Removal of Nitrate and Nitrite by Donnan Dialysis: Optimization According to Doehlert Design

Abstract

The removal of nitrate and nitrite simultaneously was investigated by Donnan dialysis (DD) using a Response Surface Methodology (RSM) approach. DD is a membrane process that consists of cross-ion exchange having the same electric charge through an ion-exchange membrane separating two solutions. In addition to being easy to handle, DD process is continuous, economical, requiring only few chemicals and low pumping energy. Statistical tools were applied to investigate the simultaneous removal of nitrates and nitrites by DD. The RSM is an efficient statistical strategy to design experiments, build models, determine the optimum conditions, and evaluate the significance of factors, even the interaction between them. A preliminary study was performed with three commercial membranes (AFN, AMX, ACS) in order to determine the experimental field based on different parameters. Then, a full-factor design was developed to determine the influence of these parameters and their interactions on the removal of nitrates and nitrites by DD. The RSM was applied according to the Doehlert model to determine the optimum conditions. The use of the RSM can be considered a good solution to determine the optimum condition compared to the traditional “one-at-a-time” method.

1. Introduction

The growth of population, industrialization, and rapid urbanization increases the pollution of nitrogen (N). Nitrate and nitrite are often used as pollutants; they present a water-quality problem [1]. Nitrate is primarily responsible for water eutrophication and infectious diseases in the environment [2]. In comparison to nitrate, the nitrite has a higher level of toxicity to human health. In the body, methemoglobin can be formed when it combines with hemoglobin, reducing oxygen transport. Additionally, this compound can be converted into carcinogenic nitrosamine, which is linked to hypertension, leukemia, brain tumors, stomach cancers, and bowel cancers [3]. The World Health Organization (WHO) suggests a maximum nitrate content of 50 mg/L in irrigation water and 0.5 mg/L in drinking water to protect the public health from the harmful effects of high nitrate and nitrite concentrations [4].

It was possible to remove nitrate and nitrite chemically using chlorination and physicochemically using coagulation-flocculation [5]. In contrast to traditional biological approaches [6], bioadsorption [7], and photocatalytic denitrification [8], most methods require oxidation in order to remove nitrite. Removing nitrate and nitrite simultaneously avoids this oxidation step, which requires chemical products to turn nitrite into nitrate.

Donnan dialysis (DD) was chosen because it is mostly economical. Because it only needs, a small amount of chemicals, pumping energy, and is simple to operate; this process is continuous and inexpensive. In the DD process, an ion-exchange membrane separates a compartment containing the solution to be treated (feed) from a compartment containing the solution that receives the target ions (receiver). The concentration gradients of the ions transported through the membrane (counter-ions) control the ion-exchange kinetics [9, 10]. The DD process involves the stoichiometric exchange of counter-ions, or ions with the same charge, over an ion-exchange membrane, and it is ended when Donnan equilibrium is attained [11].

As is common knowledge, the DD process is used to purify, concentrate, and remove various ions from wastewater and industrial effluents, including boron [12, 13], fluoride [14, 15], chromium [16, 17] and nitrates, nitrites [18, 19] using different type of anion-exchange membranes. Although Donnan dialysis has certain advantages in terms of cost and energy efficiency, industry does not adopt it primarily due to its slow kinetics. The almost of applications have been studied at laboratory scale.

The typical “one factor at a time” method of optimizing multivariate systems is not only time-consuming, but also often does not take into account the effects of cross-interactions between experimental factors. Furthermore, this approach implies that the best levels must be determined by multiplying experiments, which is not always true. By integrating Doehlert’s experimental design [20] with response surface methods to simultaneously optimize all influential parameters, these drawbacks of the single-factor optimization procedure can be avoided.

The Box and Wilson-developed Response Surface Methodology (RSM) is a set of mathematical and statistical methods for studying situations like the one that is being posed using an empirical model. The RSM is a useful tool for process optimization. The benefit of this method over the traditional one is that it takes less time and costs less. Doehlert designs have a number of advantages over other designs, such as central composite or Box-Behnken designs. Because the number of levels can vary from one variable to another, there is greater flexibility in assigning a high or low number of levels to the selected variables, saving time on studies. Additionally, adjacent hexagons may successfully occupy a space because they do not overlap, making Doehlert designs more effective at mapping space [20].

The present investigation presents the application of the RSM applying Doehlert experimental design studies to look into the simultaneous removal of nitrates and nitrites by Donnan dialysis. In order to set up the experimental field, one component (nitrate or nitrite) was first eliminated from the feed compartment using different parameters, such as the concentration of counter-ions and the concentration of nitrate and nitrite separately. In order to improve the procedure and comprehend the simultaneous transport of nitrites and nitrites, the removal of two components (nitrate and nitrite) in the feed compartment was then explored by the Response Surface Methodology (RSM) by Doehlert design.

2. Experimental

2.1 Anionic-exchange membranes

During the Donnan dialysis operation, three commercial anion-exchange membranes have been used: ACS, AMX, and AFN. These membranes have the same structural properties and are homogeneous. Table 1 presents the characteristics and properties of these membranes.

Membrane	Ion-exchange capacity (mmol/g)	Water content %	Thickness (mm)
AFN	3.00	47.8	0.12
AMX	1.30	26.0	0.13
ACS	1.85	18.9	0.15

Table 1.

The main characteristics of the three commercial anion-exchange membranes.

In order to prepare the samples for use in the Donnan dialysis, they had to be conditioned before any measurement. This was done primarily to eliminate contaminants from the production process and to stabilize their physical-chemical properties. French standard NF X 45-200 [21] was followed in doing this conditioning.

2.2 Donnan dialysis

All Donnan dialysis experiments were carried out using a laboratory cell. The device was used to study the nitrate and nitrite removal by Donnan dialysis. It is composed of a thermoregulated water bath (25.0 ± 0.1°C in this study), containing a cell with feed and receiver compartments separated by an anion-exchange membrane.

The dialysis cell consists of two detachable compartments made with polymethylmetacrylate (plexiglass) as shown in Figure 1a and b, in two formats, a photo of the mounted cell ready to be connected to the peristaltic pump, and a drawing of the different sections of this cell. It is composed of four parts joined by three stainless steel threaded rods. The centering is assured by bolsters. The two central compartments, consisting of two tubes are symmetrical. Two threaded holes penetrate each compartment and serve as support for stuffing boxes. The membrane is sandwiched between these two compartments, making a seal at the same time [22].

Figure 1.

The two-compartment cell used for the Donnan dialysis experiments. (a) Photo of an assembled cell. (b) Plan with different cuts according to three sections.

Then, the solution of nitrate or/and nitrite was prepared as a feed compartment with different concentrations varying from 50 to 500 mg/L, and the solution of chloride with different concentrations varying from 10 to 100 mg/L. The solutions are placed in volumetric flasks which are essential due to their geometry because they limit the losses of solvent by evaporation. There the circulation of these solutions in the two compartments is ensured by a peristaltic pump equipped with two identified heads controlled by a speed variator acting on his engine. The circulation of fluids takes place in flexible reference pipes. The volume flow rates Q_f and Q_R of solutions leaving the feed and receiver compartments respectively are determined using a 1000 mL flask and a stopwatch measuring their filling times. This method assumes that the transmembrane volume flux is infinitely small compared to the solution flow rate imposed in each compartment.

The residual concentration of nitrate or/and nitrite at the outlet of the receiver compartment was determined spectrophotometrically. The UV-spectrophotometry approach was used to measure the nitrite and nitrate content at the receiver compartment during the dialysis operations [23]. The sodium salicylate and nitrate reaction produce paranitrosalicylate sodium, which is yellow colored. This reaction is followed by absorbance measurements at 415 nm using a UV-visible spectrophotometer to determine the amount of nitrate present. The amino-4-benzenesulfonamide was diazotized by nitrites in an acidic medium, and when it was coupled with N-(naphthyl-1) diamino-1,2-ethane dichloride, a purple-colored complex resulted. This complex’s absorbance at 543 nm was then measured using a UV-visible spectrophotometer. The removal rate of nitrate (Y₁%) and nitrite (Y₂%) was calculated by Eq. (1):

$${\mathrm{Y}}_{1\text{or}2}\left(\%\right)=\frac{{\mathrm{C}}_0-{\mathrm{C}}_{\mathrm{e}}}{{\mathrm{C}}_0}\times 100$$E1

where C_e is the nitrate and nitrite equilibrium concentration (mg/L) and C₀ is the initial concentration of nitrate or nitrite (mg/L).

2.3 Doehlert design

A Doehlert design based on the RSM was employed as the experimental strategy in this inquiry. The values of these parameters must be simultaneously optimized in order to achieve the best system performance. The optimal scenario was found by superimposing the contours of the response surfaces in a plot with multiple responses. In the three-dimensional plots of numerous variables used to represent the graphical optimization in the experimental field, the regions of optimal response would be highlighted in red. A close match between the experimental and projected values is required [24]. The total number of experiments for k factors is N = k² + k + 1. In fifteen tests, three duplicates at center field were employed [24, 25, 26, 27].

The initial nitrate, nitrite, and counter-ion concentrations in the receiver compartment were the factors that were examined. With one component in the feed compartment, the range of these factors was set in accordance with the preliminary investigation. The experimental field of the factors under investigation is shown in Table 2.

Factors	Symbol	Range and levels
Coded variable X1	[${\mathrm{NO}}_2^{-}$]	−1	0	1
Initial concentration ${\mathrm{NO}}_2^{-}$ (mg/L)		10	55	100
Coded variable X2	[${\mathrm{NO}}_3^{-}$]	−1	0	1
Initial concentration ${\mathrm{NO}}_3^{-}$ (mg/L)		50	275	500
Coded Variable X3	[${\mathrm{Cl}}^{-}$]	−1	0	1
Concentration of ${\mathrm{Cl}}^{-}$ (mol/L)		0.1	0.3	0.5

Table 2.

Range and levels of nitrate and nitrite removal.

The variables that were examined included the initial nitrate, nitrite, and counter-ion concentrations in the receiver compartment. The range of these factors was established with one component in the feed compartment in accordance with the preliminary investigation. The experimental setting for the factors under investigation is shown in Table 2.

The Doehlert design, a matrix that can anticipate the values of the response at any point in the experimental area, can be used to estimate the coefficients of a second-order function [28]. Using a polynomial equation (Eq. (2)), the selected model represents the predicted values of the answers Y. Bi represents the estimated major effect of component i, b_ii represents the estimated second-order effects, b_ij represents the estimated interactions between the factors i and j, and X_i represents the coded variable. NemrodW® Software was used to determine the model’s coefficients.

$$\begin{array}{l}{\mathrm{Y}}_{1\text{or}2}={\mathrm{b}}_0+{\mathrm{b}}_1{\mathrm{X}}_1+{\mathrm{b}}_2{\mathrm{X}}_2+{\mathrm{b}}_3{\mathrm{X}}_3+{\mathrm{b}}_{11}{\mathrm{X}}_1^2+{\mathrm{b}}_{22}{\mathrm{X}}_2^2+{\mathrm{b}}_{33}{\mathrm{X}}_3^2+{\mathrm{b}}_{12}{\mathrm{X}}_1{\mathrm{X}}_2\\ +{\mathrm{b}}_{13}{\mathrm{X}}_1{\mathrm{X}}_3+{\mathrm{b}}_{23}{\mathrm{X}}_2{\mathrm{X}}_3\end{array}$$E2

The percentage absolute error of deviation (AED) and the regression coefficient (R²) between experimental and theoretical findings were employed as two metrics to assess the models. Eq. (3) was used to compute the AED.

$$\mathrm{AED}\left(\%\right)=\frac{100}{\mathrm{N}}.\left|\frac{{\mathrm{Y}}_{\exp }-{\mathrm{Y}}_{\text{theo}}}{{\mathrm{Y}}_{\exp }}\right|$$E3

where Y_theo represents the theoretical replies and Y_exp represents the experimental responses. N is the total number of locations where measurements were made. The validation of the model is deemed valid when R² > 0.7 and AED 10% [29].

3. One-component Donnan dialysis in the feed compartment

3.1 Counter-ion concentration effect

The effect of counter-ion concentrations in the compartment receiver on the removal of nitrate and nitrite from the feed compartment separately was an important parameter of this investigation. One of the crucial factors influencing the elimination of nitrates and nitrites across the membrane is the counter-ion. Due to its high mobility, chloride appears to be the counter-ion that is employed the most.

We investigated the effect of increasing the concentration of counter-ion ${\mathrm{Cl}}^{-}$ in the receiver compartment from 0.01 to 0.5 mol./L on the removal of nitrates (100 mg/L) and nitrites (20 mg/L) separately from the feed compartment to the receiver compartment. We have tested three membranes (AFN, AMX, and ACS), with four-hours DD operations. Figures 2 and 3 show the variation in nitrate and nitrite concentrations at the receiver solution’s outlet.

Figure 2.

The counter-ion concentration effect on nitrate removal for the three tested membranes.

Figure 3.

The counter-ion concentration effect on nitrite removal for the three tested membranes.

For the three membranes, it was observed that increasing the concentration of ${\mathrm{Cl}}^{-}$ from 0.01 to 0.5 mol/L resulted in a greater removal percentage for nitrate and nitrite. In the receiver compartment, the concentration of nitrate and nitrite was very low when ${\mathrm{Cl}}^{-}$ was 0.01–0.05 mol/L. As the ${\mathrm{Cl}}^{-}$concentration increased from 0.1 to 0.5 mol/L, nitrates and nitrites were greatly removed. To maintain electroneutrality, the cross-ion transfer between ${\mathrm{Cl}}^{-}$ and nitrates and nitrites improves because the concentration gradient of the counter-ions increases. According to Donnan dialysis with three membranes, an increase in the counter-ion concentration in the receiver compartment is associated with a significant improvement in the removal of nitrates and nitrites in the feed compartment; this is reflected by an increase in the exchange kinetics. Ben Hamouda et al. [13] and Turki et al. [30] have also reported similar conclusions.

3.2 Nitrate and nitrite concentration effect in the feed compartment

Depending on the geographic location, natural waters contained varying levels of nitrate and nitrite. Due to this, nitrate and nitrite concentrations were studied separately in the feed compartment. For this study, the initial concentration of nitrate was varied from 10 to 500 mg/L, and the initial concentration of nitrite was varied from 5 to 100 mg/L. The concentration of counter-ion ${\mathrm{Cl}}^{-}$ was 0.1 mol/L in the receiver compartment. Figures 4 and 5 show the variation of nitrate and nitrite concentrations at the receiver solution’s outlet.

Figure 4.

Nitrate concentration effect in the feed compartment on nitrate removal for the three tested membranes.

Figure 5.

Nitrite concentration effect in the feed compartment on nitrite removal for the three tested membranes.

Figure 4 shows the variation of nitrate concentration from 50 to 500 mg/L, the rate of removal significantly increased for all membranes. Nitrate removal improved when concentration going from 50 to 500 mg/L by 38 to 61% with AFN, 30 to 50% with AMX, and 34 to 50% with ACS. The highest nitrate exchange for chloride ion efficiencies was achieved with the AFN membrane when performing Donnan dialysis.

Figure 5 shows the variation of nitrite concentration from 10 to 100 mg/L, the rate of removal significantly increased for all membranes. In the feed compartment when the nitrite concentration was the lowest (10 mg/L), the removal for AFN, AMX, and ACS was 17%, 15%, and 16% respectively. The removal was enhanced with an increase in nitrite concentration from 10 to 100 mg/L from 17 to 70% with AFN, 15 to 65% with AMX, and 16 to 66% with ACS. It appears that the AFN membrane removed nitrites most effectively.

We can conclude that the improvement in the removal of nitrate and nitrite was mostly caused by the rise in the initial concentration of nitrate and nitrite, regardless of all membranes AFN, AMX, and ACS. The increase in the concentration gradient of nitrate and nitrite, which increased the chloride ion flux from the receiver compartment to the feed compartment, can be attributed to this [11]. As a result, the cross-ion transfers between ${\mathrm{Cl}}^{-}$ and nitrates and nitrites are improved.

4. Two components of Donnan dialysis in the feed compartment

4.1 Choice of the membrane

Three membranes, AFN, AMX, and ACS, were examined for the selection of anion-exchange membranes due to the intricacy of the correlation between their qualities given in catalogs and the real performances in Donnan dialysis process. With a counter-ion concentration of 0.5 mol/L, an initial nitrate concentration of 100 mg/L, and a nitrite concentration of 20 mg/L, Donnan dialysis was carried out. Figure 6 illustrates the simultaneous testing of membranes (AFN, AMX, and ACS) for the removal of both nitrites and nitrates in the same compartment.

Figure 6.

Choice of the best membrane allowing the highest removal rate for both nitrate and nitrite.

Figure 6 shows the simultaneous elimination of nitrites and nitrates by three membranes (AFN, AMX, and ACS) in the same compartment. Because a higher proportion of nitrates and nitrites ions in the feed compartment increases the overall flow, which in turn causes a higher proportion of the counter-ions to be transported from the feed to the receiver, the presence of nitrite and nitrite in the same compartment improves their elimination.

In comparison to the AMX and ACS membranes, the AFN membrane has the superior performance. The AFN membrane is a macro-porous structure with a high concentration of inorganic groups and a low amount of a cross-linked agent. This membrane has a relatively high ion-exchange capacity and the highest water content [14].

4.2 Doehlert design

The simultaneous removal of nitrate and nitrate was optimized using the membrane AFN and RSM via Doehlert design. The 15 tests of the Doehlert experimental design (Table 3) contained three replicates at the center field. Repeated measurements on the same center field can yield almost the same results for our solution, resulting in a significant lack of fit. Replication actually reduces the experimental data’s variability, increasing its significance and level of confidence [31].

N°	[${\mathrm{NO}}_2^{-}$]	[${\mathrm{NO}}_3^{-}$]	[${\mathrm{Cl}}^{-}$]	Y1 (%)	Y2 (%)
1	100	275	0.300	86.3	98.2
2	10	275	0.300	88.1	79.4
3	78	470	0.300	92.7	97.1
4	33	80	0.300	79.1	87.9
5	78	340	0.300	76.3	97.2
6	33	210	0.300	93.2	87.4
7	78	210	0.463	92.9	99.8
8	33	405	0.137	69.9	82.3
9	78	340	0.137	65.7	89.7
10	55	145	0.137	73.7	86.7
11	33	340	0.463	92.7	90.9
12	55	145	0.463	84.2	97.1
13	55	275	0.300	87.2	96.3
14	55	275	0.300	87.2	96.3
15	55	275	0.300	87.2	96.3

Table 3.

Experimental design and results of nitrate and nitrite removal.

Using the experimental findings from Table 3 and the second-order polynomial equations provided by Eqs. (4) and (5), the necessary data were fitted to the equation.

$$\begin{array}{l}{Y}_1=87.2-1.36X_1+8.70X_2+12.35X_3-2.50X_2^2-10.40X_3^2+1.33X_1{X}_2\\ +1.33X_1{X}_3+1.03X_2{X}_3\end{array}$$E4

$$\begin{array}{l}{Y}_2=96.3+9.11X_1-0.28X_2+5.94X_3-7.50X_1^2-2.50X_2^2-5.27X_3^2\\ +0.23X_1{X}_2+0.84X_1{X}_3-0.04X_2{X}_3\end{array}$$E5

According to obtained results, the coefficients are presented and show that the counter-ion concentration had an important effect (b₃ = 12.35) on the removal of nitrate. The second influenced factor was the nitrate concentration (b₂ = 8.70). Except the concentration of nitrite had a less important effect on the removal of nitrate (b₁ = 1.36). According to obtained results, the coefficients are presented and show that the nitrite concentration had an important effect (b₁ = 9.11) on the removal of nitrite. The second influenced factor was the counter-ion concentration (b₃ = 5.94). Nevertheless, the concentration of nitrate had a less important effect on the removal of nitrate (b₂ = 0.28).

The multiple polynomial model coefficients (Eqs. (4) and (5)), which reflected the effects and interactions of the many components under investigation were identified. A check of the relative importance of various coefficients in the experimental region under investigation is possible thanks to the Pareto analysis (Figure 7). The following relation (Eq. (6)) allows to calculate this analysis:

Figure 7.

Pareto analysis for nitrate (a) and nitrite (b) removal.

$${\mathrm{P}}_{\mathrm{i}}={\left(\frac{{\mathrm{b}}_{\mathrm{i}}^2}{\sum {\mathrm{b}}_{\mathrm{i}}^2}\right)}^2\times 100$$E6

The concentration of chloride and nitrate both have a positive effect on the studied response, meaning that increasing their concentrations leads to improved nitrate removal. Their contributions to the studied response were only 30.5% for chloride concentration and 15.2% for nitrate concentration. Thus, the removal can be considerably influenced by two parameters: chloride concentration and nitrate concentration. However, the other interactions have a negligible effect; they represent only 1.0% of the studied response. The concentration of nitrites and the concentration of chloride are two factors that affect nitrate elimination. The concentration of nitrite has a positive effect, which translates to improved elimination via Donnan Dialysis as nitrite concentration rises. This improvement was attributed to the rise in the concentration gradient of nitrates, which enhanced the chloride ion flux from the receiver to the feed solution, hence the cross-ion transfer between Cl⁻ and nitrate improves to maintain the electroneutrality.

Two factors positively impact the studied response, indicating that an increase in these factors leads to an improvement in nitrite removal. Their contributions to the studied response were only 33.24% for nitrite concentration and 14.13% for chloride concentration. The nitrate concentration has a negative effect, meaning that an increase in nitrate concentration leads to a decrease in nitrite removal. However, the interaction between nitrite and chloride has a negligible effect of 0.28%. Similarly, the other interactions also have a negligible effect, accounting for only 0.3% of the studied response.

The concentration of nitrites and the concentration of chloride are two factors that affect nitrite elimination. The concentration of nitrite has a positive effect, which translates to improved elimination via Donnan Dialysis as nitrite concentration rises.

This improvement was attributed to the rise in the concentration gradient of nitrates and nitrites, which enhanced the chloride ion flux from the receiver to the feed solution, hence the cross-ion transfer between ${\mathrm{Cl}}^{-}$ and nitrite improves to maintain the electroneutrality.

Using the regression coefficient (R²) and the percentage of absolute errors of deviation (AED), the validity of the model was evaluated. According to Y₁, the regression coefficient is 0.999, while according to Y₂, it is 0.997. However, the percentage absolute error of deviation, AED (%) = 0.262%, was less than 10% for nitrate removal, and AED (%) = 0.311% for nitrite removal. These confirm the validation of the models suggesting that the model is suitable to describe the removal of nitrate and nitrite.

Analysis of variance (ANOVA) was carried out to support the models’ suitability. The results are shown in Table 4 in this regard. The F-ratio is the mean square error, and the p-value is the ratio of the mean square effect.

Source model	Sum of square	Degree of freedom	Mean of square	F-value	P-value
${\mathrm{NO}}_3^{-}$
Regression	1149.0	9	1272.10	86.8865	0.0005
Residual	7.3205	5	1.464
Total	1152.2	14
${\mathrm{NO}}_2^{-}$
Regression	1806.89	9	61.93	223.9885	0.0001
Residual	1.3825	5	0.2765
Total	558.77	14

Table 4.

Variance analysis of nitrate and nitrite removal.

The P-value has been used to identify the effects that are statistically significant. The P-value is crucial since it is close to zero, which denotes the significance of the data. The Fischer value (F_{0.05, 1, 1.16}) for 5% error, 1 degree of freedom, and 16 factorial testing is 4.77, according to the Fischer table. Due to the fact that every effect’s value is higher than 4.77, it appears that they are all important. At a level of 5%, the experimental model’s Fischer value is significantly greater than the crucial F value. The model is therefore regarded as statistically significant.

The software NemrodW®s desirability function produced the optimal conditions. The optimum values for each factor are therefore 82 mg/L for nitrite concentration, 406 mg/L for nitrate concentration, and 0.412 mol/L for counter-ion concentration. The maximal elimination of nitrates (95.5%) and nitrites (100%) was achieved under these conditions. In order to confirm the accuracy of the predictions, the experiment was replicated three times under optimal conditions. Since the coefficient of repeatability was less than 1%, it can be said that Donnan dialysis’s simultaneous removal of nitrates and nitrites is repeatable. The experimental results’ statistical analysis revealed that the analysis has a normal distribution.

5. Application on real water

Simultaneous removal of nitrate and nitrite from real water was performed to ensure the feasibility and effectiveness of Donnan dialysis. We used water from the eastern side of the Tunisian Cape Bon, a region well-known for its intensive cultivation of vegetables and fruit, mainly citrus. The large quantities of chemical fertilizers have caused real problems with the quality of both surface and groundwater. Table 5 shows the anion content of water taken from a well in the town of Menzel Bouzelfa. We notice the high contents of nitrates and nitrites which are respectively of the order of 100 and 10 mg/L.

Parameters	Concentration (mg/L) or value
Fluoride	0.8
Chloride	528.9
Nitrate	98.4
Nitrite	8.9
Phosphate	5.2
Sulfate	127.5
pH	7.8

Table 5.

Composition of the groundwater from Menzel Bouzelfa.

We used the experimental parameters resulting from the optimization of this process for similar nitrate and nitrite contents, namely a chloride concentration of 0.4 mol/L and a removal time of 4 hours under agitation. We also tested the three membranes AFN, AMX, and ACS. The results obtained are given in Figure 8. It can be seen that the most efficient membrane is AFN and that the removal rates for this membrane are around 98% for nitrates and 85% for nitrites. These results are perfectly consistent with theoretical predictions and confirm that Donnan dialysis remains very efficient for the removal of these two anions.

Figure 8.

Simultaneous removal of nitrate and nitrite from the real water of Menzel Bouzelfa—Tunisia.

6. Conclusion

In order to clearly depict the impact of every factor on optimum conditions and to determine the optimal parameters for simultaneous removal of nitrate and nitrite by Donnan dialysis with a small number of experiments, RSM according to the Doehlert matrix was used. The RSM is highly efficient and makes it easier to achieve the optimum conditions while accounting for interactions between experimental parameters. These conditions were 82 mg/L for the concentration of nitrite, 406 mg/L for the concentration of nitrate, and 0.412 mol/L for the concentration of chloride, which were the best for the simultaneous removal of nitrates (95.5%) and nitrites (100%) through the AFN membrane. The removal effectiveness of the simultaneous removal of nitrite and nitrate by DD as well as the optimization of process factors for maximal removal were successfully determined by the RSM approach. Finally, the application of the DD for treating real water (from the region of Menzel bouzelfa) to remove nitrate and nitrite allowed us to confirm that AFN is the best membrane and that the elimination rates remain very high and close to those found in the theoretical study.

Acknowledgments

The researchers would like to thank the ICMPE/University Paris-Est Créteil for funding the publication of this project.