Reuse of electrocoagulated metal hydroxide sludge to 2 fluoride and arsenic removal by a fixed-bed column in 3 continuous operation 4

In the present study, Electrocoagulated Metal Hydroxide Sludge (EMHS) was analyzed 21 as adsorbent material to remove both fluoride ion (F-) and arsenic V (As(V)) from aqueous effluents. 22 This material was generated during an electrocoagulation process using Aluminum anode. It was 23 characterized by using specific surface areas and the surface morphology was studied by scanning 24 electron microscopy (SEM). Adsorbent fixed-beds are generally studied to remove different class of 25 contaminants. EMHS was evaluated using a continuous flow rate column test with an experimental 26 design. The effect of initial concentration of F(2.5-10 mg L-1) and the Empty Bed Contact Time 27 (EBCT (0.4-0.8 min)) was studied following a central composite design methodology. The 28 experimented parameters had a significant influence on saturation time, breakthrough volume, and 29 breakthrough time. A response surface analysis was a tool for analyzing the adsorption study, 30 showing interactions that are complicated to identify by others methods. The results, here reported, 31 revealed that EMHS is an efficient and promising adsorbent material in order to remove Fand 32 As(V) from water contaminated by these pollutants. 33


Introduction
The expanded distribution of diverse ionic elements such as fluoride and arsenic in water as result of diverse industry process or naturally occurring has generated interest from the scientific researches since they are affecting human health in the world.In this context, the most substantial inorganic pollutants in groundwater, stablished by the World Health Organization (WHO), are fluoride (F -) and arsenic (As) [1].F -at small amount is favorable for bone and teeth development, but in concentrations elevated are harmful to human health, provoking skeletal or dental fluorosis [2].On the other hand, As is distinguished for generate cancer in skin, lung, kidney, liver and bladder, besides gastrointestinal problems and arsenicosis [3].Therefore, the WHO recommends guideline values for F -and As concentrations with the upper permissible limit in water of 1.5 mg L -1 and 10 µg L -1 , respectively [4].
In many places worldwide, F -and As concentrations show a significant co-contamination in groundwaters whit concentrations up of 29 mg L -1 and 5300 µg L -1 for F -and As, respectively [5].In the actuality water contaminated with F -and As is a concern, which requires an efficient treatment.
Adsorption technique has shown considerable potential due to its simplicity, chemicals addition are not necessary, and its efficiency with high grade of total solids [21].Different strategies has been used employing diverse materials for individual and concurrent elimination of F -and As.Several adsorbents have been used to eliminate F -and As in both, individual and a concurrent way such as activated carbon [22], layered double hydroxides [23], aluminum hydroxide [24], ferric hydroxide [25], goethite [26], Fe-Ce oxides [20], Fe-Al doped polymers [27], Haix-Fe-Zr and Haix-Zr resin beads [28], inorganic ion exchange adsorbents [29], mesoporous aluminas [30], modified cellulose [31] and volcanic ash [32] Researches are looking simple and cost effective processes, diverse low cost materials from different sources have been applied for F -and/or As removal [33].
A candidate strategy for providing low-cost adsorbent can be Electrocoagulated Metal Hydroxide Sludge (EMHS) which is obtained from an electrocoagulation (EC) system, in this process sacrificial aluminum or iron electrodes are oxidized releasing hydroxides of these metals.Metallic hydroxides can separate some pollutants by electrostatic attraction or surface complexation [34].The disposal of these materials, which is considered as waste, are actually a critical subject because they can cause environmental and public health impacts without an appropriate management [35].The EMHS can be utilized to remove F -and As from the effluents which formed the motivation of this study.
In this investigation, EMHS was reused before disposal for a modelling and experimental research.No results of fluoride and arsenic removal in fixed bed coulum by EMHS have been presented so far.Therefore, the overall motivation for investigating the effect of this molecules on sorption is toward developing a mechanistic understanding of inorganic sorbate-EMHS interactions.
The characteristics of EMHS using specific surface areas and scanning electron microscopy techniques were analyzed.The influence of column variables such as empty bed contact time (EBCT) and fluoride ion (F -) concentration in a continuous flow fixed-bed column have been investigated using a response surface methodology.

Chemicals and analytical method
All of the reagents used in this work were analytical grade and were used without any further purification.Fluoride (F -) and arsenic (As(V)) stock solutions were made from sodium fluoride (NaF) and sodium arsenate (NaAsO3•12H2O), respectively.Hydrochloric acid (HCl) and sodium hydroxide (NaOH) were used to adjust requiered pH.Fluoride concentration was monitored using an ionselective electrode for fluoride ion (Thermo Scientific 9609 BNWP), while for As concentration a digital arsenic Test Kit (Arsenator®) was used.

Adsorbent material preparation
EMHS applied to adsorption research and carried out in this investigation was by-produced from electrocoagulation (EC) pilot plant using aluminum anodes, at the say way as is reported in a previous work [36].A flow diagram of the adsorbent by-production used in this study is shown on

Material characterization
The characterization of EMHS was analyzed by N2 adsorption-desorption isotherms at 76 K utilizing a Micromeritics ASAP 2020 surface area and porosity analyzer.N2 isotherms with the Brunauer-Emmett-Teller method (BET) was used to determine the surface area BET.The volume of liquid nitrogen corresponding to the amount adsorbed at relative pressure of P/P0 = 0.99 and was defined as the total pore volume, VT.The Dubinin-Radushkevich method was used to determine the micropore volume, Vµ, and the mesopore volume, Vm, was obtained with the difference between VT and Vµ.The average pore diameter, Dp, was calculated using the relation 4 VT /area BET and the pore size distribution by Density-Functional-Theory method (DFT).The scanning electron microscopy (SEM) assay was carried out for the adsorbent material in order to analyze their morphology.

Column studies
Adsorption processes design in a full scale involves a lot of time and expensive pilot plant evaluations.These experiments could take several years.For prevent these expensive studies quick laboratory experiments are employed.The small-scale fixed bed column test is, possibly the most useful tool to evaluate adsorbents capacity to remove pollutants.In a series of column test were studied the influence of the initial F -concentration (mg L -1 ) and empty bed contact time (EBCT, (min)).
According to several investigations [37], a prediction of the variation of adsorption contaminants rate on adsorbents as adsorbate concentration is difficult.If intraparticle diffusion is the main mechanism, an equilibrium interfacial concentration should form rapidly that is followed by slow diffusion into the adsorbent particles.Then, a simple dependence on solution concentration is not expected.
Moreover, the mathematical treatment of intraparticle diffusion does not lead a simple algebraic relationship between external solute concentration and time of reaction even when a constant saturated external layer is maintained.In this case, concentration effect is used to define the ratelimiting reaction step.When intraparticle transport limits the kinetic of an adsorption reaction, the variation in reaction rate is not expected to be linear, whereas the rates of strictly adsorptive reactions and simple diffusion-controlled processes are expected to be proportional to the of adsorbate´s concetration.
Arsenic concentration was fixed at 100 µg L -1 .The EBCT is represented by the flow rate through the column, this parameter influences the shape of breakthrough curve and the volume to breakthrough.The EBCT is resolved using equation ( 1): An acrylic column of 2 cm internal diameter and 100 cm length was filled with metal hydroxide sludge as adsorbent.Before to start column tests, a flow rate of 26 ml min -1 of deionized water was introduced in the column for 5 min with the propose of remove air bubbles and to guarantee a closely packed disposal of particles without cracks, channels or voids.After that, the pollutant solution was passed through a fixed-bed of adsorbent in down-flow mode.A flow rate of 26 ml min -1 was controlled with a peristaltic pump.Periodically, samples were collected from the effluent and then analyzed for the remaining pollutant concentration.Temperature was maintained around 21±0.8ºC, while pH of the medium was fixed at 7 in all experiments.

Experimental design
The one variable at a time or step-by-step variable strategy is the traditional way to analyzed the effect of several independent variables or factors over dependent variables or response variables, in which the variables remain constants and only one is changing.However, this strategy implicates several number of runs and interactions among factors are missed.These drawbacks can be solved by response surface methodology (RSM).This methodology is a mathematical and statistical method functional for studying and optimizing processes [37].Therefore, for explore the impact of operating conditions on diverse process the RSM can be an alternative.
Figure 3 shows the coded presentation of the CCD for 2 factors.
Factorial designs with two levels (2 K ) have advantage in the total of experimental runs compared with the 'one variable at a time' method.Here a mathematical model can be used to define the behavior of the response variable on the interest zone.In this investigation a RSM was developed based on the CCD methodology experiments.
The statistical software of Design-Expert (version 7.0, STAT-EASE Inc., Minneapolis, MN, USA) was applied for designing and analyzing the experimental results.The values of process variables and their variation limits were selected based on the preliminary experiments.For statistical purposes, the transformation of independent variables (  ) into coded corresponding variables (  ) was generated by (Eq.( 3)): Where   is the value of   at the center point and  represents the step change.The table 1 shows the codified levels.ANOVA was applied to RSM model to study the individual and combined effect of two variables X1 and X2.The response variables studied were related with F -pollutant like saturation time   ), breakthrough volume   ), and breakthrough time   ).Also, arsenic removal efficiency (% of removal) is reported.The sequential model fitting test was implemented in order to select an appropriate model.For express the response variable in the investigated domain it is appropriate an adjustment for a mathematical model equation.Generally, to describing a flat surface the first-order model can be used under to the following expression (Eq.( 4)): Where  is the response variable,   is the constant coefficient,   represents the coefficients of the linear parameters,   represents the variables studied.When interaction terms are incorporated (FI model or factor interactions), the next equation (Eq.( 5)) can represent the first-order model: Where   represents the coefficients of the interaction parameters   and   and  < .
where   expresses the coefficients of the quadratic parameter and  < .
An experiment can be optimized for examining the analysis of variance (ANOVA) statistics (R 2 , the adjusted R 2 , lack-of-fit, F-test and t-test), the residual analysis, normal plots, interaction effects and the contour plot, and this way decide the fit of the first-order or second-order model.The runs carried out are shown in Table 2.

Optimization: Desirability function.
The desirability is a multiple response method, which is useful for the process optimization.This method utilizes an objective function or desirability function (D), which change a calculated response into a value from zero to one, in other words, form least to most desirable.The optimal parameter conditions are considered with maximum desirability.The desirability function is maximized by the numerical optimization.Regulating the importance may change the attribute of a goal.The desirability function combines all goals of several responses.In the optimization, both the independent and response variables have a low and high value for every goal.

SEM analysis
Micrograph obtained from EMHS is shown in Fig. 2, where the morphology of the sample presents particle sizes between 5 and 350 micrometers.In general, the adsorbent studied showed a brightness in the regions analyzed, which indicates chemical consistency related to C, Si, O, and Al.As can be seen in Fig. 3a, according the average pore diameter of EMHS comprised mesopores and macropore, given that diameter is within the range of 2-50 nm and partially more than 50 nm.The result can be confirmed by the nitrogen adsorption/desorption isotherms in Fig. 3b.It can be seen that EMHS exhibited an IV type N2 adsorption isotherm with an evident hysteresis loop (according to the IUPAC classification), implying the existence of mesopores structures in the material [39].Furthermore, the existence of macropores is evidenced given that the hysteresis loop shifts approach relative pressure (p/p0)=1.
The adsorbent showed a total pore volume (TPV) of 0.76 cm 3 g -1 and a BET surface area of 350.19 m 2 g -1 .From the TPV amount, only 0.007 cm 3 g -1 is due to the micropore volume, and 0.753 cm 3 g -1 correspond to mesopore volume.In other words, the EMHS is composed mainly of mesopores and macropores (99%).According to the DFT and BET analysis, the average pore width was 8.43 nm.A large adsorption capacity of the EMHS is confirmed due to high porosity, small pore structures and large specific surface area, indicating its potential as adsorbent.

Column studies
The most significant conditions, which impact in the performance of pollutant removal in a column adsorption method, are initial pollutant concentration and EBCT.Experiments were carry out for varied interactions of the operational parameters using a design of experiments and this way explore the effects of factors mentioned above.For obtain the regression equations were fitted to the experimental data a linear, interactive, quadratic and cubic models.Two different tests were carried out to represent by models the F -performance in function of the response variables (tb, Vb, ts) by EMHS, from the sequential model sum of squares and model summary statistics, cubic model was aliased.The linear model showed a p-value was less of 0.01 on all response variables (data not shown), according to sequential model sum of squares.Model summary statistics showed that the excluding cubic model which was aliased, lineal model was found to have maximum ''adjusted R-Squared'' and the ''Predicted R-squared'' values on all response variables.For that reason, lineal model was selected for statistical analysis.
ANOVA method was utilized to verify the fitness as well as the significance of the models.In our ANOVA results (data not shown), the model F-values of 24.57, 38.62 and 8.57 implied that the model is significant for   ,   and ts, respectively also there was only a 0.04%, 0.01% and 1% chance that a model F-value could occur due to noise.A Prob F under 0.05 suggested that model terms are significant.A lack of fit F -value of 2.51, 5.05 and 17.90 implied the lack of fit is not significant relative to the pure error and there is a 31.24%,17.42% and 5.48% chance that a lack of fit F-value this large could occur due to noise for   ,   and ts, respectively.Given that we want the model to fit, nonsignificant lack of fit is necessary.
Predicted R 2 is an indicator of how the model forecast a response variable.Both, predicted R 2 and adjusted R 2 should be near (no more of 0.20) to be a permissible adjustment, if not the experimentation has a problem with the data.In our case, for all response variables the predicted R 2 is near with the adjusted R 2 .A signal to noise ratio or in other words a range in estimated response in relation to its error is called adequate precision, where must be at least 4 its desired value.A value of 16.647 indicated an acceptable signal in our experiments.In this model, the coefficient of variation can represent the error, declared as a percentage of the mean.In addition, for verify the significance over the coefficients the P-values were utilized, which also indicate the way of the interactions between the variables.With small values of P, the coefficient is more significant.From our results the coefficients A and B (for F -concentration and EBCT, respectively) were significant, with small Pvalues (P <0.05) (data not shown).
The results were also examined to verify the normality of the residuals.The normality of the data can be identifying with a normal probability plot, this is a strategy for calculate if a data group is normally distributed [40].The difference between the predicted and the real value is called residual.
If the data are normally distributed, the values should be located on the plot close to the straight line.
In our results the data points were reasonably aligned with a straight line (data not shown), suggesting normal distribution for the three variables of response.Table 2 shows the independent and dependent variables used and the results obtained in this study.Breakthrough time was the primary parameter estimated (  ), which is the time required for 50% of adsorbent saturation.The application of the model in terms of coded factors to the results generated the following equation (Eq.( 7)): = . − .  + . (7) The   value shows high correlation between the model (Eq 7) and the experimental data.In other words, the experimental data fit the model very well.Figure 6 plots the corresponding response surface.Both F -concentration and EBCT have a significant effect on   , which increases with high values of EBCT and when F -concentration decrease.
For estimate the volume of influent treated by gram of adsorbent (  ), the equation 8 can be used: Where   is the volume of treated influent and   is the mass of dry adsorbent used.The resulting quadratic model is (Eq.( 9)): Figure 7 shows that   is in function of F -concentration and EBCT.  increase at the maximum in the region between 2.5 and 4 mg L -1 of F -and between 0.7 and 0.8 min for EBCT.
The time required to saturate (saturation time,   ) the adsorbent is one of the most important variable used to describe the performance of an adsorption process.This time is when outlet concentration is equal to   .The selected breakthrough concentration was fixed at 10% of the inlet feed concentration.The quadratic model found for   is (Eq.( 10)): In contrast to the other response variables, the effect of both independent variables, F - concentration and EBCT on saturation time (ts) is not independent of each other, but there is a synergistic effect among independent variables.The zone where F -concentration and EBCT variables are highest represent the highest ts values (Fig. 8).According to the results obtained in this investigation, the Surface Response Methodology can be applied in to study adsorption investigations, because of interactions or synergistic effects among independent variables not can be examined by the traditional step-by step methods.

Conclusions
Material used at the present study (EMHS) is an efficient adsorbent, which can be applied to eliminate F -and As from aqueous effluents.In column studies both fluoride ion concentration and EBCT have a significant effect on the three response variables investigated (tb, Vb, and ts); experimental data fits the lineal model very well.The effects of F -concentration and EBCT on ts is synergistic, which not can be shown by others ways, like the classical step-by-step way.The results presented demonstrate that use of surface response methodology is a potential strategy for analyzing adsorption process in adsorbent-fixed beds.A better comprehension of adsorption process will help the selection of the optimal conditions of studied variables for a given application.

Figure 1 .
Figure 1.Process flow diagram of the EMHS production.

Figure 3 .
Figure 3. Nitrogen adsorption-desorption isotherm (a) and the BJH pore-size distribution curve (b) of the obtained EMHS.

Fig. 4
represents the predicted and actual values of the response variables for the removal of fluoride onto EMH.The developed model is adequate for all response variables in view the prediction residuals from the responses are distributed in a diagonal line.

Figure. 4 .
Figure. 4. Predicted response versus actual response for response variables: a)  , b)  and c) t s .

Figure 6 .
Figure 6.Variation of t with F -concentration and EBCT.

Figure 7 .
Figure 7. Variation of  with F -concentration and EBCT.

Figure 8 .
Figure 8. Variation of  with F -concentration and EBCT.

Table 1 .
Experimental range and level of independent variables.
Sometimes for an interpretation of relationships with independent variables the FI or first-order models are not convenient.For this reason, a more highly diversified, structured and flexible model like a second-order model can be applied and this way establish the optimum value.The secondorder model can be predicted by the following equation (Eq.(6)):  =   + ∑     + ∑       +

Table 2 .
Central composite design analysis: experimental conditions and results.