Review and analysis of defluoridation of drinking water by electrocoagulation

Fluoride is found in all waters, especially groundwater, in several parts of the world including Africa, Asia, southern parts of Europe, USA and the former USSR (Fawell et al. 2006; Brindha & Elango 2011). The WHO guideline for maximum fluoride concentration in drinking water is 1.5 mg/L. At least 25 countries in the world are estimated to be affected by health problems related to excess fluoride in drinking water (Andezhath & Ghosh 2000; Brindha & Elango 2011). Reliance on groundwater is increasing and it is likely that the problems associated with excess fluoride in groundwater will increase in the future. Intake of fluoride via foodstuffs and agricultural produce is also becoming a matter of concern (Mumtaz et al. 2014).

Treatment technologies used for the removal of fluoride include coagulation-precipitation, adsorption, reverse osmosis, ion exchange, nanofiltration, membrane separation, chemical treatment and electrocoagulation (Gwala et al. 2011). Singh et al. (2013) have compared these methods in detail. Electrocoagulation (EC) is a process in which the anode metal dissolves, leading to the formation of metal-hydroxide complexes. These complexes act as coagulants and are responsible for the removal of dissolved contaminants along with turbidity from water. Colloidal particles which contribute to turbidity are removed by charge neutralization and electrostatic attraction, while dissolved contaminants are removed by adsorption on coagulant floc and metal-ligand formation (Amirtharajah & O'Melia 1990; Daneshvar et al. 2006). Aluminium and iron are the most commonly used anode materials. Reactions at the anode and cathode that are widely accepted are noted below (Mollah et al. 2004; Gomes et al. 2007).

Electrocoagulation has several advantages over conventional coagulation (CC) (Zhu et al. 2005; Dolo et al. 2010). No external salts or polymers need to be added in EC; instead, the coagulant is generated in situ. The hydrogen gas evolution at the cathode raises the pH and provides a buffering effect due to which post-treatment neutralization is generally not required. In the case of CC, addition of coagulants like alum or ferric chloride results in the formation of acidic compounds which lower the pH and may need to be neutralized. A few researchers have reported better defluoridation efficiencies with EC as compared to CC (Zhu et al. 2007). However, optimization of operating parameters is required to make EC economically viable.

EC has been used successfully for treating wastewater since 1887 and for drinking water since 1889, but with limited success (Vik et al. 1984). The objective of this paper was to review electrocoagulation studies for defluoridation of water and examine the factors influencing fluoride removal in drinking water. At the time of writing, 56 peer-reviewed publications were accessed and their results are discussed in this paper. The 56 publications accessed for this review included a total of 83 experimental studies since many publications have reported multiple studies with various operating factors.

Electrocoagulation efficiency depends on several factors which are discussed below.

• Reactor design: Type of reactor, configuration, duration of experiment and hydraulic retention time.

• Electrodes: Materials, configuration and number, spacing, shape and combinations of metals/materials.

• Feed water: Solutions of distilled or deionized water, tap water, groundwater and synthetic water.

• Solution chemistry: pH, conductivity, interfering or competing ions and temperature.

Different reactor types and configurations have been used by various research groups. Contaminants are removed using electrocoagulation through four basic processes: electrolysis, flotation, settling and filtration. Most researchers have combined these processes in either one or two stages in their reactors.

Electrocoagulation can be conducted in batch or continuous-flow reactors. Batch studies can be useful for decentralized treatment, e.g., in rural areas where small quantities of water are required and can be generated as and when required in small community systems. Continuous-flow treatment can be developed as an alternative to CC in large-scale water treatment plants that are common in cities and towns.

A summary of the batch studies reviewed for this paper is provided in Table 1. Of the 83 studies reviewed for this paper, 67 studies were in batch mode (44 publications) with maximum removal efficiencies (RE) ranging from 43% to 100%.

Of the 44 publications listed in Table 1, 32 publications did not report settling. Filtration was provided by all researchers for analytical purposes. Many research groups analysed their samples immediately after electrolysis and membrane filtration (asterisks in Table 1). Ün et al. (2010) and Ün et al. (2013) analysed their samples after electrolysis and centrifugation.

Only 12 publications were found for continuous-mode EC and are summarized in Table 2. Maximum removal efficiency of 95% was achieved under optimum conditions with seven electrodes (Guzmán et al. 2016). The duration of these studies ranged from 1 to 36 hours.

Reactor configuration plays an important role especially in continuous-flow studies. The basic geometry of the reactor, placement of the electrodes, flow direction, presence of baffes, inlet and outlet positions and various other factors can influence defluoridation efficiency. It is also important to design the reactors so that the dynamic voltage (IR) drop is minimized, the accumulation of bubbles between the electrodes is avoided and mass transfer is not hindered in any way (Paul 1996).

Many researchers have used a combination of EC with EF for fluoride removal (Emamjomeh et al. 2004; Apshankar & Goel 2017). Zuo et al. (2008) used a combined EC chamber, a flocculation-enhanced chamber and an EF chamber for defluoridation. Two different sets of electrodes were connected to the same DC power supply in different configurations – bipolar in the EC cell and monopolar in the EF cell. The main objective of the EF cell was to separate the flocs formed in the EC cell. Emamjomeh & Sivakumar (2005, 2009) used a conventional reactor which included separate sedimentation and flotation tanks.

Most batch reactors were cylindrical beakers (or buckets) into which electrodes were inserted in monopolar or bipolar mode, connected to a power supply and mixed with magnetic stirrers. However, some studies were conducted with rectangular batch reactors (Mameri et al. 1998; Emamjomeh et al. 2004; Emamjomeh & Sivakumar 2005; Ghosh et al. 2008, 2011; Behbahani et al. 2011a, 2011b; Orescanin et al. 2011a, 2011b; Naim et al. 2012, 2015; Takdastan et al. 2014; Thakur & Mondal 2016). Baffles were used by several researchers in continuous-flow reactors to increase mixing and, therefore, removal efficiency (Zuo et al. 2008; Sandoval et al. 2014; Guzmán et al. 2016; Apshankar & Goel 2017). An exception in these designs was the flow column electrocoagulation reactor (FCER) used by Hashim et al. (2017), in which perforated stacked aluminium plates were used, of which six were submerged and the rest were above water (unsubmerged). An electric current was applied to the submerged electrodes while the unsubmerged electrodes acted as water diffusers and assisted in the aeration and mixing process. Apshankar & Goel (2017) found that increasing the depth–width ratio from 0.96 to 2.14 increased the removal efficiency from 36% to 46%. The laminar flow conditions in their reactor created a floc layer at the surface which enhanced defluoridation efficiency.

Reactors can also be classified as tank-type or horizontal-flow reactors and up-flow or vertical flow reactors. Tank-type reactors are most commonly used because of their ease of operation and simplicity.

Some researchers used packed-bed reactors to ensure proper contact between the contaminant and the electrodes. Others used continuous filter-press reactors when operating in continuous mode (Sandoval et al. 2014; Guzmán et al. 2016). Perforated plate reactors, solid-tube reactors and perforated-tube reactors were reported in a review paper by Mollah et al. (2004).

Andey et al. (2013) designed pilot-scale hopper bottom reactors which recirculate water in the EC tank to keep floc in suspension. External-loop airlift reactors (ELAR) were used by Bennajah et al. (2010a, 2010b), where electrolytic gases produced in situ assisted in the circulation of the reactor solution. ELARs were also used by Essadki et al. (2009, 2011). Iron cylindrical reactors served as anodes in Ün et al.'s (2010) study while a mechanical stirrer consisting of two iron blades served as cathodes. A similar model was used in their next study, where iron was replaced by aluminium in both the anode and the rotating impeller cathode with four blades (Ün et al. 2013). Vasudevan et al. (2009, 2011) used a water jacket outside their electrolytic cell to control the temperature. Other innovative designs include an all-in-one EC chamber, a flocculation enhanced chamber and an EF chamber (Zuo et al. 2008). They also inserted perforated plates in the flocculation chamber with holes of diameters varying from 1 mm at the bottom to 2 mm at the top.

Guzmán et al. (2016) used a pre-pilot-scale filter press cell in which coagulant was produced. The resulting solution was then passed through a serpentine array of perforated baffles to induce better mixing and fluid turbulence. A similar apparatus with 5 mm diameter holes was used to retain bigger floc inside the reactor (Sandoval et al. 2014). The flocculator had a serpentine arrangement to enhance the flocculation process. An unglazed clay diaphragm was used to separate the anode and cathode baths by Kodikara et al. (2015) with different volumes while maintaining the same water level.

Very few researchers have tried mathematical modelling of the hydrodynamic conditions of their reactors for any given flow. Such modelling can be useful for reactor design where different reactor geometries can be evaluated and the best possible option chosen for implementation. This can help in saving time and money involved in conducting experimental studies for different reactor geometries.

Removal of contaminants by electrocoagulation is generally dependent on two processes within the reactor: electrolysis time (also called reaction time) and settling time. Electrolysis time is the period during which coagulant is generated in situ due to the passage of current. In some studies, settling time is provided simultaneously along with electrolysis, while in other studies, settling time is separated from electrolysis time by switching off the power supply. Settling time was not provided in several studies.

Electrolysis time determines the amount of aluminium or iron ions dissolved in solution, and hence, greatly affects removal efficiency (Sivakumar & Emamjomeh 2006). Figure 1 shows the impact of electrolysis time on fluoride removal efficiency using distilled water solutions of NaF in batch reactors (Babu & Goel 2013). No settling time was provided in this study.

Settling time in batch reactors is necessarily separate from electrolysis time. The same may or may not be true in continuous-flow reactors. Results for different contaminants like turbidity, total organic carbon (TOC), nitrate and fluoride in batch studies indicate that settling is important for the removal of suspended solids, i.e., turbidity, but not for dissolved compounds like TOC, nitrate and fluoride (Adapureddy & Goel 2012; Apshankar & Goel 2017).

Hydraulic retention time (HRT) is a function of flow rate, and flow rate influences the rate of dissolution of coagulant ions into the solution, which in turn affects fluoride removal (Sivakumar & Emamjomeh 2006). Low flow rates mean that a higher coagulant dosage is available for the same concentration of pollutants, which in turn results in higher removal efficiency. When flow rates were increased, defluoridation efficiency decreased for the same voltage or current density in each of these studies (Emamjomeh & Sivakumar 2005, 2009; Sinha et al. 2012; Mumtaz et al. 2014; Sandoval et al. 2014; Guzmán et al. 2016; Kim et al. 2016; Apshankar & Goel 2017). However, Zuo et al. (2008) found in their study that HRT has a small or negligible effect on defluoridation. The minimum time tested was 20 min and 30 min was found to be optimum. After 30 min, the effect of HRT on fluoride removal was insignificant.

The removal of contaminants depends upon the type of hydroxide formed, which in turn depends on the electrode material used. Aluminium and iron or steel are the most common materials for electrodes and the number of papers along with the electrode material and highest removal efficiencies are summarized in Table 3. The choice of electrodes is based on the contaminant to be removed, cost, inertness of the electrodes, final water quality and oxidation potential. Comparatively, more electrocoagulation studies were done using aluminium electrodes than iron electrodes. As a result, mechanisms and complexes formed with aluminium have been thoroughly studied while the same with iron electrodes remain to be fully explored. Mechanisms for removal with iron electrodes have been proposed by several researchers and were reviewed by Mollah et al. (2001).

Vasudevan et al. (2009) have noted that the usage of a contaminant-free ion source allowed maximum adsorptive removal of various metals. On experimenting with magnesium, magnesium alloy, aluminium and mild steel, they obtained the highest F removal efficiency (96%) with magnesium alloy, followed closely by aluminium, and the lowest RE was for magnesium (53%). Better defluoridation removal efficiencies have been obtained with aluminium electrodes due to the formation of cationic, hydroxyl-aluminium complexes (Holt et al. 2002; Ghosh et al. 2008; Emamjomeh & Sivakumar 2009; Essadki et al. 2009; Behbahani et al. 2011a; Govindan et al. 2015). However, these complexes exist only below pH 6 (Hao & Huang 1986). Two mechanisms, namely, surface complexation or electrostatic attraction, are responsible for contaminant removal (Singh et al. 1998). Many researchers have obtained equally good removals with iron electrodes (Bazrafshan et al. 2007; Babu & Goel 2013). A few have found that both metals give similar removal efficiencies (Bazrafshan et al. 2012). Combining aluminium and iron electrodes was also found to give good results (Phalakornkule et al. 2010; Orescanin et al. 2011a; Zhao et al. 2011).

When current density is constant, increasing the surface area of the electrodes increases removal efficiency (Zhang et al. 2014).

Electrodes can be arranged in monopolar or bipolar mode. In monopolar mode, each electrode is connected to the power supply, i.e., the ones connected to the positive terminal are anodes and the ones connected to the negative terminal are cathodes. This is the most popular arrangement of electrodes. In bipolar mode, only electrodes at the ends are connected to the power supply; one to the positive terminal and the other to the negative terminal. The electrodes in between develop a charge due to induced polarization when voltage is applied to the electrodes at the end. This configuration has been found by many researchers to increase contaminant removal efficiency (Hu et al. 2003, 2007; Ghosh et al. 2008; Zuo et al. 2008; Behbahani et al. 2011a; Zhao et al. 2011; Bazrafshan et al. 2012; Naim et al. 2012, 2015; Takdastan et al. 2014). However, bipolar mode also leads to greater energy consumption and greater dissolution of the sacrificial anode.

Another important parameter that determines removal efficiency by EC is electrode surface area/reactor volume (S/V or A/V) ratio. Zhang et al. (2014) experimented with the number of plates keeping current density constant. They found that an increase in the number of plates led to an increase in defluoridation, and an increase in energy consumption. Takdastan et al. (2014) kept the voltage constant, and observed the same. Therefore, an optimum number of plates was selected to minimize energy consumption and save electrode material. The increase in area was related to current density, coagulant dose or the electrode S/V or A/V ratio. According to Mameri et al. (1998), the rate of formation of fluoro-aluminium complexes was directly related to current density up to an optimum value, which in turn, depended upon the S/V ratio. Hu et al. (2003) used a value of 40.9 m⁻¹, higher than the value recommended by Mameri et al. (1998). Thus, it was concluded that high S/V ratios were needed, whereas it was recently proved that low S/V ratios give better defluoridation efficiency (Essadki et al. 2009). Mass ratio or the ratio of electrode metal ions to fluoride ions has been considered as a criterion by some researchers (Emamjomeh et al. 2004). Others have varied the current concentration, i.e., the current-to-volume ratio (Emamjomeh & Sivakumar 2006).

In general, an optimum inter-electrode spacing was found in most studies where removal efficiency was maximum (Escobar et al. 2006; Sivakumar & Emamjomeh 2006; Ghosh et al. 2008, 2011; Adapureddy & Goel 2012; Takdastan et al. 2014; Zhang et al. 2014). Behbahani et al. (2011a) found in their study that inter-electrode distance affected cell resistance and as a result, cell voltage. Narrow gaps enhance mass transfer characteristics and decrease ohmic losses. However, a decrease in the gap also led to increased electrolyte resistance, especially when gaseous products are present (Paul 1996). Hence, the optimum distance was reported to be around 20 mm.

Several terms have been used in the literature to express the amount of charge applied. Current density is the amount of current passed per unit of effective surface area submerged in the solution (amperes/unit area), whereas charge loading is the charge passed per unit area (faraday/unit area). The current passing through an EC reactor determines the coagulant dosage, which in turn, influences removal efficiency, bubble production rate and the size and growth rate of the flocs formed (Holt et al. 2002; Bazrafshan et al. 2012). As applied voltage or current increases, removal efficiency also increases. However, the increase in current density will also increase passivation of electrodes, and hence, an optimal current density has been suggested (Zhu et al. 2007; Zhang et al. 2014). Operational costs are also directly linked to the passage of current and reaction time. Alternating current (AC) has been used to avoid passivation, but Ghanizadeh et al. (2016) found that direct current (DC) was more effective than AC while Vasudevan et al. (2011) found AC to be more effective than DC.

Fluoride removal followed first-order kinetics with respect to initial fluoride concentration (Mameri et al. 1998; Sivakumar & Emamjomeh 2006). Reaction rate changed with initial concentration and it has been suggested that defluoridation follows a pseudo-first order model (Bennajah et al. 2010a; Mounir et al. 2010).

The use of response surface methodology (RSM) as a tool in the design of experiments, model making, finding the optimum values and significance of factors is being explored by many researchers. Behbahani et al. (2011b) and Thakur & Mondal (2016) applied RSM to evaluate the effect of initial pH, reaction time, current density and initial fluoride concentration, and obtained high R² values. Behbahani et al. (2011b) found that all parameters except initial pH made a significant difference, while Thakur & Mondal (2016) found that the initial pH, run time and interaction of current density and run time were highly significant factors. However, both these studies observed current density, run time and their interaction to influence operating costs significantly.

Zhu et al. (2007) fit Freundlich and Langmuir isotherms to their data and found that the Langmuir isotherm was a better fit. Hu et al. (2007) developed a variable-order-kinetic (VOK) model using Langmuir adsorption isotherm while Bennajah et al. (2010a) showed that a VOK approach coupled with a Langmuir–Freundlich adsorption model was suitable. In another study, the model was applied successfully to an airlift reactor accounting for the degree of mixing and coagulation beyond monolayer deposition which takes place in large reactors (Mounir et al. 2010). Multi-objective optimization (Sinha et al. 2014), multivariate studies (Naim et al. 2015) and Taguchi experiments (Sinha et al. 2015; Ghanizadeh et al. 2016; Sinha & Mathur 2016) have also been used by researchers. A notable fact is that all these modelling studies were conducted with aluminium electrodes only.

An important factor that is known to influence removal efficiency is the type of feed water. The number of studies conducted with each type of feed water and their corresponding maximum removal efficiencies are shown in Table 4. Several studies were done with either tap water (TW) or drinking water (DrW), i.e., 16 out of the 83 studies reviewed in this paper (19.3%). Most studies were done with deionized (DI) water (26 out of 83 or 31.3%) or groundwater (GW) (16 out of 83, 19.3%). The remaining studies were done with either double distilled water (DD) or distilled water (DW) (19 out of 83 or 22.9%) or synthetic water (6 out of 83 or 7.2%). In all these studies, maximum removal efficiency reported was 97% for GW, 92.7% for synthetic water, 99.5% for tap water and >99% for DW or DI. The greater purity of distilled water or deionized feed water and lack of competition from other ions in solution accounts for the higher removal efficiencies observed in these feed waters.

Since pH determines the speciation of aluminium or iron ions, it plays a crucial role in all coagulation processes including EC (Stumm & Morgan 1981; Amirtharajah & O'Melia 1990). There are two major issues with respect to pH during EC:

• i.

  Change in pH during the EC process: The pH of the solution increases as electrocoagulation progresses due to hydrogen evolution at the cathodes, and liberation of hydroxide ions from metal hydroxides. The final pH is mostly basic and is a major advantage over CC where the final acidic pH often needs to be neutralized to meet the desired range of 6.5 to 8.5. However, if the final pH is more than 8.5, neutralization will again be required.

• ii.

  Effect of initial pH: Initial pH of the solution has been found to affect removal efficiency. The optimum initial pH was observed to be around 6 to 7 (Chen et al. 2000; Adhoum et al. 2004; Zuo et al. 2008; Vasudevan et al. 2009, 2011; Ün et al. 2010; Behbahani et al. 2011a; Gwala et al. 2011; Babu & Goel 2013; Ün et al. 2013). Some researchers have observed a buffering effect where the pH of the solution increases if the initial pH was acidic and decreases if the initial pH was basic (Chen et al. 2000; Adapureddy & Goel 2012; Sharma et al. 2014). However, others have found that the initial pH does not influence removal efficiency significantly (Zhao et al. 2011; Behbahani et al. 2011b; Bazrafshan et al. 2012). Under acidic conditions, below pH 4.5, Hao & Huang (1986) found that hydrogen ions react with fluoride forming HF, while at alkaline pH, they are more likely to react with HCO₃⁻ and OH⁻ ions.

Solution conductivity can be increased by adding salts or salt solution to the water thereby improving removal efficiency (Lin & Wu 1996; Ün et al. 2010; Behbahani et al. 2011b; Bazrafshan et al. 2012; Thakur & Mondal 2016). However, increasing conductivity from 100 μmhos/cm to 10,000 μmhos/cm helped reduce contact times for the same experimental conditions only marginally (Emamjomeh et al. 2004). Ün et al. (2010) reported lesser energy consumption with increase in conductivity of their solution because of decreased resistance.

The effects of competing or interfering ions on fluoride removal efficiency were studied using synthetic solutions under controlled conditions and are summarized in Table 5. In general, most anions (sulphate, phosphate, nitrate and chloride) had a negative effect on defluoridation due to competition with F⁻ especially when aluminium electrodes were used.

The anion with the most negative impact on fluoride removal efficiency was sulphate, followed by nitrate (Hu et al. 2007). Zuo et al. (2008) suggested that ion exchange competition between SO₄²⁻ and F⁻ ions may be the reason as they found traces of S during sludge analyses. In another study, phosphate was also found to have a negative impact on fluoride removal efficiency (Ün et al. 2013). Govindan et al. (2015) used salts of Ca, Mg and Al to evaluate the impact of different cations and anions on F removal. Al cations were effective as coagulants while Ca and Mg were not significantly effective. Using EC along with these salts improved fluoride removal. Comparison of Al salts showed that maximum F removal was obtained with chloride followed by nitrate and sulphate. Calcium was found to have a positive impact on fluoride removal efficiency by Zuo et al. (2008) but not by Montero-Ocampo & Villafañe (2010). Mg²⁺ and Ca²⁺ assist in electrocoagulation probably due to the formation of nano-crystalline MgF₂ and CaF₂ (Shen et al. 2003; Zuo et al. 2008; Montero-Ocampo & Villafañe 2010; Govindan et al. 2015). Silicate, carbonate and arsenate were also reported to affect defluoridation negatively (Vasudevan et al. 2011). Further insight into the influence of ions affecting defluoridation efficiency positively or negatively can help in improving fluoride removal from waters like ground and river water where a mixture of ions are present.

Initial contaminant concentration also influences its removal efficiency. Most papers indicate that an increase in initial concentration reduces F⁻ removal efficiency (Zhu et al. 2007; Zuo et al. 2008; Emamjomeh & Sivakumar 2009; Behbahani et al. 2011a, 2011b; Babu & Goel 2013; Takdastan et al. 2014). Figure 2 shows a decrease in removal efficiency from 84.9% at an initial fluoride concentration of 10 mg/L to 63.6% at an initial fluoride concentration of 50 mg/L (Babu & Goel 2013). These results were obtained after passing current through a batch reactor containing a distilled water solution for 180 min and filtering the sample for fluoride analysis through a 0.45 micron cellulose nitrate filter.

However, Vasudevan et al. (2011) found that fluoride adsorption increased on increasing initial concentration and remained constant after 90 min using AC and after 120 min using DC. Increasing fluoride concentration may have increased the ratio of aqueous fluoride to available complexation sites, thereby decreasing fluoride removal. An increase in initial concentration has also been found to increase the amount of sludge produced (Ghosh et al. 2008) and the cost of treatment (Ghosh et al. 2011).

Zhang et al. (2014) found that increasing temperature from 18 to 38 °C increases fluoride removal efficiency. In experiments with geothermal water, it was observed that removal efficiency increased and energy consumption decreased when the temperature was increased from 18 to 45 °C, resolublization of fluoride occurred when the temperature of water increased from 45 to 55 °C (Wang et al. 2017). However, Narasimham & Silaimani (1992) found that a temperature of 313 °K (39.85 °C) gave the best removal efficiencies in the range of 303–323 °K. Thus, it can be concluded that increase in temperature up to a certain point aids in removing fluoride, after which it will inhibit it. In both batch and continuous-flow experiments conducted by the first author (unpublished data), the temperature of the solution in the electrocoagulation reactor was found to increase over a period of time due to heat generated at the electrodes. In one continuous-flow experiment, temperature increased from 27.9 °C to 30.37 °C from the start of the experiment to its end (Figure 3). Temperature rise can be seen even more clearly in the batch experiment, where the temperature increased from 24.6 °C to 27.48 °C during electrolysis and decreased after that (Figure 4). Conductivity decreased after 3 hours in the batch reactor due to coagulation and settling of sludge. Temperature studies conducted by Vasudevan et al. (2011) showed that adsorption was spontaneous and exothermic in nature in the temperature range they studied (313–343°K).

Identification of elements or compounds present in sludge and electrode material can help to identify reactions taking place during electrolysis. Many researchers have analyzed sludge and electrode surfaces using SEM, EDS, TEM, FTIR and XRD.

SEM analyses of EC sludges show the presence of crystal structures (Figure 5(a) and 5(b)). Govindan et al. (2015) showed that the sludge particles or floc formed by iron electrodes were much smaller in size as compared to those formed by aluminium electrodes. Guzmán et al. (2016) reported irregular aluminium flocs of around 500 μm size. Hu et al. (2003) and Vasudevan et al. (2009) have shown SEM images with fine coagulant particles on the surfaces of their anode. The whitish by-products of sludge found by Ghosh et al. (2008, 2011) contained mostly aluminium hydroxides since the amount of dissolved electrode material was much higher than the quantity of fluoride. Essadki et al. (2009) found that the sludge generated at pH 4 was slightly more porous than the sludge generated at pH 7. A large number of indentations on the surface of the anode were observed by Vasudevan et al. (2011). Less disordered pores were formed by AC as compared to DC, which indicated uniform dissolution of aluminium electrodes as opposed to consumption of the electrodes at active sites due to oxygen generation at the surface.

The formation of alumina-fluoro complexes has been verified by EDAX analysis wherein Al, F and O were detected by Ghosh et al. (2008, 2011) and Vasudevan et al. (2011). Oxygen and aluminium were always present in the flocs analyzed by Guzmán et al. (2016). The occasional presence of silica indicated the formation of aluminium hydro-silicates and small quantities of arsenic and fluoride were also found. Arsenic, fluoride and aluminium were present in EC sludges of Thakur & Mondal (2016). Sinha et al. (2015) confirmed entrapment of fluoride within aluminium-hydroxide complexes in EDAX analysis. The elements detected in sludges depend on the type of feed water used. C, O, Si and Fe were found in sludge and deposits on electrodes (Babu & Goel 2013). However, fluoride was found only in the sludge; no traces of fluoride were detected on the electrodes. Essadki et al. (2009) obtained greater fluoride content in their sludge at pH 4, whereas the amounts of oxygen and aluminium increased as the pH increased, a fact partly confirmed by XRF-WDS analysis.

Stretching bonds of H-O-H, Al-O, Al-F-Al and Al-O-H were detected by FTIR analysis (Ghosh et al. 2008, 2011; Vasudevan et al. 2011; Govindan et al. 2015; Sinha et al. 2015). O-H, Na-F, Al-O-Si, Si-O and As-O bonds were reported by Guzmán et al. (2016), who found mainly aluminium silicates in aluminium sludge. Si-O-Al bonds were also reported in the study by Sinha et al. (2015) since they used bentonite for the removal of alumina-fluoro complexes. Thakur & Mondal (2016) found O-H, Al-H and Al-O stretching bands. Wave numbers in all studies ranged from 400 to 4,000 cm⁻¹. Mg-O and Mg-F were found in sludge by Vasudevan et al. (2009).

When aluminium electrodes dissolve, they form several different aluminium-fluoride-hydroxide complexes (Al(OH)_3-xF_x) depending on the solution pH (Sivakumar & Emamjomeh 2006; Zuo et al. 2008). These complexes have been identified by several researchers using XRD along with FTIR in their sludge analysis. XRD analysis of aluminium sludge showed that the phase was amorphous or poorly crystalline in nature (Ghosh et al. 2008, 2011; Vasudevan et al. 2011; Guzmán et al. 2016). Aluminium hydroxide flocs of β-Al(OH)₃ were confirmed by Govindan et al. (2015). Other probable compounds were AlF(OH)₂. Emamjomeh & Sivakumar (2005, 2009) have reported aluminium fluoride hydroxide and aluminium hydroxide (bayerite) in their sludge. Their XRD results indicated competition between OH⁻ and F⁻. Na₃(AlF₆), AlF₃(H₂O)₃ and AlO(OH) were reported by Sinha et al. (2015), which indicates aluminium-fluoride-hydroxide complex formation. Zhao et al. (2010) added calcium and magnesium in their feed water to analyze their effects on defluoridation and found that four wide diffraction peaks were observed in the system when magnesium was added as compared to calcium. They inferred that Mg-Al layered double hydroxides were being generated in the Mg²⁺ containing solution.

Iron oxide hydroxide crystals (Fe₂₁HO₃₂) have been reported by Govindan et al. (2015) and Apshankar & Goel (2017). However, no compound with fluoride was observed in XRD analysis with iron sludges. Govindan et al. (2015) attributed this to the similar ionic radii of fluoride and the replaced oxygen atom. This may also be due to the low concentration of fluoride relative to Fe.

Use of electrocoagulation for wastewater treatment has been studied to a far greater extent than for drinking water treatment. A limited number of studies were found in the literature regarding the use of electrocoagulation for fluoride removal in drinking water treatment. Several factors were found to influence defluoridation efficiency and the most important factors are reactor configuration, applied voltage, electrolysis time, initial contaminant concentration, pH and the presence of competing ions. All studies reviewed here are bench-scale or pilot-scale studies and no reports of full-scale drinking water treatment using EC for defluoridation were found. Most studies have been done with aluminium, and fluoride removal mechanisms with aluminium are understood to some extent. The mechanisms and chemistry of defluoridation with iron electrodes are still not clear.