Application of Electrocoagulation for the Removal of Transition Metals in Water

Abstract

Urban and industrial effluents, stormwater, road runoff, agricultural runoff, urban or mine waste deposits and fuel storage sites can lead to the contamination of water sources with compounds that are hard to biodegrade, such as heavy metals, whose removal requires advanced and expensive technologies. The Sustainable Development Goals (SDGs) established by the UN and the current requirements in terms of energy efficiency, reduction of carbon emissions, water reuse, waste valorization and preservation of public health, have led to a rethink concerning the typology of technologies for the treatment of water and the production of drinking water. Electrocoagulation (EC) stands out in this scenario due to its high efficiency in the removal of several pollutants, production of low sludge volumes and adaptability to the use of renewable energies. This is in addition to the ease with which it can be combined with other water treatment technologies. This work presents a literature review to systematize the use of EC for the removal of transition metals in water to produce drinking water, since these elements are present in several natural water sources and are parameters used in the legislation of many countries for the quality control of drinking water. The works found were analyzed in detail, and relationships between pre-set variables and categories were determined through regression analysis. Generally, it was found that EC is a highly efficient technology for the removal of transition metals in water (above 75% for most metals), specifically through parallel plates technologies with iron and aluminum electrodes with a minimum spacing of 1 cm and density minimum applied current of 10 A/m².

1. Introduction

Water availability from natural sources has been decreasing due to the growing extraction for use in agriculture, as well as industrial and urban uses. At the same time, water pollution is increasing due to the growth of activities that generate wastewaters (e.g., urban, domestic and industrial activities) and sites that generate leachates (e.g., sanitary landfills, road runoff, mining and agricultural activities and fuel storage sites), which reach surface or groundwater. The main pollutants of concern are organic molecules such as catechol, resorcinol, p-nitrophenol, 4-chlorophenol, sulfamethazine, 1H-benzotriazole, 4-methyl-1H-benzotriazole, carbamazepine, perfluorooctane sulfonate (PFOS), perfluorooctanoate, polyethylene, polypropylene, polystyrene [1,2,3,4,5,6], heavy metals [3,7,8,9] and nutrients (especially nitrogen and phosphorous) [10,11,12,13], as well as pathogenic microorganisms (viruses, bacteria, fungi, protozoa, helminth) [14,15,16,17], which are also a threat to public health.

Nowadays, emergent pollutants (EPs), namely pharmaceutical compounds, pharmaceutical and personal care products (PPCPs), antibiotics, flame retardants, industrial additives, endocrine-disrupting chemicals (EDCs), hormones, and pesticides [18,19,20], and new or opportunist pathogens are a growing threat for the environment as well as for public health, namely antibiotic-resistant bacteria (ARB) and enteric opportunistic pathogens (EOPs) [21,22,23]. The NORMAN network identified around 700 substances in the European aquatic environment that were classified into 20 classes of EPs [24].

New threats of pandemic-based viruses are emerging, such as the ARS-CoV-2 RNA, which has been detected in wastewaters in different countries. Furthermore, the risk of exposure for sewage workers, potential contamination of surface waters and long-term persistence of viral RNA in wastewater has been reported [25,26,27,28,29]. Pashaei et al. (2022) [30] assessed pharmaceutical contamination in 25 countries and microplastic pollution in 13 countries. They identified 75 Eps in surface water, groundwater, and wastewater influents and effluents and concluded that the COVID-19 pandemic enhanced contamination from EPs in aquatic environments.

The expansion of urbanization and population growth is exerting water demand pressure and a deterioration of quality. Unsustainable practices of soil fertilization and management and treatment of waste and wastewater are among factors that lead to deterioration of the quality of natural waters. This also has economic impacts when producing drinking water, because the more pollutants that are in the raw water, the greater the need for technologies and chemicals for their treatment in water treatment plants (WTP) and the higher the cost for users.

The last United Nations’ report [31] on valuing of water expresses that more than 140 low- and middle-income countries do not have universal access to safe drinking water and sanitation. Further, it is necessary to incorporate decision-making tools to control how water is valued to achieve sustainable and equitable water resources management and Sustainable Development Goal (SDG) No. 6 (“Ensure access to water and sanitation for all”) set by UN for 2030.

Among several contaminants produced in urban and industrial activities, transition metals may pose a threat to water quality at the source and to public health. Industries with activities in the areas of electromechanical, chemical, pulp and paper, textile, paint and pesticide production, tanneries, foundries, smelters, oil refineries, petrochemical plants, some mining operations, and also road runoff and stormwater runoff, produce effluents with a varied and wide concentration of heavy metals, which can be toxic or harmful to humans and other biological systems when discharged into water sources [32,33], as shown in Figure 1. Metals are also present in natural waters due to their dissolution from soil minerals, but normally at concentrations that are not harmful [34,35]. Therefore, anthropogenic metal sources are responsible for the increase of their concentration in natural waters (both surface and groundwater) [7,8,32,33,34].

Transition metals or transition elements can be described as the elements in the d-block (i.e., groups 2A to 3A) on the periodic table [36], as shown in Figure 2. They are chemical elements with valence electrons in two shells instead of only one (i.e., electrons that can participate in the formation of chemical bonds). The term “transition” has no chemical significance, but it means there are close atomic structures, resulting in similar properties among the metals (e.g., excellent conductors of heat and electricity, malleable, tend to be very hard and have high melting points).

Transition elements of the fourth period (22  ≤  Z  ≤  30), scandium (Sc), titanium (Ti), vanadium (V), chromium (Cr), manganese (Mn), iron (Fe), cobalt (Co), nickel (Ni), copper (Cu) and zinc (Zn), are the most common in waters contaminated with metals through anthropogenic sources [7,8,32,33,34]. However, cadmium also appears (Cd, Z = 48, fifth period) as well as mercury (Hg, Z = 80, sixth period).

When those metals are dissolved in water, they are difficult to degrade and can accumulate in sediments and in water plants and animals, inhibiting some of all water uses (e.g., irrigation, environmental or social uses, aquaculture or even abstraction for producing drinking water), thus resulting in a lasting threat to public health [32,33]. Those elements can be precursors to several diseases such as myocardial infarctions [37], hemochromatosis and liver cancer [38] due to excessive and prolonged exposure to Fe, lung cancer due to hexavalent Cr [39] and nasal cancer due to Ni [40]. Cd, Cr, Hg and Pb are very toxic to biological life and are precursors of several cancers [41,42]. In addition, the excess of metals can cause organoleptic rejection of water consumption, since iron imparts color, turbidity and flavor, and incrustation problems in materials and equipment used for water transport, treatment and storage [43].

High concentrations of Cd, Ni, Pb and Zn have been observed in water in rainy and dry periods according to serval research works [44,45]. Zhou et al. (2020) [34] analyzed data on 12 heavy metal concentrations presenting in 168 rivers and 71 lakes, between 1972 and 2017, in different countries in Asia, Europe and Africa, nine of which were transient elements (Cd, Co, Cr, Cu, Fe, Hg, Mn, Ni, and Zn). They concluded that, generally, their concentrations have been increasing in sediment and water over the years because of anthropogenic pollution. Alves et al. (2010) [46] compared measurements of Cd, Cr, Cu, Mn and Zn in surface water and sediment of the Monte Alegre river (São Paulo, Brazil), in 2004 and 2007, and concluded that concentrations of Zn increased in water and concentrations of Cu and Zn increased in sediment. High concentrations of Cd and Ni were observed by Ferraz et al., 2018 [47], in a study that involved the determination of heavy metals in six wells in Vitoria da Conquista (Bahia, Brazil) over seven months and were associated with both rock formations and anthropogenic sources.

There are several processes for removing transition metals from water, such as ion exchange [48,49], physical and chemical adsorption [50,51,52,53], biosorption [54], biological processes based on activated sludge technologies [55], flotation [56], filtration [41,57,58], application of activated carbon [59], phytoremediation [60,61,62], chemical coagulation [58,63] and electrocoagulation [64,65].

Electrocoagulation (EC) is gaining popularity among these due to the lower volume of sludge generated in treatment, low operating and implantation costs, and the range of treated compounds and possibilities of combination with renewable energy sources and other technology [66,67].

EC is an electrolytic process which consists of applying electric current between two electrodes (an anode and a cathode) that are submerged in an aqueous solution inside a tank/cell. The anode is normally made of a corrosive metal (e.g., aluminum or iron) and, once connected to a power source, it starts to oxidize and corrode into solution as metal cations. Simultaneously, water at the cathode (normally made of aluminum or iron) is hydrolyzed into hydrogen gas and hydroxyl groups. When the electric current is activated, metal ions complex with hydroxyl groups in water and entrain other contaminants such as suspended solids, emulsified oils, organic compounds and microorganisms, producing visible flocs. These flocs are heavier than water and settle into the bottom of the tank/cell [64,65,66,67]. Therefore, contaminants are separated from water through a set of processes involving coagulation, flocculation or filtration to achieve a required water quality objective. During the process, the anodes undergo strong dissolution and need to be replaced before the next use.

The removal of pollutants is mainly determined by the amount of electrical current applied, with the lower the current applied, the lower the formation of bubbles at the cathode, and the smaller the flotation and the greater the sedimentation [65,67]. However, other parameters, such as the type of material, pH, conditions of the environment and disposition of anodes and cathodes and concentration of pollutants are also relevant. EC is a viable alternative for removal of heavy metals and other pollutants (e.g., turbidity, apparent color, organic molecules, and pathogens) from water due to its treatment flexibility and high pollutant removal power [64,66].

The main advantages of the technology are facility in removing water pollutants without the need for chemicals; ease of handling with water quality variations; allowing very high removal efficiency and the removal of several compounds and microorganisms simultaneously; pollutants such as metals and nutrients (e.g., N and P) can be recovered for valorization; low sludge (by-product) production; and it can be very effective for low levels of electrical current (current density). Among the main disadvantages are the intensity of the process, which places a lot of strain on the electrodes themselves, resulting in their wear and tear; the need for regular cleaning and maintenance; and fouling and passivation of the electrodes, especially for applications with high contaminant concentrations; and the operational costs [64,68,69,70,71].

EC had its beginnings in the mid-20th century, being applied, essentially, for removing organic compounds from industrial wastewaters [72]. Although some studies were developed for the removal of organic molecules in water through EC, this technology was not economically competitive against the application of chemicals (e.g., aluminum sulfate or ferric chloride) in coagulation-flocculation processes [73]. The costs associated with using energy in EC are often referred to as a major disadvantage. However, technology has been improving in recent years to reduce costs associated with operating systems. Alavijeh et al. [71] carried out a comparative economic analysis, including operational costs and capital costs, among EC, chemical coagulation (CC), integrated EC-CC, and combinations of these techniques with ultrafiltration (UF), for treating baker’s yeast wastewater. They concluded that EC attained better performance, but with greater operating costs and capital costs than the other technology types. However, the technology has evolved in the last decade, with sensors of low production cost and with lower energy consumption having been developed, thus making it competitive in relation to physical and chemical technologies, especially when considering different operating parameters [64,68,74,75]. The low intensities of electrical current used are not dangerous for animals or plants.

Thus, this study intends, through a bibliometric search, to select works on the application of EC to remove transient metals from the fourth period (22  ≤  Z  ≤  30) of the periodical table that are present in waters and to evaluate statistical relationships between the main operational parameters.

2. Materials and Methods

2.1. Bibliometric Search and Analysis

Bibliometric analysis consisted of searching publications in the Scopus database, as it is considered the most comprehensive research database, using the following keywords: “EC ^ water treatment”, “Electrocoagulation ^ water treatment”, “EC ^ WTP” and “Drinking water ^ EC”, using filters to limit the areas of knowledge to: “Biochemistry, Genetics and Molecular Biology”, “Business Management and Accounting”, “Chemical Engineering”, “Chemistry”, “Civil Engineering”, “Computer Science”, “Earth and planetary Science”, “Energy”, “Economics, Econometrics and Finance”, “Engineering”, “Environmental Engineering”, “Environmental Science”, “Mathematics”, “Materials Science”, and “Multidisciplinary”. The search considered all existing Scopus publications up to 31 December 2021.

The selected papers were afterwards evaluated in two stages, configuring the inclusion criteria of the study. The first criterion was reading the title and abstract to verify whether the work had EC as a treatment technology for removing pollutants in any liquid fluid (e.g., water, wastewater or other stream). The second criterion consisted of reading the papers to include only works that investigated the application of EC on surface water, groundwater or manipulated water simulating natural waters or drinking water, thus excluding those that focused on wastewater, marine water, stormwater or other streams, and the following operating conditions (divided into “variables” and “categories”):

Metal (P): only works that investigated the application of EC for removing seven transition metals of the fourth period (22  ≤  Z  ≤  30), Co, Cr, Cu, Fe, Mn, Ni and Zn (Figure 2), because they are the most common in waters contaminated with anthropogenic sources of pollution [7,8,32,33,34];

Experimental arrangement: with parallel plates (PP) and other different arrangements (D), because PP are the most common arrangements [64,68,69,74];

Initial concentration (Ci): as the values can vary greatly due to the different types of water used, it was decided to choose three categories of analysis: below 1 mg/L (Ci ≤ 1), which is closer to the limits defined by the WHO, European Union, Brazil and EPA [43,76,77,78,79] for drinking water (

Table 1. Permissible limits of metal concentrations (mg/L) in drinking water according to WHO, EU, Brazil and EPA.

Metal	WHO [76]	EU [79]	Brazil [43]	EPA [77,78]
Cr	0.05	0.025	0.05	0.1
Co	NL	NL	NL	NL
Cu	2.0	0.002	2.0	1.0
Fe	0.3	0.2	0.3	0.3
Mn	0.5	0.05	0.1	0.05
Ni	0.07	0.02	0.07	0.02
Zn	3.0	NL	5.0	5.0

) and also appears in low contaminated waters [34,80,81,82]; between 1 and 20 mg/L (1 < Ci ≤ 20), as they are the concentrations that most appear in medium contaminated waters [8,44,83,84,85]; and above 20 mg/L (Ci > 20) for high contaminated natural waters [81,83,84];

Inter-electrode distance (Id): it was considered lower than 1 cm (Id < 1) and greater than 1 cm (Id ≥ 1), because most devices use electrodes with Id around or above 1 cm, which are considered more efficient [70,86,87,88];

Current density (CD): it was considered lower than 10 A/m² (CD < 10) and greater than 10 A/m² (CD ≥ 10), because most of the devices use CD around or above 10 A/m², which are considered more efficient [88,89,90,91,92];

pH: it was considered values in acid (pH < 7), neutral (pH = 7) and basic (pH > 7) conditions;

Removal efficiency (RE): it was considered lower than 95% (RE < 95%) and greater than 95% (RE ≥ 95%), because EC is considered a very efficient technology for RE above 95% [64,68,69,74,93];

Cathode material (CM): the most common are in aluminum (CM Al) or iron (CM Fe), but other materials (CM Other) can also be used [68,69,70,74,93];

Anode material (AM): the most common are in aluminum (AM Al) or iron (AM Fe), but other materials (AM Other) can also be used [68,69,70,74,93];

Treatment time (TT): the most common is lower than 60 min (TT < 60), but it can appear applications for more than 60 min (TT ≥ 60) [69,70,93].

The second criterion involved the application of ten variables and 29 categories, according to the type of metal, expected concentration range in water, electrode characteristics and operating conditions of the experiment, as shown in

Table 2. Variables and categories employed in the bibliometric analysis.

Variable	Category	Abbreviation
Metal (P)	Cobalt (Co)	P. Co
	Copper (Cu)	P. Cu
	Chromium (Cr)	P. Cr
	Iron (Fe)	P. Fe
	Manganese (Mn)	P. Mn
	Nickel (Ni)	P. Ni
	Zinc (Zn)	P. Zn
Experimental arrangement (EA)	Parallel plates (PP)	EA. PP
	Different arrangement (D)	EA.D.
Initial concentration (Ci)	≤1 mg/L	Ci ≤ 1
	1 to 20 mg/L	1 < Ci ≤ 20
	>20 mg/L	Ci > 20
Inter-electrode distance (Id)	≥1 cm	Id ≥ 1
	<1 cm	Id < 1
Current density (CD)	≥10 A/m²	CD ≥ 10
	<10 A/m²	CD < 10
pH	Acid	pH < 7
	Neutral	pH = 7
	Basic	pH > 7
Removal efficiency (RE)	≥95%	RE ≥ 95%
	<95%	RE < 95%
Cathode material (CM)	Aluminum (Al)	CM Al
	Iron (Fe)	CM Fe
	Other	CM Other
Anode material (AM)	Aluminum (Al)	AM Al
	Iron (Fe)	AM Fe
	Other	AM Other
Treatment time (TT)	≥60 min	TT ≥ 60
	<60 min	TT < 60

. A matrix database worksheet was built where the columns were the variables and the rows the papers that shifted the categories. The number of rows was related to the pollutants removed by the EC, i.e., the more pollutants worked on in the same study, the more rows for the paper, with the optimal operating variables and the removal efficiency being analyzed for each type of pollutant. A statistical analysis was applied to the matrix as presented in Section 2.2.

2.2. Statistical Analysis

Correlations and dependencies of the categories of Table 2 were analyzed using regression analysis. The measure of correlation between events followed the method proposed by Obreshkov (1963) [94] and detailed by Dimitrov (2010) [95]. The regression coefficient of an event $\mathrm{A}$ in relation to an event $\mathrm{B}$ is defined by Equation (1)$.$

$${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)=\mathrm{P}\left(\mathrm{A}|\mathrm{B}\right)-\mathrm{P}\left(\mathrm{A}|\underset{\_}{\mathrm{B}}\right)$$

(1)

where $\mathrm{P}\left(\mathrm{A}|\mathrm{B}\right)=\mathrm{P}\left(\mathrm{A}\cap \mathrm{B}\right)/\mathrm{P}\left(\mathrm{B}\right)$ is the conditional probability of event $\mathrm{A}$ given event $\mathrm{B}$, such that $\mathrm{P}\left(\mathrm{B}\right)>0$ and $\underset{\_}{\mathrm{B}}$ is the complementary event of $\mathrm{B}$. The conditional probability $\mathrm{P}\left(\mathrm{A}|\underset{\_}{\mathrm{B}}\right)$ and the regression coefficient ${\mathrm{R}}_{\mathrm{A}}\left(\mathrm{B}\right)$ are defined in a similar manner.

According to Dimitrov (2010) [95], the regression coefficient can be interpreted both from the probabilistic point of view (Equation (2)) and from the point of view of regression models (Equation (3)).

$$\mathrm{P}\left(\mathrm{A}|\mathrm{B}\right)=\mathrm{P}\left(\mathrm{A}\right)+{\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)\left(1-\mathrm{P}\left(\mathrm{B}\right)\right)$$

(2)

In probability theory, an event A is said to be independent of an event B if $\mathrm{P}\left(\mathrm{A}|\mathrm{B}\right)=\mathrm{P}\left(\mathrm{A}\right)$, i.e., considering that the occurrence of an event $\mathrm{B}$ does not change the probability of occurrence of an event $\mathrm{A}$. On the other hand, an event $\mathrm{A}$ is positively dependent on an event $\mathrm{B}$ if $\mathrm{P}\left(\mathrm{A}|\mathrm{B}\right)>\mathrm{P}\left(\mathrm{A}\right)$, i.e., the information of the occurrence of an event $\mathrm{B}$ positively alters the probability of the occurrence of an event $\mathrm{A}$, and similarly an event $\mathrm{A}$ is negatively dependent on an event $\mathrm{B}$ if $\mathrm{P}\left(\mathrm{A}|\mathrm{B}\right)<\mathrm{P}\left(\mathrm{A}\right)$.

Then, thorough Equation (2), it can be observed that when ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)$= 0, an event $\mathrm{A}$ does not depend on an event $\mathrm{B}$; when ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)>0$, an event $\mathrm{A}$ positively depends on an event $\mathrm{B};$ and when ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)<0$, an event $\mathrm{A}$ negatively depends on an event $\mathrm{B}$.

The regression coefficient (Equation (3)) can be interpreted as the slope of a simple linear regression model relating two indicator variables associated with events $\mathrm{A}$ and $\mathrm{B}$.

$${\mathrm{I}}_{\mathrm{A}}=\mathsf{\alpha }+{\mathsf{\beta }\mathrm{I}}_{\mathrm{B}}+\mathrm{e}$$

(3)

where ${\mathrm{I}}_{\mathrm{A}}=1,$ if A occurs, and ${\mathrm{I}}_{\mathrm{A}}$= 0, if the contrary, and ${\mathrm{I}}_{\mathrm{B}}$ is defined similarly.

After the definition and interpretation of the regression coefficient ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)$, and knowing how it measures the dependence of an event $\mathrm{A}$ in relation to an event $\mathrm{B}$, some of its properties can be stated as follows:

• The coefficient ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)$ has a value within the range −1 to 1, that is $-1\le {\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)\le 1$.
• The coefficient ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)$ = 1 only when an event $\mathrm{A}$ coincides with an event $\mathrm{B}$, that is, they are equivalent.
• ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)$ = −1 when an event A coincides with an event $\underset{\_}{\mathrm{B}}$.
• ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)$ = 0 if, and only if, events $\mathrm{A}$ and $\mathrm{B}$ are independent. ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)=-{\mathrm{R}}_{\underset{\_}{\mathrm{B}}}\left(\mathrm{A}\right)=-{\mathrm{R}}_{\mathrm{B}}\left(\underset{\_}{\mathrm{A}}\right)$.
• The signs of ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)$ and ${\mathrm{R}}_{\mathrm{A}}\left(\mathrm{B}\right)$ are equal, that is, if an event $\mathrm{A}$ has positive dependence in relation to an event $\mathrm{B}$, then an event $\mathrm{B}$ also has positive dependence in relations to an event $\mathrm{A}$. The same property is true if ${\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)$ is negative.

Using the regression coefficients, Obreshkov (1963) [94] defined a measure of correlation between two events $\mathrm{A}$ and $\mathrm{B}$ through Equation (4)

$${\mathrm{R}}_{\mathrm{AB}}=\mathrm{sign}\left({\mathrm{R}}_{\mathrm{A}}\left(\mathrm{B}\right)\right)\sqrt{{\mathrm{R}}_{\mathrm{A}}\left(\mathrm{B}\right){\mathrm{R}}_{\mathrm{B}}\left(\mathrm{A}\right)}$$

(4)

where $\mathrm{sign}\left({\mathrm{R}}_{\mathrm{A}}\left(\mathrm{B}\right)\right)=1$ if ${\mathrm{R}}_{\mathrm{A}}\left(\mathrm{B}\right)\ge 0$ and $\mathrm{sign}\left({\mathrm{R}}_{\mathrm{A}}\left(\mathrm{B}\right)\right)=-1$ if ${\mathrm{R}}_{\mathrm{A}}\left(\mathrm{B}\right)<0$.

The correlation measure ${\mathrm{R}}_{\mathrm{AB}}$ is a geometric mean of the dependence of an event $\mathrm{A}$ in relation to an event $\mathrm{B},$ and an event $\mathrm{B}$ in relation to an event $\mathrm{A}$, with the following properties:

6.

$-1\le {\mathrm{R}}_{\mathrm{AB}}\le 1$.

7.

${\mathrm{R}}_{\mathrm{AB}}$ = 1 only when events $\mathrm{A}$ and $\mathrm{B}$ are equivalent.

8.

${\mathrm{R}}_{\mathrm{AB}}$ = −1 when an event $\mathrm{A}$ coincides with an event $\underset{\_}{\mathrm{B}}$.

9.

${\mathrm{R}}_{\mathrm{AB}}$ = 0 if, and only if, events $\mathrm{A}$ and $\mathrm{B}$ are independent.

10.

${\mathrm{R}}_{\mathrm{AB}}={\mathrm{R}}_{\underset{\_}{\mathrm{A}\mathrm{B}}}=-{\mathrm{R}}_{\mathrm{A}\underset{\_}{\mathrm{B}}}=-{\mathrm{R}}_{\underset{\_}{\mathrm{A}}\mathrm{B}}$

It is possible to make a probabilistic interpretation of the correlation measure in a similar way to Equation (2), consideration the relation expressed in Equation (5).

$$\mathrm{P}\left(\mathrm{A}|\mathrm{B}\right)=\mathrm{P}\left(\mathrm{A}\right)+{\mathrm{R}}_{\mathrm{AB}}\sqrt{\mathrm{P}\left(\underset{\_}{\mathrm{B}}\right)\mathrm{P}\left(\mathrm{A}\right)\mathrm{P}}$$

(5)

${\mathrm{R}}_{\mathrm{AB}}$ can also be written as the Pearson correlation $\mathsf{\rho }$, which measures the linear correlation between the indicator variables ${\mathrm{I}}_{\mathrm{A}}$ and ${\mathrm{I}}_{\mathrm{B}}$ defined in Equation (3) through Equation (6) [95].

$${\mathrm{R}}_{\mathrm{A}\mathrm{B}}=\mathsf{\rho }\left({\mathrm{I}}_{\mathrm{A}},{\mathrm{I}}_{\mathrm{B}}\right)=\frac{\mathrm{P}\left(\mathrm{A}\cap \mathrm{B}\right)-\mathrm{P}\left(\mathrm{A}\right)\mathrm{P}\left(\mathrm{B}\right)}{\sqrt{\mathrm{P}\left(\mathrm{A}\right)\left(1-\mathrm{P}\left(\mathrm{A}\right)\right)\mathrm{P}\left(\mathrm{B}\right)\left(1-\mathrm{P}\left(\mathrm{B}\right)\right)}}$$

(6)

Dimitrov (2010) [95] considers the ranges of intensity of dependencies based on their absolute distances (R) in relation to zero as a way of categorizing the results of correlations and regression coefficients, namely:

• $0<\mathrm{R}\le 0.05$: almost independent;
• $0.05<\mathrm{R}\le 0.20$: weak dependence;
• $0.20<\mathrm{R}\le 0.45$: moderate dependence;
• $0.45<\mathrm{R}\le 080$: medium dependence;
• $0.80<\mathrm{R}<1$: strong dependence
• when R = 0 and R = 1 variables are independent and totally dependent, respectively.

3. Results and Discussion

A total of 6600 works (mainly research papers) were obtained in the search of the Scopus database with application of the keywords defined in Section 2.1. The application of the first criterion reduced the sample to 2215 works (i.e., works with application of EC), and the application of the second criterion reduced it to 22 research papers (i.e., works with EC, natural or manipulated water, and transient metals of the fourth period of the periodical table), which were selected for the statistical analysis (Figure 3). These works were published between 2008 and 2021. Information for the ten variables and the 29 categories (Table 2) was collected from the 22 research papers and is presented in

Table 3. Characteristics of EC application for water treatment in the 22 research papers.

Author	Water Type	Metal	EA	Ci (mg/L)	Id (cm)	CD (A/m²)	pHo (pHi–pHf)	RE (%)	CM	AM	TT (min)
Ghosh et al. (2008) [96]	Manipulated	Fe	PP	20.00	0.5	0.40	7.5 (7.5–7.8)	100.00	Aluminum	Aluminum	35
Abdulhadi et al. (2020) [97]	Manipulated	Fe	PP	10.00	0.5	30.00	7.0 (4.0–10.0)	99.90	Aluminum	Aluminum	50
Das e Nandi (2020) [98]	Manipulated	Fe	PP	20.00	1.0	43.10	7.0 (5.0–9.0)	99.96	Aluminum	Aluminum	60
Das e Nandi (2019) [65]	Manipulated	Fe	PP	20.00	1.0	20.00	6.3 (6.3–6.6)	98.60	Aluminum	Aluminum	45
Doggaz et al. (2018) [99]	Manipulated	Fe	PP	20.00	1.0	25.00	8 (5.0–8.0)	97.00	Aluminum	Aluminum	60
Silva et al. (2017) [100]	Manipulated	Fe	PP	20.00	0.8	12.50	6.3 (6.0–8.0)	100.00	Aluminum	Aluminum	80
Pan et al. (2016) [101]	Manipulated	Cr	PP	0.50	2.0	10.00	6.0 (4.0–8.0)	99.00	Iron	Iron	30
Thella et al. (2008) [102]	Manipulated	Cr	PP	15.00	1.0	50.00	2.0 (2.0–8.0)	100.00	Steel	Iron	40
Hamdan and El-Naas (2014) [103]	Groundwater	Cr	D	5.00	-	76.10	8.0 (5.0–8.0)	100.00	Iron	Iron	30
Martín-Dominguez et al. (2018) [104]	Groundwater	Cr	D	19.00	0.3	30.00	6.2 (6.2–7.0)	99.70	Carbon Steel	Carbon Steel	0.5
Ferreira et al. (2013) [105]	Groundwater	Cu	PP	12.00	1.5	42.00	7.7 (7.7–7.9)	90.00	Aluminum	Aluminum	180
Ferreira et al. (2013) [105]	Groundwater	Zn	PP	20.00	1.5	42.00	7.7 (7.7–7.9)	90.00	Aluminum	Aluminum	180
Ferreira et al. (2013) [105]	Groundwater	Ni	PP	20.00	1.5	42.00	7.7 (7.7–7.9)	90.00	Aluminum	Aluminum	180
Shafaei et al. (2011) [106]	Manipulated	Co	PP	100.00	1.0	62.50	7.0 (2.0–8.0)	99.00	Aluminum	Aluminum	30
Mateen et al. (2020) [107]	Manipulated	Cu	D	27.80	2.0	46.40	7.0 (6.0–8.0)	95.29	Iron	Iron	5.4
Hernández et al. (2012) [108]	Groundwater	Ni	PP	41.00	0.5	16.00	7.4 (7.4–8.4)	93.90	Aluminum	Aluminum	120
Hernández et al. (2012) [108]	Groundwater	Cr	PP	23.00	0.5	16.00	7.3 (7.3–8.0)	82.60	Aluminum	Aluminum	120
Ganesan et al. (2013) [109]	Manipulated	Mn	PP	2.00	0.3	20.00	7.0 (3.0–11.0)	98.00	Stainless Steel	Magnesium	110
Gomes et al. (2010) [110]	Manipulated	Cu	PP	10.00	-	30.00	6.0 (6.0–9.1)	99.90	Iron	Iron	40
Hussain et al. (2021) [111]	Groundwater	Co	PP	0.02	1.5	1000.00	7.8 (7.8–8.4)	75.70	Aluminum	Iron	5
Hussain et al. (2021) [111]	Groundwater	Cr	PP	0.04	1.5	1000.00	7.8 (7.8–8.4)	100.00	Aluminum	Iron	5
Hussain et al. (2021) [111]	Groundwater	Ni	PP	0.05	1.5	1000.00	7.8 (7.8–8.4)	100.00	Aluminum	Iron	5
Hussain et al. (2021) [111]	Groundwater	Mn	PP	0.02	1.5	1000.00	7.8 (7.8–8.4)	100.00	Aluminum	Iron	5
Hussain et al. (2021) [111]	Groundwater	Cu	PP	0.04	1.5	1000.00	7.8 (7.8–8.4)	100.00	Aluminum	Iron	5
Shahreza et al. (2018) [112]	Manipulated	Mn	PP	360.00	2.0	10.00	9.0 (3.0–10.0)	92.00	Aluminum	Aluminum	190
Pan et al. (2017) [113]	Manipulated	Cr	PP	2.00	2.0	10.00	6.0 (6.0–9.0)	100.00	Iron	Iron	40
Vasudevan et al. (2009) [114]	Manipulated	Fe	PP	25.00	0.5	12.00	6.0 (4.0–9.0)	98.40	Galvanized Iron	Magnesium	60
Vasudevan et al. (2011) [92]	Manipulated	Cr	PP	5.00	0.5	20.00	7.0 (2.0–12.0)	99.60	Aluminum	Aluminum	240
Vasudevan et al. (2012) [115]	Shallow water	Cu	PP	10.00	0.5	2.50	7.0 (4.0–12.0)	98.50	Aluminum	Aluminum	35
Vasudevan et al. (2012) [115]	Manipulated	Co	PP	10.00	0.5	2.50	7.0 (2.0–10.0)	100.00	Galvanized Iron	Magnesium	30
Vasudevan et al. (2012) [115]	Manipulated	Cu	PP	10.00	0.5	2.50	7.0 (2.0–10.0)	99.00	Galvanized Iron	Magnesium	30
Vasudevan et al. (2012) [115]	Manipulated	Cr	PP	10.00	0.5	2.50	7.0 (2.0–10.0)	97.00	Galvanized Iron	Magnesium	30

EA: Experimental arrangement; PP: Parallel plates; D: Different arrangements; Ci: Initial concentration; RE: Removal efficiency; pHo: Optimal pH; pHi: Initial pH; pHf: Final pH; Id: Inter-electrode distance; CD: Current density; CM: Cathode material; AM: Anode material; TT: Treatment time.

. The type of water and range of pH was also added. Manipulated water used in 17 studies consisted mainly of drinking water with addition of a metal solution to obtain an initial metal concentration (Ci). For variables with only two categories, the results of the correlations give equal values but with opposite sign. For example, the R value for the variable “RE” applied to “Al CM” was 0.41 for “RE < 95%” and −0.41 for “RE ≥ 95%”. Thus, to avoid Figure 3 becoming heavy and difficult to read, it was decided to present only the result for “RE ≥ 95%”.

Analysis of the 22 works confirmed the application of EC to seven transient metals of the fourth period of the periodical table (Co, Cr, Cu, Fe, Mn, Ni and Zn) present in surface water and groundwater. The seven metals are referred to 32 times in the 22 studies, as shown in Table 3, which gives the following percentage of distribution: three references to Co removal (3/32 = 9.4%), nine references to Cr removal (9/32 = 28.1%), six references to Cu removal (6/32 = 18.8%), seven references to Fe removal (7/32 = 21.9%), three references to Mn removal (3/32 = 9.4%), three references to Ni removal (3/32 = 9.4%), and one reference to Zn removal (1/32 = 3.1%).

To estimate $R_B\left(A\right)$ (through Equation (1) to Equation (3)) and $R_{AB}$ (through Equation (6)) for each pair of categories $A$ and $B \left(\mathrm{events}\right)$ of the variables under study (Table 2), they were associated with the variables indicators $I_A$, $I_B$ and $I_{A\cap B}$ defined for each observation (the 22 works) in such way that:

$$\begin{array}{c}\hspace{1em}\hspace{1em}{I}_A=\{\begin{array}{c}1,\mathrm{i}\mathrm{f} \mathrm{c}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{g}\mathrm{o}\mathrm{r}\mathrm{y} \mathrm{A} \mathrm{o}\mathrm{c}\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{s}\\ 0, \mathrm{i}\mathrm{f} \mathrm{c}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{g}\mathrm{o}\mathrm{r}\mathrm{y} \mathrm{A} \mathrm{d}\mathrm{o}\mathrm{e}\mathrm{s} \mathrm{n}\mathrm{o}\mathrm{t}\mathrm{ }\mathrm{o}\mathrm{c}\mathrm{c}\mathrm{u}\mathrm{r}\end{array}\mathrm{and}\hfill \\ \hspace{1em}\hspace{1em}{I}_{A\cap B}=\{\begin{array}{c}1, \mathrm{i}\mathrm{f} \mathrm{c}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{g}\mathrm{o}\mathrm{r}\mathrm{i}\mathrm{e}\mathrm{s} \mathrm{A} \mathrm{a}\mathrm{n}\mathrm{d} \mathrm{B} \mathrm{o}\mathrm{c}\mathrm{c}\mathrm{u}\mathrm{r} \mathrm{t}\mathrm{o}\mathrm{g}\mathrm{e}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{r}\\ 0, \mathrm{i}\mathrm{f} \mathrm{c}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{g}\mathrm{o}\mathrm{r}\mathrm{i}\mathrm{e}\mathrm{s} \mathrm{A} \mathrm{a}\mathrm{n}\mathrm{d} \mathrm{B} \mathrm{d}\mathrm{o} \mathrm{n}\mathrm{o}\mathrm{t} \mathrm{o}\mathrm{c}\mathrm{c}\mathrm{u}\mathrm{r} \mathrm{t}\mathrm{o}\mathrm{g}\mathrm{e}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{r}\end{array}\hfill \\ \mathrm{and}{I}_B\mathrm{defined}\mathrm{similarly}\mathrm{to}{I}_B.\hfill \end{array}$$

From the indicator variables and applying Equation (1) and Equation (6) probabilities $P\left(A\right)$, $P\left(B\right)$ and $P\left(A\cap B\right)$ were estimated as $\frac{k_A}{N}$, $\frac{k_B}{N}$ and $\frac{k_{A\cap B}}{N}$, respectively, where $k_A$, $k_B$ and $k_{A\cap B}$ were defined by the sum of the indicators $I_A$, $I_B$ and $I_{A\cap B}$, respectively, representing the number of times the categories $A$, $B$ and $A\cap B$ occurred among the total of N observations (22 papers). The estimation of $R_A\left(B\right)$ was made in a similar way to $R_B\left(A\right)$.

The estimates for the results of Table 3 were computed through statistical software R version 4.1.3. with the aid of the corrplot package to graphically represent the correlation matrix between categories, which is necessary only to present the values below the lower diagonal of the correlation matrix of all the categories of interest, considering that the matrix of correlations is symmetric on the diagonal. The results are shown in Figure 4, with blue color indicating a positive relationship (i.e., an increase in category A causes an increase in category B) and red color indicating a negative relationship (i.e., an increase in category A causes a decrease in category B).

EC showed higher application at laboratory scale (96.7% of the works) using more batch flow (78.8% of the cases) and useful volume of up to 5 L (81.3% of the cases). Therefore, laboratory-based EC systems are normally used previously their scale-up to semi-industrial or real-scale systems.

Most of the technologies used parallel plates (90.6%), consisting of two or more plates in series or parallel and only three studies (9.4%) used a different plates arrangement (helical with rings [103], cylindrical and concentric electrodes [104] and bar with wrapped rings [107], namely for removing Cr and Cu.

Average initial concentration of metals (Ci) varies as follows:

• 0.02 $\le$ Co $\le$ 100 mg/L
• 0.04 $\le$ Cr $\le$ 23 mg/L
• 0.04 $\le$ Cu $\le$ 27.8 mg/L
• 10 $\le$ Fe $\le$25 mg/L
• 0.02 $\le$ Mn $\le$ 360 mg/L
• 0.05 $\le$ Ni $\le$ 41 mg/L
• Zn = 20 mg/L

Approximately 84.4% of the works found Ci above 1 mg/L (63.7% using manipulated water). The Fe and Zn minimal concentrations were 10 mg/L and 20 mg/L respectively, very high values for untreated waters, but maximum values of 360 mg/L for Mn and 100 mg/L for Co show how natural waters can be easily contaminated by anthropogenic pollution sources. In the lower category of initial concentration (Ci ≤ 1 mg/L), metals distribution was as follows: Co (16.7%) < Cr (22.2%) < Co (33.3%) < Ni (33.3%). In the intermediate category (1 < Ci ≤ 20 mg/L) the variation was Co (33.3%) < Ni (33.3%) < Mn (33.3%) < Cr (66.7%) < Cu (66.7%) < Fe (85.7%) < Zn (100.0%). In the upper category (Ci > 20 mg/L), the variation was Cr (11.1%) < Fe (14.3%) < Cu (16.7%) < Co (33.3%) < Ni (33.3%) < Mn (66.7%).

Treatment time less than 60 min was observed in 63.6% of the studies, ranging from 30 s (continuous flow) to 240 min (batch flow), with the average time being 64 min. A continuous flow device presented the longest TT. As observed by Martin-Dominguez et al. (2018) [104], systems that use EC in continuous flow usually have shorter TT, since they analyze only the passage between the electrodes, compared to batch flow. Among the 22 analyzed works, batch flow systems were used in most of the studies (78.8%) followed by continuous flow systems (21.2%), probably because they are easier to assemble and to apply. According to Holt et al. (2002) [75], continuous flow systems become cheaper when they are scaled-up to treat larger volumes of water. Thus, metal RE (A) was found to be dependent on TT, and for times less than 60 min (B). The regression coefficients were R_AB = 0.49 R_B(A) = 0.58, R_A(B) = 0.41, fitting a medium dependence, although this was the highest value found for a relationship between the categories for RE ≥ 95%. For batch reactors, the longer the TT, the greater the removal of metals, corroborating the studies of Hashim et al. (2017) [116] who, after a multiple regression analysis, concluded that TT and CD were the key variables for controlling the dissociation of material from the electrodes, when at least one of the variables was incremented.

The most common electrode materials used in reactors are aluminum and iron because they are very effective in removing metals, easy to find and have lower acquisition cost [67,69,70], which was corroborated by the present research, with aluminum being used in at least one of the electrodes in 62.5% of cases and iron in 34.4% of the studies. The preference for aluminum can be explained by the generation of the hydrolyzed species with Al(III), which is more effective in destabilizing particles than the species generated by iron electrodes with the Fe(II) species. In addition, aluminum follows the dissociation estimate of Faraday’s law, which makes employability more predictable than for iron due to the solubility of the species [117]. Iron electrodes are more used for the removal of hexavalent chromium [67], as also observed in the analyzed works where 55% of the studies used this material for removing Cr.

The analysis on the relationships between electrode material types reveals that cathodes (A) are well correlated with medium intensity with anodes (B) when iron is used, and the regression coefficient indicates that the dependence of cathode on anode is less than anode on cathode (R_AB = 0.59, R_B(A) = 0.45, R_A(B) = 0.78). When observing the anodes of the process, which is where EC reactions effectively occur, it is noted that for aluminum (A) electrodes present a medium association with TT ≥ 60 min (B) (R_AB = 0.50, R_B(A) = 0.49, R_A(B) = 0.51), and moderate association to iron (B) (R_AB = 0.41, R_B(A) = 0.34, R_A(B) = 0.49), RE < 95% (B) (R_AB = 0.41, R_B(A) = 0.34, R_A(B) = 0.50), arrangement of parallel plates (B) (R_AB = 0.30, R_B(A) = 0.17, R_A(B) = 0.52) and Ci between 1 and 20 mg/L (B) (R_AB = 0.21, R_B(A) = 0.20, R_A(B) = 0.22). For the iron anode (A), medium intensity associations were observed for the categories Id ≥ 1 cm (B) (R_AB = 0.60, R_B(A) = 0.62, R_A(B) = 0.58), TT < 60 min (B) (R_AB = 0.60, R_B(A) = 0.62, R_A(B) = 0.58), and Ci ≤ 1 mg/L (B) (R_AB = 0.60, R_B(A) = 0.45, R_A(B) = 0.78), while moderate intensity associations were found for CD ≥ 10 A/m² (B) (R_AB = 0.31, R_B(A) = 0.23, R_A(B) = 0.40), metal Cr (B) (R_AB = 0.28, R_B(A) = 0.26, R_A(B) = 0.30), RE ≥ 95% (B) (R_AB = 0.22, R_B(A) = 0.19, R_A(B) = 0.26), and different experimental arrangement (B) (R_AB = 0.22, R_B(A) = 0.36, R_A(B) = 0.13).

Regarding the electrode materials, approximately 59% of the studies used aluminum in both CM and AM and of these 46.2% to evaluate Fe removal. In all cases Fe was removed with RE ≥ 95%. Ni, Mn and Zn have been removed with RE < 95%. Electrodes with CM and AM in iron were used in 22.7% of the studies and a good relationship was found for RE ≥ 95% for Cr and Cu. Figure 4 shows a positive statistical correlation between the application of aluminum electrodes and RE ≥ 95% for Fe metal. For iron electrodes the correlation is statistically higher for Cr metal removal with RE ≥ 95% (80% of cases).

Experiments on EC normally use a supporting electrolyte to the medium for increasing electrolyte conductivity, which leads to anodic dissolution of the electrodes. Ferreira et al. (2013) applied electrodes of aluminum and iron to a low conductivity type manipulated for removing Cu, Ni and Zn. All metals were removed with RE ≥ 95% for TT > 60 min.

The use of other types of electrodes was also found in smaller quantities, namely: carbon steel (1), stainless steel (1), magnesium (3) and galvanized iron (2), with RE above 97%, and especially used for the removal of Cr. For anodes of other types of material (A) there were observed relationships of medium intensity for Id < 1 cm (B) (R_AB = 0.58, R_B(A) = 0.73, R_A(B) = 0.46) and CD < 10 A/m² (R_AB = 0.45, R_B(A) = 0.42, R_A(B) = 0.49), and moderate intensity associations for initial neutral pH (B) (R_AB = 0.37, R_B(A) = 0.44, R_A(B) = 0.31), RE ≥ 95% (B) (R_AB = 0.25, R_B(A) = 0.27, R_A(B) = 0.24) and Ci between 1 and 20 mg/L (B) (R_AB = 0.21, R_B(A) = 0.26, R_A(B) = 0.17).

Electrodes with Id ≥ 1 cm occurred in 60.6% of the studies. All metals, excepted Zn, presented RE between 92% and 100% when electrodes with different materials were used. In studies with CM and AM in aluminum and Id > 1 cm, RE was below 90%. A good correlation was observed between the smallest spacings (A) and greater RE, which can be measured at (R_AB = 0.13, R_B(A) = 0.11, R_A(B) = 0.15) (weak intensity) with RE ≥ 95% (B), mainly due to the strength of the electric field generated between the plates, in addition to having less space for coagulant dispersion [115,118]. However, Mansoorian et al. (2014) [119] and Nguyen et al., 2014 [120] found another optimal inter-electrode distance, between 1.0 to 3.0 cm.

Therefore, the spacing between electrodes (Id) allows different possibilities of reactor configuration, and, at full scale, the distance must be as large as possible if it does not compromise the efficiency of metal removal or cause instability in the electromagnetic field. About 62.5% of the analyzed works showed Id ≥ 1 cm (A) for the continuous treatment flows (which are closer to a treatment system for large flows), being also associated with the application of larger CD (B) of mean intensity (R_A(B) = 0.71) when analyzing the dependence of density on spacing. For the association between Id and CD, the relationship is of moderate intensity (R_B(A) = 0.38). The association occurs positively between the highest Id and the highest applied CD, which is explained by the need to maintain a good electromagnetic field between the electrodes, essential for a good removal of metals, but having a higher energy cost.

According to Lu et al. (2015) [121], the direct electric current applied between the electrodes is regulated by the electric CD (i.e., a regulator of the reaction rate of the anodes and cathodes), and its increase is proportional to the increase in metal removal. In about 80% of the studies that had RE ≥ 95%, a CD ≥ 10 A/m² was applied, thus, the greater removal of metals is associated with the greater density of applied current. This circumstance can lead to a high electrical consumption and high operational costs and, therefore, the DC must be optimized to obtain good metal removal, fulfilling the potability standards, with the lowest possible energy cost, which can be more expensive than the gain in removal efficiency. On the other hand, high CD can lead to the solubility of the electrode material and its destruction [117]. In approximately 15.6% of the studies, CD < 10 A/m² (A) was applied, with moderate relationships between CD and RE ≥ 95% (B) (R_AB = 0.22, R_B(A) = 0.26, R_A(B) = 0.20).

Metal RE was greater than 75% in all studies and only 24.3% was below 95%. In almost half of the studies (45.4%) the RE was above 99%, especially for Fe, which showed RE ≥ 95% in all studies. The results of the 22 studies show that EC is efficient for removing transition metals from water, with concentrations above 1 mg/L, and even from public supply water, with concentrations below the potability limits shown in Table 1 (i.e., trace concentrations). In the study by Hussain et al. (2021) [111], trace concentrations of Cr (0.04 mg/L), Cu (0.04 mg/L), Ni (0.05 mg/L) and Mn (0.02 mg/L) were completely removed with aluminum (CD) and iron electrodes (AM), for 1.5 cm Id and 1000 A/m² CD.

Therefore, RE ≥ 95% (A), in addition to presenting a mean correlation with TT < 60 min (R_AB = 0.49, R_B(A) = 0.58, R_A(B) = 0.41), presented a moderate correlation with neutral pH (B) (R_AB = 0.36, R_B(A) = 0.40, R_A(B) = 0.32), acidic pH (B) (R_AB = 0.31, R_B(A) = 0.32, R_A(B) = 0.29), Fe metal (B) (R_AB = 0.28, R_B(A) = 0.28, R_A(B) = 0.28), electrodes of other types of materials, with the cathode (B) (R_AB = 0.28, R_B(A) = 0.28, R_A(B) = 0.28) and the anode (B) (R_AB = 0.25, R_B(A) = 0.24, R_A(B) = 0.27), CD < 10 A/m² (B) (R_AB = 0.23, R_B(A) = 0.20, RA(B) = 0.26) and also iron electrodes ((R_AB = 0.23, R_B(A) = 0.20, R_A(B) = 0.26) for cathode and (R_AB = 0.22, R_B(A) = 0.26, R_A(B) = 0.19) for the anode).

Regarding the initial pH, the following distribution of acidity was observed: 43.8% (basic), 31.2% (neutral) and 25.0% (acid), but always with values close to neutral (see Table 3). Where the pH was acidic, the median values were close to 6.0, with basification of the aqueous medium during the treatment (especially with aluminum anode since it tends to neutralize the pH). In cases of basic pH, the median was 7.7 and the hydrolyzed species, depending on the electrode material, were not of low coagulant power [93]. The ER for Cr, Cu and Fe are high for initial pH ranging from 2 [102] to 9 [112], indicating that this parameter did not have a high correlation with the ER of metals. However, Vepsäläinem and Sillanpää (2020) [122] indicate that the initial pH plays an important role in the removal of metals, since it determines the hydrolyzed species formed by the metals in the electrodes, with the coagulation performance in EC being very similar to that of conventional chemical coagulation.

Approximately 22.7% of the studies (five studies) used natural groundwater without pH readjustment for the treatment (in natura), with the initial pH always close to neutrality, although slightly basic, and the application of various materials for the electrodes. Among the electrodes made with the same material, only those made of iron(CM)+ iron(AM) and carbon(CM)+ carbon(AM) showed removal greater than 99% and for the metal Cr. The association of aluminum(CM)+ iron(AM) was excellent for the removal of Cr, Cu, Co, Mn and Ni, having been completely removed.

The pH appears with good correlation with the moderate or average CD, which electrode materials, whatever they are, and especially related to the spacing of the electrodes and the current density applied in the system.

The studies were mostly developed at laboratory scale, and, therefore, more studies on the use of EC technologies at real scale is needed. Nevertheless, it can be observed that there are many correlations between the categories of the variables studied with different degrees of correlation, demonstrating that there is not a very determining factor in the performance of EC technologies, but the association of variables such as TT, CD, pH and electrode material, as well as the type of flow.

4. Conclusions

This work has found 22 research studies on the application of EC for removing seven transition metals of the fourth period (22  ≤  Z  ≤  30) from water and the most common metals were Cr, Cu and Fe. Parallel plate devices, with current density greater than 10 A/m², and with spacing between aluminum electrodes greater than 1 cm, were the most used. The studies showed removal efficiencies greater than 75% for all metals, and in the majority a removal above 95% was observed. Removal efficiencies presented several intensity relationships with the different categories, with stronger values for current density, pH, electrode material (especially for aluminum in both cathode and anode) and treatment time.