Drinking water should contain certain chemicals only in limited quantities. Boron, one of these chemicals which is considered as a threatening compound and is difficult to eliminate from water. The purpose of the research is to study the major role of reverse osmosis (RO) and nanofiltration (NF) process which can contribute to the removal of this inorganic element from brackish water. For this reason, two RO and two NF membranes have been used to reduce boron and total salinity of brackish water. Since boron is a problematic compound, its elimination would also not be easy. To optimize these removal rates, two solutions based on the change of the chemical state of boron in water, either by varying the pH solution or by a complexation reaction, have been proposed in this work. Obtained results show that for both solutions, the boron removal percentage has clearly improved and reached 95% for the first case and 70% for the second. It is worth noting that the Spiegler–Kedem model was applied to fit the found experimental results.

INTRODUCTION

With the increasing demand of good quality water and the reduction of freshwater resources, human beings must turn to alternative solutions such as desalination of sea water and brackish water. When using these alternative resources, several pollutants begin to appear. These pollutants include boron which is an inorganic pollutant, and is problematic and difficult to remove from water. The name of boron comes from the borax compound (sodium borate). It was discovered for the first time by Gay-Lussac and Louis Jacques Thenard in 1809. It was identified as a chemical element by Jons Jacob Berzelius in 1824.

Boron is a trivalent metalloid which is very abundant in the environment. It is found naturally in over 80 minerals and constitutes 0.001% of the Earth's crust. It is widely distributed in nature with concentrations of about 10 mg/kg in the Earth's crust and from 4 to 5 mg/L in ocean waters (Tanaka & Fujiwara 2008). The amount of boron in fresh water depends on factors such as the proximity of the coastal regions, the contribution of industrial and municipal effluents, and the geochemical nature of the watershed. Water with high boron content is located in areas that contain sediments and sedimentary rocks rich in clay.

In aqueous solutions, boron is mainly in the form of boric acid. Consideration of the dissociation constant of boric acid (pKa = 9.14 to 25 °C) suggests that at neutral or acidic pH, it is the H3BO3 form which predominates, while in the high pH it is the form which prevails. Both forms exist in equilibrium in a pH range of 7.0 to 11.5.

Humans can be exposed to the boron via fruits, vegetables, water, air and other consumer products. This element is associated with many metabolic activities such as the metabolism of sexual hormones and certain nutrients such as calcium, magnesium, phosphorus, molybdenum and vitamin D (Devirian & Volpe 2003). It is necessary for bones growth (Devirian & Volpe 2003). When a person consumes a huge quantity of water or food containing boron, its concentration in the body can lead to tragic consequences. Boron can infect the stomach, liver, kidneys and brain and can eventually lead to death (WHO 2011). The presence of very low quantities of boron appears to be required for all plants, but at higher concentration, it is highly toxic for vegetation. An excess of this element in irrigation water can be responsible for the appearance of spots on fruits and vegetables (Tanaka & Fujiwara 2008). According to medico-biological investigations, the boron compounds belong to the second class of the toxicological hazard (WHO 2011). The advanced research concerning the boron and its toxicity allow evolving the standards and guidelines for the maximum allowable concentration in water intended for human consumption. The adverse effects of boron on human health have been assessed by the World Health Organization (WHO). In 1993, a maximum concentration of 0.3 mg/L of boron was established. In 1998, this value has been revised and brought back to 0.5 mg/L (Tu et al. 2010). A new guideline value of 2.4 mg/L was adopted in 2011 (WHO 2011). At a global scale, boron standards in drinking water vary widely from 0.5 to 4 mg/L depending on the region (Tu et al. 2010).

Currently, boron removal from water is needed in many situations and has been the subject of many studies (Redondo et al. 2003; Jocob 2007; Koseoglu et al. 2008; Dydo et al. 2012). But until now, no method has proved to be satisfactory. Among the most used methods, experiments of elimination of boron in the water have been conducted by implementing the reverse osmosis (RO) and nanofiltration (NF) processes. Indeed, several characteristics make these techniques more efficient and more economical compared with other water treatment processes.

Over the past few decades, the rapid expansion in desalination sector imposes the importance of boron removal. Although the RO and NF membranes deliver a significant efficacity in salts reduction, the efficiency of boron removal by existing trade membranes remains relatively low (Taniguchi et al. 2004; Koseoglu et al. 2008; Tu et al. 2010). To solve this problem, various studies have focussed on the synthesis of high boron rejection membranes by altering the active layer or by adding another layer of chemical coating (Comstock 2009; Du et al. 2015). This option remains limited because changing the active layer leads to a very narrow structure and causes a significant decrease of permeability and amount of water produced. This work would not focus on the active layer modification, but it is intended for the change of the chemical state of boron in the water to be treated with the aim of optimizing their elimination percentages.

In the present study, optimizing boron removal by NF and RO would be investigated using two RO membranes (AG and SG) and two NF membranes (NF-90 and HL). The performance of these membranes in terms of boron reduction has been studied with brackish water from Oued Boudhebane underground in the city of El'Fahs in Tunisia. Results of this study will lead to the selection of suitable membranes for the treatment of brackish water and to assist the best choice for large scale application. The efficiency of desalination and boron removal by the studied membranes was evaluated at different transmembrane pressure. The Spiegler–Kedem model was used to check the experimental results and to calculate the phenomenological parameters, that is, the reflection coefficient (σ) and the solute permeability (Ps) of the membrane to the brackish water.

MATERIALS AND METHODS

RO and NF experiments

A mini pilot-scale RO/NF system was used for treatment of brackish water. It includes, as shown in Figure 1, a feed tank, a cartridge microfilters (5 μm), valves, manometers, a high pressure pump (NU.ER.T) and RO/NF modules. Operation was in a recycle mode, that is, the concentrate as well as permeate was recycled to the feed tank.
Figure 1

Descriptive scheme of the RO/NF system.

Figure 1

Descriptive scheme of the RO/NF system.

Four commercial RO/NF thin-film composite membranes were used in this study. The RO membranes were AG 2514 and SG 2514 and the NF membranes were HL 2514 and NF-90 2540. Characteristics of the above-mentioned membranes as given by producers are presented in Table 1.

Table 1

Characteristics of the four membranes used in this study

MembranesAGSGNF-90HL
ManufacturerGE OsmonicsGE OsmonicsDow FilmTecGE Osmonics
Surface area (m20.6 0.6 2.6 0.6
Module Length (inch) 14 14 40 14
Diameter (inch) 2.5 2.5 2.5 2.5
Maximum pressure (bar) 30 41 41 30
Maximum temperature (°C) 50 50 45 50
Operating pH range 2–11.5 2–11.5 2–11 2–11.5
Salts rejection (%) 99.5a 98.5a >97b 98b
MembranesAGSGNF-90HL
ManufacturerGE OsmonicsGE OsmonicsDow FilmTecGE Osmonics
Surface area (m20.6 0.6 2.6 0.6
Module Length (inch) 14 14 40 14
Diameter (inch) 2.5 2.5 2.5 2.5
Maximum pressure (bar) 30 41 41 30
Maximum temperature (°C) 50 50 45 50
Operating pH range 2–11.5 2–11.5 2–11 2–11.5
Salts rejection (%) 99.5a 98.5a >97b 98b

aNaCl rejection.

bNa2SO4 rejection.

Before each experiment, the membranes were cleaned with pure water (<1 μS/cm), during half an hour, at an operating pressure of 5.0 bar. The pure water permeability (Lp) was determined by statistical linear regression of permeate flux Jv versus transmembrane pressure ΔP. The linear evolution is described by a slope corresponding to the pure water permeability.

The mean values of Lp were 3.45 L/h/m2/bar for AG, 3.88 L/h/m2/bar for SG 11.45 L/h/m2/bar for NF-90 and 9.01 L/h/m2/bar for HL.

The performance of the tested membranes was measured in terms of rejection (R) and flux (Jv). Membrane rejection is calculated by Equation (1):
1
where Cp and C0 (mol/L) are permeate and feed concentrations, respectively.
Permeate flux Jp (L/h/m2) was calculated as follows:
2
where V (L) is the volume of permeate collected in a given time interval t (h) and A (m2) is the membrane area.

To determine the performance of each membrane, before and after optimization, 50 L of brackish water are used. For each test, three measurements were taken; average values were calculated.

Spiegler–Kedem model

The solute transfer through the RO and NF membranes can be described by using the principles of irreversible thermodynamics. For a two component system, consisting of water and a solute, the irreversible thermodynamics approach leads to two basic equations:
3

4
where Jv and Js are the water flux and solute flux, C0, Cm, and Cp represent, respectively, the concentrations of solute in the initial effluent, in the membrane and in the permeate. Ps and σ are the permeability of solute and reflection coefficient of the membrane, respectively. Lp represents the hydraulic permeability. ΔP and Δπ defined, respectively, the transmembrane pressure and the difference of osmotic pressure between each side of the membrane.
With constant fluxes and constant transport parameters, integration of Equation (4) on the membrane thickness, in terms of the real salt rejection, gives the following rejection expression:
5
with:
6
According to the film theory, the relationship between the observed rejection rate (Rob) and the true rejection (R) may be expressed as:
7
where k is the mass transfer coefficient.
Substitution of Equation (5) into Equation (7) and rearranging results in the following equation:
8

This equation can be used to estimate the transfer parameters σ and Ps and the mass transfer coefficient k and subsequently to determine the theoretical results of retention rates. In this paper, the curves giving boron rejection depending on operating pressure or permeate flux were adjusted by the Spiegler–Kedem model. Experimental data used are marked as solid symbols, whereas dash lines represent the Spiegler–Kedem model.

Brackish water characterization

A real application of NF and RO process was performed on brackish water, containing boron, which is taken from Oued Boudhebane underground in the city of El'Fahs in Tunisia. The characteristics of the brackish water after treatments by sand and microfiltrations are shown in Table 2. Three measurements were taken; average values were calculated and noted in this table.

Table 2

Physicochemical proprieties of Oued Boudhebane water

 Salinity (mg/L) 3,710 Sodium (mg/L) 1,087 Potassium (mg/L) 33.2 Calcium (mg/L) 88 Magnesium (mg/L) 125 Chloride (mg/L) 1,406 Fluoride (mg/L) 33 Nitrate (mg/L) 0.51 Sulfate (mg/L) 367 Boron (mg/L) 3 Turbidity (NTU) <1 pH 7.62
 Salinity (mg/L) 3,710 Sodium (mg/L) 1,087 Potassium (mg/L) 33.2 Calcium (mg/L) 88 Magnesium (mg/L) 125 Chloride (mg/L) 1,406 Fluoride (mg/L) 33 Nitrate (mg/L) 0.51 Sulfate (mg/L) 367 Boron (mg/L) 3 Turbidity (NTU) <1 pH 7.62

CHEMICAL AND ANALYTICAL METHODS

All chemicals reagents were analytical graded and used without further purification. The pH change was effected by addition of NaOH (99.8%, Merck) solutions.

Boric acid and borate compounds are capable of reacting with complexing agents containing multiple hydroxyl groups (polyols) forming stable complexes. The choice of complexing was carried out according to the availability and harmlessness. Two reagents were tested: mannitol and sorbitol. These are two isomers of molecular formula C6H14O6. They are a part of alcohols having hypertonic and sweetening qualities. They have a lower sweetening power than regular sugar.

The concentration of boron in the samples was determined by molecular absorption spectrometry in the UV–visible range using Azomethine-H. Other parameters such as pH and salt concentrations were measured by pH-meter Orion 2 Star and ionic chromatography Metrohm, respectively.

RESULTS AND DISCUSSION

Brackish water permeability

The permeate flux is an important parameter in the design and economic feasibility of NF and RO processes. When the elimination of solutes is performed, it is a fundamental factor in optimization process. The experimental data for the permeate flux, with brackish water, as a function of the operating pressure are given in Figure 2 for the tested NF and RO membranes.
Figure 2

Dependency of permeate brackish water flux on operating pressure for NF and RO membranes (T = 27 °C).

Figure 2

Dependency of permeate brackish water flux on operating pressure for NF and RO membranes (T = 27 °C).

The permeate flux increases linearly with the applied motive force, especially the pressure, indicating that the increase of the applied pressure will improve the water driving force inside the membrane and subsequently will defeat the membrane resistance.

The permeability of the studied membranes follows this order: HL> NF-90> SG> AG. The values of pure water permeability (Lp) and brackish water permeability (L′p) and the critical pressure (Pc) for the four membranes are reported in Table 3, where the critical pressure is the efficient pressure for which the first droplets of permeate solution observed.

Table 3

Values of pure water permeability (Lp), brackish water permeability (L′p) and critical pressure (Pc) for the tested membranes

MembranesLp (L/h/m2/bar)L′p (L/h/m2/bar)Pc (bar)
AG 3.45 1.59 3.52
SG 3.88 1.87 3.35
NF-90 11.45 6.14 1.6
HL 9.01 6.81 1.06
MembranesLp (L/h/m2/bar)L′p (L/h/m2/bar)Pc (bar)
AG 3.45 1.59 3.52
SG 3.88 1.87 3.35
NF-90 11.45 6.14 1.6
HL 9.01 6.81 1.06

It is observed from Table 3, the saline water permeability (Lp) is lower than pure water permeability (Lp). The presence of the salts in the water makes the membrane surface more compact due to contraction of pores, resulting in the decreasing in the permeability through the membranes (Huang et al. 2008).

Besides, the NF membranes may operate at low transmembrane pressure (1 bar for the HL membrane and 1.5 bar for the NF-90 membrane), while the two RO membranes begin their products at a higher pressure (3 bar). Indeed, the NF membranes have the advantage of offering a partial demineralization in correlation with a weak effect of the osmotic pressure compared with the RO membranes due to the small difference in concentration between the two sides of the membrane.

Brackish water desalination

This study is performed to assess the effectiveness of studied membranes for the desalination of brackish water and to select those that can reach the salt rejection required at low pressures and with the higher permeate flux. The effectiveness of four membranes for the desalination of brackish water is given in Figure 3 which represents salts retention depending on the transmembrane pressure at a conversion rate of 25%.
Figure 3

Dependency of salinity retention on operating pressure for NF and RO membranes (Y = 25%, T = 27 °C).

Figure 3

Dependency of salinity retention on operating pressure for NF and RO membranes (Y = 25%, T = 27 °C).

As expected, the RO membranes have shown the best performance. Their rejections of salts are the highest and almost identical for both membranes tested. The HL membrane shows lower rejection of salts conversely to the NF-90 membrane which shows higher rejection of salts. The following order of salts retention was observed for the four membranes, AG > SG > NF-90 > HL. The molecular weight cutoff (MWCO) values of these membranes can be used to express this sequence. The MWCO values of the AG, SG, NF--90 and HL membranes are respectively 145 Da, 173 Da, 163 Da and 314 Da (Bejaoui et al. 2011; 2014,Tabassi et al. 2013; Mnif et al. 2015). Generally, the MWCO of the NF membranes due to the wider pores are responsible for the lower salts retention such as the case of HL membrane. The NF-90 membrane has rather small MWCO, close to the RO side of the NF region. For this reason, its salts retention proportions are similar to the RO membranes.

Four commercial RO/NF thin-film composite (TFC) membranes were used in this study. The used membranes, such as the majority of TFC polyamide membranes, are negatively charged, as announced by Norberg et al. (2007). They studied the surface characterization of AG, SG, NF-90 and HL membranes and they found that zeta potentials of these membranes are, respectively, −10.8, −7.6, −14.2 and −8. According to these zeta potential values, the NF-90 membrane is more negatively charged compared to the HL NF membrane. This result confirms its high salts retention. The same result was shown using the AG membrane which is more negatively charged than the SG RO membrane.

The Spiegler–Kedem model was applied to be compatible with the desalination of brackish water. The phenomenological parameters (σ, Ps) for the four membranes are summarized in Table 4. Furthermore, the variation of the salinity retention depending on the transmembrane pressure is adjusted by the theoretical Spiegler–Kedem model and is shown in Figure 3.

Table 4

Salinity transfer parameters (σ, Ps) present in brackish water

MembraneAGSGNF-90HL
σ 0.983 0.975 0.962 0.532
Ps (L/h) 0.138 0.214 1.981 36.273
MembraneAGSGNF-90HL
σ 0.983 0.975 0.962 0.532
Ps (L/h) 0.138 0.214 1.981 36.273

The σ and Ps values depend on the type of the membrane. The AG, SG and NF-90 membrane present higher σ values and lower salt permeability followed by the HL membrane. It seems that the retention behavior of NF-90 is comparable with that of the RO membranes. A good fit was obtained for the retention values of tested membranes for brackish water used.

Boron removal from brackish water

This brackish water contains 3 mg/L of boron; a concentration which exceeds the limits fixed by the World Health Organization and the majority of international guidelines. The boron removal selectivity for the NF and RO membranes depending on the transmembrane pressure was studied and is shown in Figure 4. The found experimental results are justified by the Spiegler–Kedem model.
Figure 4

Dependency of boron retention on operating pressure for NF and RO membranes (Y = 25%, T = 27 °C).

Figure 4

Dependency of boron retention on operating pressure for NF and RO membranes (Y = 25%, T = 27 °C).

From this figure, we can notice that the boron removal by the four studied membranes does not exceed 40%. At a fixed conversion rate and operating pressure, the order of boron retention is the following: AG > SG > NF-90 > HL, the same sequence obtained for the elimination of salinity.

The results of Spiegler–Kedem model summarized in Table 5 fit well with the experimental curves. Concerning the values of Ps, the higher quantity of boron passing through the membrane corresponds to the HL membrane, while the lowest amounts correspond to the two RO membranes AG and SG. These values offer a good verification of the experimental data of boron removal by all membranes.

Table 5

Boron transfer parameters (σ, Ps) present in brackish water

MembraneAGSGNF-90HL
σ 0.391 0.407 0.429 0.579
Ps (L/h) 4.76 3.25 33 472
MembraneAGSGNF-90HL
σ 0.391 0.407 0.429 0.579
Ps (L/h) 4.76 3.25 33 472

To check if the water produced by these four membranes complies with the boron standards fixed by the World Health Organization and the European Union (EU), the variation of boron concentration in the permeate depending on the transmembrane pressure as well as the boron standards set by the EU and WHO are represented in Figure 5.
Figure 5

Dependency of boron concentration in permeates of the four membranes on operating pressure (Y = 25%, T = 27 °C) and standards set by EU and WHO.

Figure 5

Dependency of boron concentration in permeates of the four membranes on operating pressure (Y = 25%, T = 27 °C) and standards set by EU and WHO.

The waters produced by the AG, SG and NF-90 membranes meet the WHO standard especially the boron concentration. However, with these operating conditions, the standard required by the EU is still away. The aim of this work is to achieve this standard, and therefore, looking for the boron removal optimization for the used membranes.

Optimization of boron removal

pH modification

As reported previously, with the four membranes used and at a neutral pH condition, the boron rejection remains low. Increasing the pH by the addition of sodium hydroxide solution, must be performed to improve boron rejection by favoring the dissociation of the boric acid which is a weak acid (pKa = 9.2). Rodriguez et al. (2001), Redondo et al. (2003) and Tomaszewska & Bodzek (2013) show in their research that the boron removal by RO and NF membranes is strongly dependent on the pH; it increases as a function of pH. These studies also report that boron is effectively eliminated at pH values between 10 and 11. Increasing of pH is effective in terms of economics; it sometimes fails when water is loaded with salts. For these waters, the adjustment of pH between 10 and 11 may affect the durability of the membranes threatened by the clogging phenomenon by the deposits of insoluble carbonates, hydroxides, and mixed salts. To reform this limitation, an integrated system with a RO or NF process to a two-stage process has been proposed. The pH adjustment is performed at the end of the first step. Figure 6 shows the experimental design used in this application. This experimental apparatus comprises two modules with same reference. The first reduces the amount of salts present in the water. At the outlet of the first module, a sufficient quantity of NaOH is added to reach the optimum pH for removal of boron. The produced water is sent to a second module in order to eliminate the remaining amount of boron in the water. The membrane HL is excluded because of its low desalting rates. The permeate of each of three membranes (AG, SG and NF-90), obtained at 10 bar and with a conversion rate of 50%, is sent to another membrane with the same reference after a pH adjustment.
Figure 6

Experimental design used for this application.

Figure 6

Experimental design used for this application.

Figure 7 illustrates the variation of boron concentration after the second treatment depending on pH. The pressure is fixed at 10 bar and the total conversion rate is 25%.
Figure 7

Dependency of boron concentration in water produced after second step on operating pressure (AG, SG and NF-90 membranes, Y = 25%, P = 10 bar and θ = 27 °C) and standard set by EU.

Figure 7

Dependency of boron concentration in water produced after second step on operating pressure (AG, SG and NF-90 membranes, Y = 25%, P = 10 bar and θ = 27 °C) and standard set by EU.

As shown in Figure 7, by increasing the pH, a boron concentration in water produced can reach 0.1 mg/L at pH of 11 with the total elimination percentages equal to 97% for the AG membrane, 95% for the SG membrane and 96% for the NF-90 membrane. From pH = 9, the EU standard is met for all membranes.

Boron complexation

Boric acid and borate compounds can react with chemical compounds containing multiple hydroxyl groups (polyols), such as mannitol and sorbitol, to produce anionic complexes which can be easily removed by RO and NF membranes (Geffen et al. 2006). Boric acid is a weak acid; the complexation with polyols compounds intensifies its acidity. This method has been used for many years for the quantitative analysis of boric acid that cannot be detected by direct titration. The complexation of boron with polyols has been used successfully for boron removal by ion-exchange processes (Simonnot et al. 2000; Jacob 2007). However, this solution has not been frequently used in RO and NF technologies. In a recent study, Geffen et al. (2006) argued the addition of mannitol in the feed solution to increase the boron rejection by complexation with mannitol compound. Dydo et al. (2012) has shown that the use of N-methylglucamine leads to an improved boron rejection by RO membranes.

Complexation equilibrium of boric acid with polyols compounds is dependent on pH. Indeed, there are four mechanisms (Figure 8); the first two correspond to the complexation of boric acid (K1 and K2), while the remaining two correspond to the complexation of the borate ion (K3 and K4).
Figure 8

Complexation equilibrium of boric acid/borate with the polyol compounds.

Figure 8

Complexation equilibrium of boric acid/borate with the polyol compounds.

The stability constants of the borate complex (K3 and K4) were determined by several studies (Van Duin et al. 1984; Makkee et al. 1985), while those of the boric acid complexes are still not available in the literature. These studies have demonstrated that the constant K3 (1,060 for mannitol and 6,840 for sorbitol) is considerably higher than K4 (150 for mannitol and 80 for Sorbitol). It is worth noting that K2 and K4 would be higher when the molar concentration of mannitol is greater than that of boron. Thus, the nature of the complexes formed, depends on the complexing agent used, molar ratios complexing/boron (nC/nB) and pH.

Experiments of the complexing agent determination and the optimal dose were performed on samples of brackish water. Different doses of mannitol and sorbitol were added. The effect of molar ratio on the elimination percentages of boron by the studied membranes is presented in Figure 9. The ratio (nC/nB)) varies from 0 to 4, the pressure at 10 bar and the conversion rate at 25%.
Figure 9

Dependency of boron retention on molar ratios nC/nB for NF and RO membranes (P = 10 bar, Y = 25%).

Figure 9

Dependency of boron retention on molar ratios nC/nB for NF and RO membranes (P = 10 bar, Y = 25%).

Figure 9 shows that the addition of complexing agents improves the removal of boron by the four membranes. By adding 25 mg/L of mannitol (nC/nB = 0.5), the boron removal rates have increased from 7.5 to 17% for the HL membrane, from 32 to 39% for the NF-90 membrane, from 32 to 43% for the SG membrane and from 34 to 45% for the AG membrane. For the other doses, there is not a great difference concerning the boron removal rates. Moreover, the boron removal rates, in the presence of sorbitol, are highest compared with mannitol. This may be explained by the fact that the sorbitol complexes are more stable (K3(sorbitol) > K3(mannitol)). Same dose of sorbitol (25 mg/L) is effective in achieving the removal percentages of 20%, 48%, 51% and 52%, respectively, by the four membranes HL, NF-90, SG and AG.

Thus, a small amount of sorbitol 25 mg/L may be sufficient to achieve a significant development of boron rejection by the four membranes. According to the market overview given by CIHI (www.icis.com/chemicals/sorbitol/), sorbitol price in the last quarter of 2014 was $0.65 to$0.75 per kilogram. Therefore, the addition of polyols such as sorbitol to improve boron rejection in natural waters can be a practical approach to the boron removal.

The pH is a determining factor that affects both the surface of the membrane and the distribution of species in solution. Despite not having information on the stability of the boric acid-polyols complex, stable complexes can be formed with the borate ions. On the other hand, polyols increase the acidity of the boric acid, that is to say decrease the pKa value which surrounds it to 9.14. Therefore, a slight increase in pH can contribute to the development of boron removal rates. Figure 10 shows that increasing pH from 7.6 to 8.5 could clearly improve the boron removal efficiency (by about 20%) to reach 70%, 69%, 68% and 38%, respectively, for the AG, SG, NF-90 and HL membranes.
Figure 10

Variation of boron reduction rates by the four membranes for three pH values and at a dose of 25 mg/L of sorbitol (P = 10 bar, Y = 25%, T = 27 °C).

Figure 10

Variation of boron reduction rates by the four membranes for three pH values and at a dose of 25 mg/L of sorbitol (P = 10 bar, Y = 25%, T = 27 °C).

The selectivity of the AG, SG, NF-90 and HL membranes to the removal of boron present in brackish water, after complexation with the sorbitol, was studied as a function of the transmembrane pressure and is shown in Figure 11. The transfer model Spiegler–Kedem was applied to fit the reduction of boron results after complexation. Phenomenological parameters (σ, Ps) are given in Table 6. The variation of boron retention rates after complexation, depending on the transmembrane pressure for studied membranes is adjusted by the theoretical model of Spiegler–Kedem and shown in Figure 11.
Table 6

Boron transfer parameters (σ, Ps) present in brackish water after complexation

MembraneAGSGNF-90HL
σ 0.72 0.72 0.76 0.56
Ps (L/h) 1.06 1.64 11.3 49
MembraneAGSGNF-90HL
σ 0.72 0.72 0.76 0.56
Ps (L/h) 1.06 1.64 11.3 49
Figure 11

Variation of the boron removal rates by as a function of the transmembrane pressure for the four membranes (25 mg/L of sorbitol, pH = 8.5 and Y = 25%).

Figure 11

Variation of the boron removal rates by as a function of the transmembrane pressure for the four membranes (25 mg/L of sorbitol, pH = 8.5 and Y = 25%).

The model will verify the experimental data of boron removal after complexing with all membranes. The highest values of σ were obtained for the AG, SG and NF-90 membranes, and with reverse proportionality, with regard to Ps values. Comparing the results obtained without complexing and after addition of sorbitol, it is clear that the complexation of boron increases its retention by the four membranes (increase of σ) and decreases its passage through the membranes (decrease of Ps).

CONCLUSION

The study of boron removal from the brackish water (Oued Boudhebane) using the AG, SG, HL and NF-90 membranes shows that the boron retention rate depends strongly on the type of the membrane. The AG membrane has a slightly higher effectiveness compared with membranes SG and NF-90. On the basis of the dissociation equilibrium of boric acid, better removal of boron requires a pH between 10 and 11. This pH value represents a problem when applied to this brackish water. Insoluble deposits of carbonates, hydroxides and mixed salts may be formed and facilitate the membrane clogging. Two solutions have been proposed to facilitate the elimination of boron from this brackish water without touching the sustainability of the tested membranes. The first solution consists of using a device of two AG, SG or NF-90 membranes and the pH adjustment is performed after the reduction in salinity by passage through the first membrane. In this case, the boron removal percentages using the AG, SG and NF-90 membranes exceed 95%. The EU standard is met for all membranes at pH = 9, P = 10 bar and Y = 25%. The second solution is to add the sorbitol as complexing agent. The addition of 25 mg/L of sorbitol and the changing of pH from 7.6 to 8.5 increase the percentage of boron removal from 35 to 70% for the AG, SG and NF-90 membranes and from 10 to 43% for the HL membrane.

REFERENCES

REFERENCES
Bejaoui
I.
Mnif
A.
Hamrouni
B.
2011
.
Desalination and Water Treatment
29
(
1–3
),
39
46
.
Bejaoui
I.
Mnif
A.
Hamrouni
B.
2014
.
Separation Science and Technology
49
(
8
),
1135
1145
.
Comstock
D. L.
2009
Method of Treating Reverse Osmosis Membranes for Boron Rejection Enhancement
.
North Pacific Research
,
USA
.
Devirian
T. A.
Volpe
S. L.
2003
.
Critical Reviews in Food Science and Nutrition
43
(
2
),
219
231
.
Du
X.
Meng
J.
Xu
R.
Shi
Q.
Zhang
Y.
2015
.
Journal of Membrane Science
467
,
205
215
.
Dydo
P.
Nems
I.
Turek
M.
2012
.
Separation and Purification Technology
89
,
171
180
.
Geffen
N.
Semiat
R.
Eisen
M. S.
Balazs
Y.
Katz
I.
Dosoretz
C. G.
2006
.
Journal of Membrane Science
286
(
1
),
45
51
.
Huang
R.
Chen
G.
Sun
M.
Gao
C.
2008
.
Separation and Purification Technology
58
,
393
399
.
Jacob
C.
2007
.
Desalination
205
(
1–3
),
47
52
.
Koseoglu
H.
Kabay
N.
Yüksel
M.
Sarp
S.
Arar
O.
Kitis
M.
2008
.
Desalination
227
(
1
),
253
263
.
Makkee
M.
Kieboom
A. P. G.
Van Bekkum
H.
1985
.
Recueil des Travaux Chimiques des Pays-Bas
104
(
9
),
230
235
.
Mnif
A.
Tabassi
D.
Ben Sik Ali
M.
Hamrouni
B.
2015
.
Environmental Progress and Sustainable Energy
34
(
4
),
982
989
.
Norberg
D.
Hong
S.
Taylor
J.
Zhao
Y.
2007
.
Desalination
202
,
45
52
.
Redondo
J.
Busch
M.
De Witte
J. P.
2003
.
Desalination
156
(
1–3
),
229
238
.
Rodríguez
P. M.
Ferrándiz
R. M.
Chillón
F.
Prats
R. D.
2001
.
Desalination
140
(
2
),
145
152
.
Simonnot
M. O.
Castel
C.
Nicolai
M.
Rosin
C.
Sardin
M.
Jauffret
H.
2000
.
Water Research
34
(
1
),
109
116
.
Tabassi
D.
Mnif
A.
Hamrouni
B.
2013
.
Desalination and Water Treatment
52
(
7–9
),
1792
1803
.
Tanaka
M.
Fujiwara
T.
2008
.
Pflügers Archiv-European Journal of Physiology
456
(
4
),
671
677
.
Taniguchi
M.
Fusaoka
Y.
Nishikawa
T.
Kurihara
M.
2004
.
Desalination
167
,
419
426
.
Tomaszewska
B.
Bodzek
M.
2013
.
Desalination
319
,
99
106
.
Tu
K. L.
Nghiem
L. D.
Chivas
A. R.
2010
.
Separation and Purification Technology
75
(
2
),
87
101
.
Van Duin
M.
Peters
J. A.
Kieboom
A. P. G.
Van Bekkum
H.
1984
.
Tetrahedron
40
(
15
),
2901
2911
.
WHO
2011
Guidelines for Drinking Water Quality, 4th edn
,
World Health Organization, Geneva, Switzerland
.