Influence of chemical speciation on the separation of metal ions from chelating agents by nanofiltration membranes

ABSTRACT The simultaneous separation of various metal ions (nickel, copper, calcium, and iron) from chelating agents (EDTA and citric acid) in water streams using Nanofiltration membranes is analyzed. Assuming that multiply-charged species are highly rejected, chemical speciation computations reproduce the observed patterns of metal and ligand rejection at different pH values and concentrations. The separation of metal ions from citric acid is achieved in acidic conditions, where multiply-charged free metal ions and neutral or singly charged free chelating species are abundant. Overall, speciation studies help to evaluate the applicability of Nanofiltration for recycling chelating agents used for metal extraction.


TABLE OF CONTENTS LISTING
The table of contents for the journal will list your paper exactly as it appears below: Influence of chemical speciation on the separation of metal ions from chelating agents by nanofiltration membranes T. Balanyà, J. Labanda, J. Llorens, and J. Sabaté

Introduction
Several techniques that include precipitation, adsorption, ion-exchange, electrochemical treatments, and membrane filtration have been proposed to separate or concentrate heavy metals from contaminated waters 20 produced by several industries. [1][2][3][4][5] Among them, Nanofiltration is a membrane process that show high retention of multiply-charged ions but their efficiency is influenced by factors such as other ions (e.g., sodium, calcium, iron, sulfate, nitrate, and chloride), pH, and 25 ionic strength. [6,7] Further, some water streams contain chelating agents because they are used to prevent the formation of insoluble salts incrustations or to extract metal ions from liquid streams or from solids such as contaminated soil, sludge, and sediment. 30 Nanofiltration of solutions containing metal ions such as Cu, Ni, Mn, Sr, Co, rare earth elements, and radionuclides in the presence of various complexing agents (e.g., EDTA, NTA, DTPA, citric acid, NH 3 , CN -, and phosphates) using membranes with different molecular 35 weight cut-off (MWCO) values (i.e., 150-1000 Da) has been studied by several authors. [7][8][9][10][11][12][13][14][15][16] Most of them reported that the addition of a chelating agent increases metal rejection because the chelate species is larger than the free metal ions. Consequently, metal rejection rises 40 as the pH increases due to complexation between the metals and the chelating agent. Clearly, the chelating agents must be separated from the metal ions for their recovery. This process involves fragmentation of the complex, which can be achieved at low pH. 45 Additionally, Nanofiltration membranes must facilitate very distinct rejection for metal ions and ligands. Most of the aforementioned studies investigated only metal ion rejection, and little attention was given to the rejection of chelating agents.

50
In a previous work, [17] the present authors examined the separation of copper ions from citric acid in synthetic solutions using several Nanofiltration membranes of different pore size and nature. Membranes with narrow pore sizes (e.g., SelROⓇ MPF-34, Koch) simulta-55 neously rejected both the copper ions and citrate. However, both species leaked through a ceramic membrane with larger pores (i.e., CERAM INSIDEⓇ, Tami). Using a SelROⓇ; MPF-36 (Koch) at pH≈ 2, Cu 2+ and citric acid experienced high and low rejection, respec-60 tively, even though they have similar radius: 0.365 [18] and 0.344 nm, [19] respectively. According to the manufacturer, [20] this membrane has a MWCO of 1000 Da, an estimated mean pore radius of 0.86 nm, [21] an isoelectric point (IEP) of~5-6, [21] and is positively Indeed, understanding the mechanisms that control the Nanofiltration process can be helpful to increase the predictive capability and improve the efficiency of these 75 separations. The most reported mechanistic model in this regard is the Donnan, Steric and Dielectric Exclusion Pore Model (DSDE-PM), which assumes that Nanofiltration membranes are porous and can maintain an electrical charge. [22][23][24][25][26] Further, the 80 DSDE-PM considers that Steric Hindrance, Donnan Equilibrium, and Dielectric Exclusion determine the partitioning of solutes between the solutions and membrane surfaces, [27] and that the extended Nernst-Planck equation, which considers diffusion, convection, and 85 electro-migration of solutes, governs the transport across the membrane. [28] Multi-ionic solutions are a more accurate representation of real streams, and their Nanofiltration study and modeling has yielded some notable findings. [21,23,26,[29][30][31][32][33] 90 On the one hand, theoretical models explained that the presence of a multiply-charged ion, which is strongly expelled from the membrane by dielectric exclusion, increases the concentration of singly charged ions of the same sign in the membrane in order to fulfill electroneu-95 trality and the Donnan potential. Thus, the rejection of singly charged ions diminishes, and in some cases can reach a negative value. The energy required to pump these singly charged ions through the membrane interface is provided by the Donnan potential, which is generated by 100 the strongly expelled multiply-charged ions. On the other hand, the presence of very mobile ions, such as H + , reduces the electromigration of other cations and augments their rejection. [21,34] Further, the models also predict a surprising feature of ions with an electrical charge 105 opposite to the membrane: smaller counterions (with high mobility) are rejected more than larger counterions (with low mobility) with the same charge as long as the steric factor is low. [21,34] This feature can enhance the separation of heavy metals and chelating agents at low 110 pH where some Nanofiltration membranes exhibit positive electrical charge; the presence of an anion, such as Clfrom HCl added to achieve low pH, should reduce the rejection of the larger and less mobile counterions, such as some singly charged chelating anions.

115
The main objective of the present study is to investigate the simultaneous separation of several metal ions from chelating agents by Nanofiltration membranes. Further, we calculate the speciation of the solutions to filter and test its capability to explain the tendencies of 120 the observed rejections. The target synthetic solutions contain two heavy metals (nickel and copper) and either calcium or iron. This ion mixture could occur in the extract of a contaminated soil. We choose two common extracting agents (i.e., EDTA and citric acid) 125 that form chelates with very different stabilities and determine the performance of their separation and recycling ability. We filter solutions that contain several metals together because multi-solutes solutions are closer to real streams, they allow one to observe and 130 analyze the interactions and competitions between metals and it is a way to assess the conclusions raised from single metal experiments. The experimental instrumentation used herein has been described elsewhere. [21] Flat organic membranes ( stream circulated tangentially to the membrane. A previous study [17] indicated that increasing transmembrane pressure and therefore the flow rate favors the copper retention. Then a high value of transmembrane pressure (14 bar) was selected. The cross-flow velocity was high

165
(1 m s −1 ) to reduce the concentration polarization layer.- [21] The temperature was maintained at 25 ± 0.5°C. The experimental device consisted of a reservoir tank, a pump, a closed-pipe pressure dampener to prevent pressure oscillation, pressure gauges, a filtration 170 cell, and flow meters for the retentate and permeate.

Filtration runs
First, the system was filled and run with pure water. To check the state of the membrane, the membrane permeability, L p , was calculated as the ratio between the 175 permeate flow and the transmembrane pressure.
Specific quantities of solutes and water were added to obtain the prescribed concentrations. The initial feed volume was 5 L. Both the permeate and retentate were returned to the stirred feed tank for the steady-state 180 experiments, while only the retentate was returned for the batch runs. The feed pH was established by adding HCl solution dropwise and was continuously measured. After 1 h of establishing the experimental conditions, which is sufficient to achieve steady state, the permeate 185 and retentate were sampled and diluted for analysis.
Between the runs, the membranes were cleaned in situ with pure water, then with HCl or NaOH solutions, and then again with pure water.

Experimental design
190 To analyze the effects of the heavy metal concentrations, the ligand to metal ratio and the stability of chelates evolved we designed experiments to assess the separation of Ni 2+ , Cu 2+ , Ca 2+ , and Fe 3+ from two ligands ( Table 1). Ni 2+ and Cu 2+ are two heavy metals; Fe 3+ is a trivalent 195 ion that forms very stable chelates; and Ca 2+ possesses a relatively low tendency for complexation. Moreover, both Fe 3+ and Ca 2+ are usually present in soils and interfere in heavy metal extraction. [35] In these experiments, we prepared mixtures of several metals together 200 with each chelating agent and filtered them at different pH values, which were obtained by adding HCl. All experiments except #2 were performed with membranes from the same batch, which had a permeability of 3.0 × 10 -3 L m -2 s -1 bar -1 and a glucose rejection of 205 approximately 0.3. The properties of these membranes differed slightly than expected from the manufacturer's description (i.e., 1.60 × 10 -3 L m -2 s -1 bar -1 and a glucose rejection between 0.3 and 0.5). [20] The membrane used in experiment #2 came from another batch and had narrower 210 pores; its permeability was 1.90 × 10 -3 L m -2 s -1 bar -1 and its glucose rejection was 0.44.

Computation of speciation
When metals and chelating agents are mixed the resulting solution contains a wide variety of species with distinct 215 electrical charges and molecular sizes that contribute to the equilibria of several chemical reactions. The chemical speciation of the retentate was calculated using OLI Studio. [36] This software takes equilibrium constants (K a , K ps , stability constants of complexes) and redox potentials 220 compiled from literature and it theoretically predicts the thermodynamic behavior of aqueous electrolytes using the HKF (Helgeson-Kirkham-Flowers) model. [37][38][39][40] It considers activity coefficients and therefore it makes rigorous computations in a wide range of ionic strength.
225 Table 2 shows the relevant stability constants of the chelates. [41] Clearly, chelates with EDTA 4are more stable than those with citrate (Cit 3-), and the preferences of both chelating agents for the target metals follow the sequence: Partially protonated species of both EDTA and citric acid can also bind metal ions, but their affinities are not as strong as those of the completely ionized species. The successive pKa values for citric acid (H 3 Cit) are 3.13, 4.76, and 6.40; the analogous values for EDTA 235 (H 6 EDTA 2+ ) are 0.98, 1.48, 2.17, 3.12, 6.80, and 11.34.
The software inputs were pH and solutes concentrations: metal salts and ligands. The concentrations of the different species present in the retentate solution were calculated. It includes the following species: Cl − , Na + ,

SEPARATION SCIENCE AND TECHNOLOGY
240 H + , OH − , free metal ions, free ligands (EDTA or citric), metal chelates, metal hydroxide and chloride complexes and even though some polynuclear metal chelates that are marginal. All species containing metal atoms can affect metal rejection on account of their complexes 245 charge, which may differ from those of the free metal ions.
Given the multifaceted chemical speciation of the solutions studied herein, the calculations attempt to explain the tendencies of the metal and chelating agent rejections rather than make accurate predictions 250 about specific experiments. For each metal and ligand, the fraction of multiply-charges species (X) is calculated as: where C is the species concentration in the solution to be filtered. They were always lower than 1.7, 1.9, 2.8, 2.3 and 2.9% for copper, nickel, iron, calcium and TOC, respectively.
The determined values of all the analytes in retentate samples, that cover a wide range of pH and metals and 270 ligand concentrations, always differed less than 4% of the expected values. It indicates that the presence of other solutes and the formation and breaking of chelates produces little interference in the analytes determination.

Steady-state experiments without chelating agents
275 First, filtrations of metal ions without ligands were tested. Iron filtration was performed separately from the other metals because precipitation was observed above pH 3 and a colored solid accumulated on the membrane surface, which fouled and damaged it. Figure 1 shows the experi-280 mental rejections and the fraction of multiply-charged species according to the speciation calculation. Iron rejection is higher than the rejections of the other metals because of the occurrence of trivalent ions. The rejections of calcium, copper, and nickel follow similar patterns, 285 reaching a rejection value of 0.91 between pH values of 2 and 4. Above pH 4, the rejection of these metals declines due to the decrease in the membrane's positive electrical charge, since the IEP of the membrane is~5-6. [21] At very low pH, the rejections also decrease due to the high ionic 290 strength of the solution and its high Clconcentration, which screens the positive charge of the membrane. At pH 1.2, the rejection of copper (0.78) is slightly lower than that of nickel or calcium (0.81). At this pH, the speciation computations indicate that 13% of the copper present 295 occurs as the monovalent CuCl + ion, whereas nearly all of the nickel and calcium remain as Ni 2+ and Ca 2+ , respectively. Hydroxide complexes are negligible for calcium, nickel, and copper. However, the acid behavior of Fe 3+ yields significant amounts of FeOH 2+ . The computations 300 give that FeOH 2+ represents 46% and 83% of the total iron at pH values of 2.3 and 3.0, respectively. These results agree with the Fe 3+ pK a values reported in literature, that fall between 2.2 and 3.0 depending on the ionic strenght. [42] 305

Steady-state experiments with EDTA
For the highest concentration of metals, a solution containing 1 mM of Cu 2+ , Ni 2+ , and Ca 2+ and a stoichiometric concentration of EDTA was filtrated. Nearly all of the calcium was rejected (Fig. 2), but experimen-310 tal rejections of copper, nickel, and EDTA were much lower and increased slightly as the pH increased. The speciation computations showed that more than 99.8% of the calcium is present as Ca 2+ at all of the experimental conditions. At pH values of 1.8 and 2.0, nearly 315 all of the nickel and copper are chelated as monovalent

325
Experiments with lower concentrations of metals (0.1 mM) were used to analyze the interference from Fe 3+ for the Cu 2+ and Ni 2+ complexes. All experimental rejections were low and similar, which would make the separation process difficult (Fig. 3). The rejections of iron, 330 nickel, copper, and EDTA follow the same pattern; they showed a minimum at pH 1.8, which increased gradually with increasing pH from this point and increased more markedly with decreasing pH. The fraction of multiply-charges species curves of nickel, copper, and EDTA repro-335 duce this shape, although they overestimate the rejections, particularly for copper and nickel at very acidic conditions. It should be pointed out that iron and most EDTA rejections were negative. Negative rejections of ions on 340 membrane processes have been reported, explained and modeled. In some cases a negative rejection of an ion is caused by an increase of its concentration along the membrane phase, which is favored by a low electrical potential in the membrane. In electrolyte mixtures, a 345 high concentration of mobile counterions reduces the electrical potential and therefore contributes to the negative rejections of bigger and less mobile counterions. [34,43] Further, the occurrence of negative rejection was explained for slightly charged membranes 350 when Diffusivity coefficients of coions are bigger than those of counterion. [44] The speciation calculation indicates the existence of single charge big counterions such as MLand a high concentration of both chloride and proton, a more 355 mobile counterion and a very mobile coion, respectively. All these conditions together can lead MLions to undergo a negative rejection. At pH 1.8, for instance, 67%, 83%, and 85% of the iron, copper, and nickel were present in the forms of FeEDTA -, CuHEDTA -, and 360 NiHEDTA, respectively, and those species together represented 78% of the total EDTA content They could lead to the observed negative rejections of iron and EDTA, as their multiply-charged ions are not relevant. In contrast, copper and nickel do not show nega-365 tive rejections because a significant fraction of them occur as Cu 2+ and Ni 2+ , that are strongly rejected. For instance, at pH 1.2, speciation indicates that 90% of the nickel and 77% of the copper had already been released as Ni 2+ and Cu 2+ , respectively. At pH 1.8, 14% of them 370 occur as Ni 2+ and Cu 2+ .   385 for 90% and 72% of the copper at pH values of 1.8 and 2.7, respectively. The remaining copper was in the form of CuH 2 Cit + , CuHCit, CuCit -, or CuCl + . The behavior of nickel was intermediate between those of copper and calcium. The influence of pH on copper and citric acid 390 rejection agrees with the results observed in a previous study considering only Cu 2+ and citric acid. [17] We analyzed the interference from iron competition with copper and nickel through experiments with lower concentrations of metals (0.1 mM). Two ratios of the 395 chelating agent to metals were tested: the stoichiometric ratio and one with 5 times more citric acid. Figures 5  and 6 show that the resulting rejections follow the same tendency in both experiments. Citric acid rejection was low in all of the pH conditions studied. The iron rejec-400 tion pattern clearly differs from the rejection of nickel and copper. Achieving a high rejection of iron requires a pH of 1.2, whereas copper and nickel can be retained at more moderate pH.

Steady-state experiments with citric acid
Low citric rejections agree with the speciation com-405 putation, which found that multiply-charged citric acid ions (i.e., Cit 3and HCit 2-) were negligible and only neutral or singly charged ions (i.e., H 3 Cit, H 2 Cit -, FeCit, CuH 2 Cit + , CuHCit, CuCit -, NiH 2 Cit + , NiHCit, and NiCit -) show high concentrations. The X curves 410 follow the tendency of the experimental rejections. Nevertheless, the speciation computation predicts that copper and nickel chelates would fragment at higher pH values than the experimental rejections suggest. Given the high stability of the iron chelates, their frag-415 mentation and subsequent Fe 3+ release requires the addition of more acid than the release of Cu 2+ and Ni 2+ . At pH 1.2, the speciation computation predicted that the iron-citrate chelate is completely fragmented; experimentally, however, the iron rejection is lower 420 than the rejection observed in the absence of ligand, which reduces the separation factor.

Batch filtration with citric acid
To obtain additional information about the possibility of separating metals from a chelating agent, we carried 425 out a batch concentration experiment. The results of the steady-state experiments showed that it was very difficult to separate EDTA and that interference from iron is more important than interference from calcium. Therefore, a batch concentration experiment was per-    used to make the conditions for the target separation more demanding. The feed pH was 1.2 and up to 88% 435 of the initial volume was filtered. Successive aliquots of the permeate were analyzed, but only a few samples of the retentate were analyzed to avoid disturbing the system. Throughout the entire process, no significant varia-440 tion in experimental rejections of any solute was observed. In a batch filtration where the rejection remains constant, the evolution of the retentate concentration is described by Equation 2: where C and V are the retentate concentration and 445 volume, respectively, and R is the presumed rejection.
The accumulated permeate concentrations (C p ) can be calculated by Equation 3, as deduced by mass balance: The instantaneous permeate concentrations (C pi ) can be calculated by Equation 4: 450 Fitting Equation 4 to the experimental results (Fig. 7) yielded the following rejection estimates: 0.76 for nickel, 0.73 for copper, 0.68 for iron, and 0.19 for citric acid. These values are close to those obtained in the steady-state experiment with the same initial condi-455 tions: 0.78 for nickel, 0.76 for copper, 0.67 for iron, and 0.24 for citric acid. The calculated evolution of the accumulated permeate concentration is plotted in Fig. 8. After filtering 90% of the initial volume, the accumulated permeate shows a high concentration of 460 citric acid (94% of C o ). Although the metal concentrations became lower, they are still significant: 46%, 51%, and 56% of C o for nickel, copper, and iron, respectively.
These results indicate that several filtration steps would be necessary to achieve a final permeate with a low 465 metal content and a high citric acid content that can be reused. That is illustrated in the following computations.
Real separations are never absolute. The comparison of the ratio between the concentrations of chelating 470 agent and the metal ion before filtration, (C ch /C m ) o , and after filtration in the permeate, (C ch /C m ) p , gives an indication of the effectiveness of the real separation. The relationship (C ch /C m ) p /(C ch /C m ) o must be greater than unity and the greater this relationship, the greater 475 the effectiveness of the separation. After doing n successive filtrations, this relationship can be calculated by Equation 5: where R ch and R m are the rejections of chelating agent and metal ion, respectively. After performing successive 480 filtrations the volume of permeate decreases and can be calculated by Equation 6: As an example, if R ch = 0.19 (citric), R m = 0.73 (copper), and V o /V = 10, the final volume of permeate will be 90%, 81%, and 73% of the initial volume for n = 1, 2 and 3,

485
respectively. The value of (C ch /C m ) p /(C ch /C m ) o will be 1.8, 3.3 and 5.9, respectively. Then, a final reusable permeate  with a low metal and a high citric acid content, can be achieved after several filtratrion steps.
Conclusions 490 In this work, the influence of chemical speciation on the separation of metal ions (i.e., copper, nickel, calcium, and iron) from chelating agents (EDTA and citric acid) by Nanofiltration is evaluated. A semi-quantitative reproduction of the observed 495 rejection patterns of metals and ligands at different pH values and concentrations can be drawn assuming that only multiply-charged species are highly rejected. The results reveal that knowledge of chemical speciation is a helpful tool to predict Nanofiltration perfor-500 mance, even though solutions containing numerous solutes are involved in the equilibria of the chemical reactions.
The separation of metal ions from EDTA was poor due to the high chelate stability constants. Separating nickel, 505 copper, and calcium from citric acid was achieved at pH≈ 2, but pH≈ 1.2 is needed for significant iron separation. At pH≈ 1.2, rejections of nickel, copper, iron, and citric acid in batch concentration filtrations were approximately 0.761, 0.726, 0.685, and 0.191, respectively. These values are simi-510 lar to those obtained in the steady-state experiments. In those conditions, speciation calculations indicate an abundance of multiply-charged free metal ions and neutral or singly charged free chelating species.
These results indicate that the separation of multi-515 ply-charged cations and chelating agents requires a Nanofiltration membrane with low steric hindrance and a high contribution of electrostatic and dielectric interactions. Further, the membrane pore radius should be larger than the species radii, and the membrane 520 should be able to maintain an electric charge in the conditions amenable to chelate fragmentation. Overall, these considerations and results can be used to evaluate the applicability of using Nanofiltration for recycling chelating agents employed to extract heavy metals from 525 contaminated waters or solids.