0.1 -10 g/L. The

Selectivity of Sr adsorption on RF resin
Another experiment was conducted to investigate the selectivity of Sr adsorption on RF resin when the same concentration of Sr, Ca, Mg and K (10 mg/L) was used at a RF resin dose of 1 g/L. The amount of each metal adsorbed was calculated using Eq. (5). The removal efficiency (%) of a metal ion by RF was determined by dividing the amount adsorbed by the amount initially present and multiplying by 100. The distribution coefficient (Kd) (L/g) of a metal was calculated by dividing the amount of metal adsorbed (meq/g) by the equilibrium concentration of the metal in solution (meq/L) [23,25,28] and the selectivity coefficient of Sr with respect to another metal (Ks) was determined by dividing the Kd of Sr by the Kd of the other metal [31]. The higher the Ks value of Sr, the higher the selectivity of adsorption of Sr with respect to the other metal.

Column adsorption experiments
The column adsorption experiments were done using a fixed-bed column consisting of a Pyrex glass tube where a stainless-steel sieve was attached to the bottom followed by a layer of glass beads to provide a uniform flow of the solution through the column. A known quantity of RF resin (10 g) was packed above the glass beads in the column consisting of 1.8 cm inner diameter to yield the desired bed height (6 cm). Feed solutions of 10 mg Sr/L with and without co-ions at concentrations that are found to occur in seawater (see section 2.4.5) were pumped upward through the column at a desired filtration velocity (10.6 mL/min, 2.5 m/h) which was controlled by a peristaltic pump. The experiment was deliberately conducted with a shorter bed height to obtain a faster breakthrough of metals. Since a shorter bed height was used, a smaller filtration velocity was used instead of a normally used velocity of 10 m/h. The effluents at the outlet of the column were collected at regular time intervals and the Sr concentrations were measured. The breakthrough curve shows the loading behaviour of Sr to be removed from the solution in the column and is usually expressed in terms of adsorbed Sr concentration (Cad), inlet Sr concentration (Co), outlet Sr concentration (C) or normalised concentration defined as the ratio of outlet Sr concentration to inlet Sr concentration (C/Co) as a function of time. The maximum column adsorption capacity, q total (meq of Sr/g), for a given feed concentration and filtration velocity is equal to the area under the plot of the adsorbed Sr concentration, Cad (Cad = Co-C) (meq/L) versus effluent time (t, min) and was calculated manually using Microsoft Excel spreadsheet according to Eq. (7).
The breakthrough curves for the co-ions Ca, Mg and K were also determined by measuring their concentrations in the effluent. The amounts of these metals adsorbed on RF were determined using Eq. (7).
Equilibrium Sr uptake qeq (meq/g) or maximum Sr capacity of the column is qtotal divided by the mass of the adsorbent. The column adsorption data were fitted to the Thomas model (Eq 8) which has been described elsewhere [32]. ln (C o /C − 1) = k Th q o M/Q − k Th C o t (8) where, kTh is the Thomas rate constant (L/min.meq), qo is the maximum solid-phase concentration of the solute (meq/g), mc is the mass of adsorbent in the column (g), and Q is the volumetric flow rate (L/min). kTh and qo were calculated from the model fit to the data. The accuracy of the model fit to the experimental data was determined by applying error analysis methods such as the sum of the squared error (SSE) and root mean squared error (RMSE) described in the following equations [33].
where, n is the number of experimental data points, yc is the predicted (calculated) data, ye is the experimental data, and y represents the ratio C/Co

Desorption
Sr desorption experiments were conducted using 1 M and 2 M of NaCl and NaOH as desorbing agents. The RF resin containing the adsorbed Sr was washed with MQ water and used for the desorption experiments. As with the Sr adsorption experiment, 1 g of the RF resin containing adsorbed Sr was added to 100 mL of the desorbing solution and agitated for 4 h and then the supernatant was filtered using 0.12 mm syringe filters. The concentration of Sr in the desorbed solution was measured in the filtrate.

Characterisation of the materials
The X-ray diffraction analysis showed that the RF resin had no diffraction peaks, indicating that the resin was poorly crystalline (amorphous) as reported previously [34]. The FTIR spectrum of the RF resin is shown in Fig. 1. The broad FTIR absorption band at 2800-3500 cm -1 and the peak at 1640 cm -1 correspond to the symmetrical and anti-symmetrical stretching vibrations of water molecules with hydrogen bonding and free water, respectively.
The absorption peak occurring at 1200 -1500 cm -1 is due to OH bending of the phenolic group and the characteristic of the phenyl group. The band at 1000-1150 cm -1 is due to C-O stretching and the peak at 600-800 cm -1 is assigned to the presence of aromatic rings. The FTIR pattern with the above characteristic peaks agreed with the previous results of RF resin [34]. The ion exchange capacity of RF determined by the method of Banerjee et al. [28] was 7.0 ± 0.3 meq/g and that by the method of Samantha et al. [23] was 6.7 ± 0.2 meq/g. These values are very close to the values of 6.0 -6.9 meq/g reported by others [23,26] for the ion exchange capacity of RF.

Comparison of the adsorption capacities of different RF resins
The results of adsorption experiments with virgin RF resin and RF resins conditioned with three metal hydroxides (1 M NaOH, 1 M KOH, 1 M Ca(OH)2) showed that the NaOH conditioned RF resin had higher Sr removal efficiency than the other resins and virgin RF resin.
The Sr removal efficiencies for the RF resins conditioned by these three metal hydroxides were 95%, 50%, and 45%, respectively compared to 40% for the virgin RF resin at the dose of 1 g/L NaOH was used in the remaining adsorption experiments.
The higher adsorption capacity of the NaOH conditioned RF towards Sr could be due to the increased number of negatively charged phenolate groups generated in the NaOH system as shown by the zeta potential data (sec 3.2.3). Another study showed that conditioning RF with elevated concentrations of NaOH in the presence of constant Na + concentration maintained by a high concentration of NaNO3, increased the adsorption of Cs. This was explained as being due to increased ionisation of phenolic OH groups [23].

Comparison of different adsorbents for Sr adsorption
The adsorption capacities of four commonly used adsorbents were compared with RF in an experiment on adsorption of Sr from a solution with Sr concentration of 10 mg/L. The results showed that of the five adsorbents tested, RF had the highest adsorption capacity (Fig. 2).
Approximately 96% removal of Sr was achieved with a resin dose of 1 g/L and the removal efficiency reached 99% and 100% with resin doses of 5 g/L and 10 g/L, respectively. Under the same adsorption conditions, only Amberlite and AMP could achieve about 80% removal efficiency at doses ≥ 5g/L.

Effect of pH on Sr adsorption
The adsorption capacity of RF resin in the pH range of 2.5 to 10 revealed that pH exerted a strong effect on Sr adsorption capacity (Fig. 3a). The adsorption capacity remained constant from pH 2.5 to 4.5 and thereafter an abrupt increase was observed up to pH 7.5 and then remained constant up to pH 9.5. Others have also reported that Sr had a sharp increase in adsorption capacity around neutral to slightly alkaline pH on other adsorbents [35,36]. forces (non-specific adsorption or outer-sphere complexation [17]). However, Fig 3b also indicated that in the presence of Sr, the zeta potential became less negative above pH 5 compared to RF in the presence of a similar concentration of NaCl (10 -3 M). The zeta potential value between pH 5 and 7.5 for Sr showed negligible change (Fig. 3 (b)), yet the pH effect on Sr adsorption capacity at this pH range showed a significant increase (Fig. 3a). This suggests that Sr was adsorbed on RF by specific adsorption or inner-sphere complexation [17] and not driven entirely by the coulombic charge balance of the RF surface as dictated by outer-sphere complexation. Inner-sphere complexation may have occurred at neutral surface sites of the RF producing positive surface charges. Such surface complexation mechanism which had less correlation of Sr ++ adsorption to surface charge development dominated the bonding between Sr ++ and the FeOOH surface [35]. The positive charges formed on RF surface balanced part of the surface negative charge resulting in a reduction in negative zeta potential.
The sharp increase in adsorption from pH 4.5 to 7.5 was not considered to be due to adsorption of Sr(OH) + species or surface precipitation of Sr(OH)2 as the formation of these species or precipitation do not occur below pH 11 [35,36]. The high Sr adsorption at high pH was explained as being due to complexation of Sr with the adsorbent's surface OH groups or by the mediation of the high concentration of OHions at high pH [36]. Samanta and Misra [30] explained the higher affinity of Cs towards resorcinol formaldehyde compared to other alkali metals as a consequence of the specific interaction of Cs with the phenolic OH groups of the resin. They stated that this was due to the weak hydration of Cs compared to the strong hydration of the smaller alkali metals. This explanation can also apply to Sr which has the The low adsorption capacity of Sr at low pH is due to fewer negative charges on the RF resin surface as shown by the zeta potential data (Fig. 3b). Also, competition between Sr and the high concentration of protons could be involved in reducing the Sr adsorption capacity.
conducted at this pH range.

Batch adsorption modelling
The adsorption of Sr on RF showed that the data fitted successfully to the Langmuir model (R 2 = 0.98) (Fig. 4). The maximum adsorption capacity of 2.28 meq/g obtained from the data fit to the Langmuir model is one of the highest Sr adsorption capacity values reported for adsorbents in the literature ( Table 1). The data fit to the Langmuir model indicates that the adsorption sites on RF were homogeneous with monolayer adsorption coverage. The adsorption capacity of Sr was much higher than that of the other divalent cations, Ca and Mg and the monovalent cation K (Fig. 4). The adsorption of Ca and Mg also fitted well to the Langmuir adsorption model (R 2 of 0.97 and 0.90, respectively) but the adsorption maxima calculated from the model for Ca of 1.25 meq/g and for Mg of 1.15 meq/g were much lower than that for Sr. This is consistent with the zeta potential data where the surface negative potential on RF decreased in the order Sr > Ca > Mg > K (Fig. 3b). This order reflects the extent of inner-sphere complexation of the cations on the RF surface in addition to the charge difference of the cations (monovalent K vs the divalent ions) controlling the adsorption capacities of the cations.

Co-ions' effect on Sr adsorption
Information on the adsorption behaviour of Sr in the presence of other ions that are present in seawater is useful for the practical application of the adsorption process in recovering Sr from seawater. To obtain this information, an experiment was conducted on Sr adsorption in the presence of the major cations present in seawater such as Na, Ca, Mg and K at concentrations ranging from 10 mg/L to 40 mg/L while that of Sr was fixed at 10 mg/L. The results showed that the adsorption capacity of Sr was not affected by the presence of Na and K even at concentrations four times that of Sr (Fig. 5). However, in the presence of divalent cations, the Sr adsorption reduced by 4, 7, and 10% for Mg concentrations of 10, 20, and 40 mg/L, respectively and 8, 11, and 14% for the corresponding concentrations of Ca. This shows that the divalent Ca and Mg are competing with Sr for adsorption. Between the two ions, Ca reduced Sr adsorption more than Mg did at the same concentration expressed as mg/L which was also found for other Sr adsorbents [44]. With Ca having a higher atomic weight of 40 than Mg of 24, at the same concentration of the two ions expressed as mmol/L, Ca would be even more competitive than Mg for Sr adsorption. Zeta potential data ( Fig. 3(b)) where Ca reduced the negative zeta potential more than Mg did ( Fig. 3(b)), is consistent with the relative competion outcomes of the two ions on Sr adsorption.
The amounts of Ca and Mg adsorption at the same concentration of Sr of 10 mg/L were 0.37 meq/g and 0.90 meq/g, respectively compared to the adsorption of 0.27 meq/g of Sr (Table   S1). However, the initial concentration of Sr expressed as mmol/L was 0.23 meq/L which was  The monovalent K had less effect on Sr adsorption than the divalent Ca and Mg because of the charge difference between the ions and the inner-sphere complexation behaviour of the divalent cations (Fig. 3b). The amount of K adsorbed at the same 10 mg/L solution concentration of K was 0.09 meq/g which is much lower than the amount of Sr adsorbed (0.27 meq/g) confirming that K has less competitive effect on Sr adsorption because K was adsorbed mainly by outer-sphere complexation as shown by the zeta potential results (Table S1).

Selectivity of Sr adsorption on RF resin
The results of the selectivity experiment showed that the removal efficiency of Sr was higher than that of the other metals from solution containing all the four metals together (Fig.   6). hydration energy leads to more specific interaction with RF [30]. Zeta potential data showed that the surface negative potential on RF decreased in the order Sr > Ca > Mg > K indicating the order of the strength of adsorption (Fig. 3).

Sr adsorption in seawater medium
To investigate the Sr adsorption behaviour of the RF resin in real seawater conditions, an experiment was conducted on Sr adsorption in synthetic seawater containing Na, Mg, Ca, K, and Sr concentrations (g/L) of 20, 2, 1, 0.85, and 0.01, respectively. The RF resin demonstrated excellent efficiency of nearly 95% Sr recovery at a dose of 1 g/L in deionised water containing 10 mg/L of Sr (Fig. 7a). But the efficiency declined to about 40% in synthetic seawater medium with the same dose of adsorbent. However, the removal efficieny of other metals at the same dose of 1 g/L were < 10% (Fig. S2a) . This indicates that at the same concentration of metals in solution, Sr might be adsorbed selectively compared to other metals. At this RF dose, the amounts of Sr, Ca, Mg and K adsorption were 0.16 meq/g, 3.0 meq/g, 3.0 meq/g and 0.26 meq/g, respectively (Fig. S2b). The higher adsorption capacities of the co-ions compared to Sr

Metal ions
Adsprotion capacity

Removal efficiency
is due to the presence of other metals in the synthetic seawater at much higher concentrations than Sr. Similar results were reported with an alginate microsphere adsorbent where the Sr adsorption capacity was drastically reduced in seawater medium [21].

Desorption
Desorption of Sr from RF resin was studied to obtain information on the percentage of adsorbed Sr that can be recovered from the desorbed solution. Also, it was necessary to evaluate whether the adsorbent can be reused without any loss of Sr adsorption capacity. Sr desorption efficiency was tested with 1 M and 2 M NaCl and NaOH on a RF sample with adsorbed Sr.
The Sr desorption efficiency was found to be higher with 2 M NaCl (100%) than with 1 M NaCl (90%) when 0.1g RF with adsorbed Sr was shaken with 100 mL of NaCl for 3 h. In contrast, even a higher concentration of NaOH (2M) could desorb only 30% of adsorbed Sr (Fig. S3). The inefficiency of the NaOH to desorb Sr is because of the high Sr adsorption in alkaline condition (Fig. 3). The weakly acidic phenolic -OH groups of the resin responsible for ion exchange were expected to fully realise their ion-exchange properties only if the -OH groups were ionised to the highest extent which occurred in alkaline solutions [23]. At a lower pH produced by NaCl (pH 6.5), the adsorption of Sr is weaker and Na at the high concentration of 1 M and 2 M exchanged with Sr and the latter is effectively desorbed.

Column study
The results from the three column experiments are presented in the form of breakthrough curves in Fig from the seawater by precipitation as CaCO3 and Mg(OH)2 prior to the RF adsorption process [1]. In the first experiment, the breakthrough of Sr occurred after 417 bed volumes and increasing concentration of Sr was found in the effluent with increasing bed volumes reaching a saturation of the column with Sr after 920 bed volumes (Fig. 8a). On the other hand, in the second experiment the breakthrough commenced from the start of the experiment and column saturation occurred much earlier (Fig. 8b). The breakthrough curve was steeper than in the first experiment. This clearly shows that the co-ions at much higher concentrations than Sr as found in seawater competed with Sr adsorption on RF, thus confirming the batch experimental results.
It is interesting to note that in between 200 and 400 bed volumes, C/Co is almost constant. It The data fitted very satisfactorily to Thomas model when only Sr was present in the influent solution (Fig. 8a). The model fits to the data for Sr adsorption in the presence of other metals in the influent solution were not satisfactory (Fig. 8b, 8c), especially when Ca and Mg were present (Fig. 8b). This may be because of competition of Ca and Mg with Sr for adsorption on RF, which the Thomas model does not take into account. The error analysis of the model fits to the data showed that both the SSE and RMSE were high when other metals were present with Sr, especially Ca and Mg (

32
The breakthrough curves for the co-ions in the second experiment showed that these ions were also adsorbed simultaneously on RF in the column (Fig. S4). The C/Co values were lower for Sr compared to other metals at all bed volumes indicating that the ratio of metal adsorbed to metal concentration in solution was highest for Sr as found in the batch selectivity experiment (section 3.2.6). However, the amount of Sr adsorbed on RF calculated from the breakthrough curve (0.14 meq/g) was lower than that of the co-ions (Mg 30 meq/g and Ca 5.4 meq/g) because of the much higher concentrations of these co-ions than Sr in the synthetic seawater. Despite K had much higher concentration than Sr in synthetic seawater medium, amount of K adsorbed (0.13 meq/g) was lower than Sr. This indicates that Sr has higher selectivity for adsorption than K as found in the batch selectivity experiment (section 3.2.6).