Enhanced Heavy Metal Removal from Acid Mine Drainage Wastewater Using Double-Oxidized Multiwalled Carbon Nanotubes

Due to the unique properties of carbon nanotubes (CNTs), they have attracted great research attention as an emergent technology in many applications including water and wastewater treatment. However, raw CNTs have few functional groups, which limits their use in heavy metal removal. Nevertheless, their removal properties can be improved by oxidation processes that modify its surface. In this study, we assessed the capacity of oxidized and double-oxidized multiwalled carbon nanotubes (MWCNTs) to remove heavy metals ions from acidic solutions. The MWCNTs were tested for copper (Cu), manganese (Mn), and zinc (Zn) removal, which showed an increment of 79%, 78%, and 48%, respectively, with double-oxidized MWCNTs compared to oxidized MWCNTs. Moreover, the increase in pH improved the sorption capacity for all the tested metals, which indicates that the sorption potential is strongly dependent on the pH. The kinetic adsorption process for three metals can be described well with a pseudo-second-order kinetic model. Additionally, in multimetallic waters, the sorption capacity decreases due to the competition between metals, and it was more evident in the removal of Zn, while Cu was less affected. Besides, XPS analysis showed an increase in oxygen-containing groups on the MWCNTs surface after oxidation. Finally, these analyses showed that the chemical interactions between heavy metals and oxygen-containing groups are the main removal mechanism. Overall, these results contribute to a better understanding of the potential use of CNTs for water treatment.


Introduction
Water scarcity has become a major problem worldwide. Nowadays, population growth, urbanization, pollution, and climate change are some of the main factors responsible for water scarcity [1][2][3]. In particular, Chile has a variable climate and, consequently, a diverse industrial activity along it [4], i.e., mining activity is very common in the central and northern regions of Chile. The mining industry consumes around 7.2% of all water used in Chile and represents more than 50% of the water used in northern regions [4][5][6]. Besides, mining activity results in widespread pollution related to acid generation and heavy metal release known as acid mine drainage (AMD), which adversely affects the quality of surface and ground water resources [7,8]. Thus, new alternatives to improve the water availability are becoming increasingly necessary.
AMD is characterized by a low pH and a high concentration of heavy metals, such as iron (Fe), copper (Cu), manganese (Mn), and zinc (Zn), among others [7,8]. Commonly, treatment systems the reflux temperature. In general, it has been reported that the average length of CNTs decreases significantly as the oxidation time and temperature increases [44][45][46]. However, other studies have shown that the CNTs structure does not change considerably when oxidation times are between 0.5 and 7 days [47]. Thus, the greatest differences are observed within the first hours of oxidation. The chemical functionalities added to the surface of the MWCNTs by oxidation processes create electrostatic stability that allows a colloidal dispersion [43]. This dispersion improves the surface contact between the MWCNTs and the surrounding medium, allowing more interactions between the functional groups and the ions present in the solution.
Given the current context of water scarcity and the growing demand for water and wastewater treatment technologies, the objective of this work was to explore the technological application of oxidized MWCNTs for the remediation of acidic waters contaminated with high concentrations of heavy metals. For this, modified MWCNTs were studied for the removal of Cu, Mn, and Zn. Successive oxidations were tested in order to reduce the pH PZC and increase the removal efficiency at low pH. The adsorption capabilities were evaluated at different pH values, and the competitive adsorption of three heavy metal ions on oxidized and double-oxidized MWCNTs was studied. The focus of this study was on exploring the use of CNTs in complex waters that emulate real AMD waters. The relevance of this work lies in providing a better understanding of MWCNTs and their adsorption properties in multimetallic acidic solutions as an alternative for the removal of heavy metals from AMD runoffs. It is necessary to understand the metals adsorption/desorption mechanisms at low pH values for the development of applicable technologies for a real treatment system.

Infrared (IR) Spectroscopy
The infrared (IR) spectroscopy analysis was carried out to study the functional groups present on MWCNTs after oxidation processes. Figure 1 shows the IR spectrum of raw, oxidized, and double-oxidized MWCNTs. The three samples show the same three major signals: 3435 cm −1 , 1635 cm −1 , and 1050 cm −1 . The peak of 3435 cm −1 indicates the O-H bond and is more intense in raw than in oxidized MWCNTs. Raw MWCNTs show an intense peak in 1635 cm −1 corresponding to the C=C bond in benzene rings [48]. It can be inferred that according to the intensity of the signal, raw MWCNTs have a higher amount of C=C bonds, and when MWCNTs were oxidized, these bonds were broken, resulting in the formation of new bonds [49]. The peak around 1050 cm −1 corresponds to C-O bond related to carboxyl groups introduced in the oxidation processes. The peaks near to 2300 cm −1 are present in all three samples and correspond to carbon dioxide, which is a typical signal of this analysis [48]. Appendix B shows the IR spectrum for samples of oxidized and double-oxidized MWCNTs after adsorption processes, peaks corresponding to O-H and C=C functional groups can be observed.

Brunauer-Emmet-Teller Analysis (BET)
The BET analysis gives information on the surface properties and the sorption capacity for different types of MWCNTs. The BET surface area, pore volume, pore diameter, maximum adsorption capacity, and point of zero charge are summarized in Table 1. The BET surface area increased from 157.34 to 179.53 m 2 /g after the first oxidation step, but it decreased during the second oxidation showing an S BET value of 105.54 m 2 /g. The decrease in the BET surface area in acid treated CNTs has been previously reported. This occurs mainly because the defective sites of the CNTs are occupied by the functional groups incorporated during the oxidation process [50,51]. Even so, these values are high compared to other values reported using different oxidation methods [30,52], which suggests a high sorption potential. The same behavior observed for S BET values was observed for the maximum adsorption capacity, which increased from 36.15 to 41.25 cm 3 /g and then reduced to 24.25 cm 3 /g with double oxidation. However, it should be noted that the BET analysis gives information about the material potential, but in practice, the sorption effectiveness will depend on the amount and type of functional groups that are present in MWCNTs, as well as on the dispersion that they can reach in solution.
Molecules 2020, 25, x FOR PEER REVIEW  4 of 22 about the material potential, but in practice, the sorption effectiveness will depend on the amount and type of functional groups that are present in MWCNTs, as well as on the dispersion that they can reach in solution.  In addition, the point of zero charge of a surface is an essential factor in determining the metal adsorption effectiveness, because waters with a high metallic content generally have a low pH. Thus, the surface charge of MWCNTs can be estimated by pHPZC values. Untreated nanotubes have a pHPZC value of about 4.65, which means that for pH values above the pHPZC, the MWCNTs surface will be negatively charged, and therefore will have an affinity/potential for cations removal. Inversely, below this value, the surface will be positively charged, and the interaction with metals will be minimal [33,34]. Furthermore, the pHPZC of oxidized MWCNTs is of about 3.70, and the value for doubleoxidized MWCNTs is even lower. These pHPZC values represent a decrease compared to raw MWCNTs. This suggests that the pH range at which oxidized and double-oxidized MWCNTs can remove metals is bigger compared to untreated MWCNTs. Successive oxidation steps partially opens the MWCNTs tips, which enhances the dispersion of MWCNTs and generates structural defects due to the attachment of oxygen-containing groups [53]. The presence of these strongly acidic groups can be evidenced by the great and sustained decrease in the pHPZC values with successive oxidations [53,54]. Therefore, the decrease in pHPZC values from 3.7 to 1.2 with double oxidation can promote a best interaction and removal of metals from acidic waters.

Scanning Electron Microscopy (SEM)
The morphology of raw and oxidized MWCNTs was investigated with scanning electron mcroscopy coupled with energy-dispersive X-ray spectroscopy (SEM-EDX), and field emission scanning electron microscopy (FESEM). SEM/FESEM images are presented in Figure 2. It is possible  In addition, the point of zero charge of a surface is an essential factor in determining the metal adsorption effectiveness, because waters with a high metallic content generally have a low pH. Thus, the surface charge of MWCNTs can be estimated by pH PZC values. Untreated nanotubes have a pH PZC value of about 4.65, which means that for pH values above the pH PZC , the MWCNTs surface will be negatively charged, and therefore will have an affinity/potential for cations removal. Inversely, below this value, the surface will be positively charged, and the interaction with metals will be minimal [33,34]. Furthermore, the pH PZC of oxidized MWCNTs is of about 3.70, and the value for double-oxidized MWCNTs is even lower. These pH PZC values represent a decrease compared to raw MWCNTs. This suggests that the pH range at which oxidized and double-oxidized MWCNTs can remove metals is bigger compared to untreated MWCNTs. Successive oxidation steps partially opens the MWCNTs tips, which enhances the dispersion of MWCNTs and generates structural defects due to the attachment of oxygen-containing groups [53]. The presence of these strongly acidic groups can be evidenced by the great and sustained decrease in the pH PZC values with successive oxidations [53,54]. Therefore, the decrease in pH PZC values from 3.7 to 1.2 with double oxidation can promote a best interaction and removal of metals from acidic waters.

Scanning Electron Microscopy (SEM)
The morphology of raw and oxidized MWCNTs was investigated with scanning electron mcroscopy coupled with energy-dispersive X-ray spectroscopy (SEM-EDX), and field emission scanning electron microscopy (FESEM). SEM/FESEM images are presented in Figure 2. It is possible to observe Molecules 2020, 25, 111 5 of 22 morphological differences between the raw and oxidized MWCNTs. The raw MWCNTs have a homogeneous surface. After the oxidations processes, the nanotube surface is heterogeneous, and, with the double oxidation, the filaments looked thinner and were divided into smaller segments. Also, it is possible to observe an increase in the oxygen weight percentage (wt %) after the oxidation processes, which may be explained by the addition of new functional groups (e.g., hydroxyl and carbonyl) ( Figure 2). Furthermore, the average diameter of the MWCNTs decreases with the oxidation processes, from a diameter of 31.65 nm for untreated MWCNTs to 29.21 nm and 27.64 nm for oxidized MWCNTs and double-oxidized MWCNTs, respectively. The reduction in diameter with successive oxidations indicates the removal of impurities such as metals and amorphous carbon, and sidewall MWCNTs damage resulting from oxidation [43].
In consequence, the MWCNTs characterization analysis showed that the main differences derived from the oxidation treatment involve the formation of new bonds due to the union of functional groups to the surface. These functional groups contain oxygen, which was confirmed by an increment in the content of this element. Moreover, the surface changed considerably due to the defects introduced during the oxidation, which could improve the adsorption potential of MWCNTs.
Molecules 2020, 25, x FOR PEER REVIEW 5 of 22 to observe morphological differences between the raw and oxidized MWCNTs. The raw MWCNTs have a homogeneous surface. After the oxidations processes, the nanotube surface is heterogeneous, and, with the double oxidation, the filaments looked thinner and were divided into smaller segments. Also, it is possible to observe an increase in the oxygen weight percentage (wt %) after the oxidation processes, which may be explained by the addition of new functional groups (e.g., hydroxyl and carbonyl) ( Figure 2). Furthermore, the average diameter of the MWCNTs decreases with the oxidation processes, from a diameter of 31.65 nm for untreated MWCNTs to 29.21 nm and 27.64 nm for oxidized MWCNTs and double-oxidized MWCNTs, respectively. The reduction in diameter with successive oxidations indicates the removal of impurities such as metals and amorphous carbon, and sidewall MWCNTs damage resulting from oxidation [43].
In consequence, the MWCNTs characterization analysis showed that the main differences derived from the oxidation treatment involve the formation of new bonds due to the union of functional groups to the surface. These functional groups contain oxygen, which was confirmed by an increment in the content of this element. Moreover, the surface changed considerably due to the defects introduced during the oxidation, which could improve the adsorption potential of MWCNTs.

Adsorption Experiments
The adsorption experiments were performed with acidic-metallic solutions with varying concentrations of divalent metals (Cu 2+ , Mn 2+ , and Zn 2+ ). The experiments are shown only for the oxidized and double-oxidized MWCNTs since raw MWCNTs showed a negligible removal capacity (data not shown). It has already been reported that raw MWCNTs have a negligible sorption capacity. Some studies present sorption values that do not exceed 1.6 mg/g [22,23].
Experimental data were fitted with Langmuir and Freundlich isotherm models. It can be seen that the Langmuir model has a better agreement with the observations compared to the Freundlich isotherm. Furthermore, we should recall that the Langmuir isotherm defines the equilibrium parameters of homogeneous surfaces and monolayer adsorption [55], while the Freundlich model describes a multilayer adsorption onto a heterogeneous surface [56]. Figure 3 shows that the sorption capacity significantly increased with the double oxidation (double-oxidized MWCNTs) for the three metals studied, compared with a single oxidation (oxidized MWCNTs). In the case of Cu 2+ , the maximum sorption capacity increased from 7.8 to 14 mg/g. These results imply an increment of 79% in the total removal within 14 h of contact time. For Mn 2+ and Zn 2+ , the maximum sorption capacity increased from 3.7 to 6.6 mg/g and from 2.7 to 4.0 mg/g, which correspond to an increment of 78% and 48%, respectively. For all three metals, the best removal capacity was achieved with the double-oxidized MWCNTs, which confirms the improvement in the removal potential due to the double oxidation. Similar studies show variable sorption capacities, depending on the type of treatment used in the CNTs and the specific experimental conditions. Form previous studies, sorption capacities between 3.5-29 mg/g for Cu 2+ , 0.2-2.0 mg/g for Mn 2+ , and 0.27-58 mg/g Zn 2+ have been reported for MWCNTs [15,19,20,30,[57][58][59][60]. This strongly supports our statement that the oxidation process improves the sorption capacity of CNTs.

Adsorption Experiments
The adsorption experiments were performed with acidic-metallic solutions with varying concentrations of divalent metals (Cu 2+ , Mn 2+ , and Zn 2+ ). The experiments are shown only for the oxidized and double-oxidized MWCNTs since raw MWCNTs showed a negligible removal capacity (data not shown). It has already been reported that raw MWCNTs have a negligible sorption capacity. Some studies present sorption values that do not exceed 1.6 mg/g [22,23].
Experimental data were fitted with Langmuir and Freundlich isotherm models. It can be seen that the Langmuir model has a better agreement with the observations compared to the Freundlich isotherm. Furthermore, we should recall that the Langmuir isotherm defines the equilibrium parameters of homogeneous surfaces and monolayer adsorption [55], while the Freundlich model describes a multilayer adsorption onto a heterogeneous surface [56]. Figure 3 shows that the sorption capacity significantly increased with the double oxidation (double-oxidized MWCNTs) for the three metals studied, compared with a single oxidation (oxidized MWCNTs). In the case of Cu 2+ , the maximum sorption capacity increased from 7.8 to 14 mg/g. These results imply an increment of 79% in the total removal within 14 h of contact time. For Mn 2+ and Zn 2+ , the maximum sorption capacity increased from 3.7 to 6.6 mg/g and from 2.7 to 4.0 mg/g, which correspond to an increment of 78% and 48%, respectively. For all three metals, the best removal capacity was achieved with the double-oxidized MWCNTs, which confirms the improvement in the removal potential due to the double oxidation. Similar studies show variable sorption capacities, depending on the type of treatment used in the CNTs and the specific experimental conditions. Form previous studies, sorption capacities between 3.5-29 mg/g for Cu 2+ , 0.2-2.0 mg/g for Mn 2+ , and 0.27-58 mg/g Zn 2+ have been reported for MWCNTs [15,19,20,30,[57][58][59][60]. This strongly supports our statement that the oxidation process improves the sorption capacity of CNTs. The major difference in the sorption capacities occurs at high metal concentrations, while at low concentrations, the difference is less significant. It has been reported that heavy metals are absorbed at specific sorption sites for low metal concentrations [61]. Furthermore, at low metal concentration, The major difference in the sorption capacities occurs at high metal concentrations, while at low concentrations, the difference is less significant. It has been reported that heavy metals are absorbed at specific sorption sites for low metal concentrations [61]. Furthermore, at low metal concentration, the competition for binding sites is lower [61]. Therefore, at low metal concentrations, the removal capacity should be similar for different nanotubes (oxidized and double-oxidized MWCNTs), which was consistent with our results. At higher metal concentrations, it is possible to observe the differences in the affinity of MWCNTs with the different metals studied. The good affinity of MWCNTs with Cu 2+ , compared to other metals, has been reported on several studies [35,62].
SEM images ( Figure 4) confirm the hypothesis of metal stabilization on the surface of MWCNTs. Similarly to what we observed in the adsorption isotherms, the metal retention is greater for double-oxidized MWCNTs than for oxidized MWCNTs. Although the SEM-EDS analysis is not quantitative, we observed an increase in the metals weight percentage as determined by EDX analysis in double-oxidized MWCNTs compared with oxidized MWCNTs. The identified spectrum showed that there were 0.37% of Cu, 0.09% of Mn and 0.12% of Zn atoms on the surface of the double-oxidized MWCNTs, while on the surface of oxidized MWCNTs a lower weight percentage was detected with values of 0.62%, 1.16%, and 0.2% for Cu, Mn, and Zn, respectively. Although the examined area was very small compared to the total sample, we observed that the metals distribution on the surface was homogeneous. These results reinforce the idea that double-oxidized MWCNTs are better sorbents for divalent metals from acidic waters. the competition for binding sites is lower [61]. Therefore, at low metal concentrations, the removal capacity should be similar for different nanotubes (oxidized and double-oxidized MWCNTs), which was consistent with our results. At higher metal concentrations, it is possible to observe the differences in the affinity of MWCNTs with the different metals studied. The good affinity of MWCNTs with Cu 2+ , compared to other metals, has been reported on several studies [35,62]. SEM images (Figure 4) confirm the hypothesis of metal stabilization on the surface of MWCNTs. Similarly to what we observed in the adsorption isotherms, the metal retention is greater for doubleoxidized MWCNTs than for oxidized MWCNTs. Although the SEM-EDS analysis is not quantitative, we observed an increase in the metals weight percentage as determined by EDX analysis in doubleoxidized MWCNTs compared with oxidized MWCNTs. The identified spectrum showed that there were 0.37% of Cu, 0.09% of Mn and 0.12% of Zn atoms on the surface of the double-oxidized MWCNTs, while on the surface of oxidized MWCNTs a lower weight percentage was detected with values of 0.62%, 1.16%, and 0.2% for Cu, Mn, and Zn, respectively. Although the examined area was very small compared to the total sample, we observed that the metals distribution on the surface was homogeneous. These results reinforce the idea that double-oxidized MWCNTs are better sorbents for divalent metals from acidic waters. The metal sorption capacity has also been studied for other types of adsorbents. Recently, Soltani et al. [63] measured sorption capacities of about 7.30 mg/g for Cu using clay as the adsorbent, and of about 22 mg/g using polyaniline-clay. It should be noted that these concentrations are much higher than the ones studied in this work (between 50 and 1000 mg/L). Moreover, other adsorbents have been used successfully for the removal of various metals, such as the graphene oxide-maize amylopectin adsorbent that reached a sorption capacity of 30.56 mg/g for Cu and of 7.92 mg/g for Mn [64]. In addition, a new imprinted magnetic biosorbent reached a sorption capacity of 69.34 mg/g for lead [28].

Competitive Adsorption
The sorption experiments were carried out for the same metal concentration and pH value, using a monometallic and a multimetallic solution. The removal rates for each case are illustrated in Figure  5. In order to quantify the effect of competition between different metals in acidic waters, we The metal sorption capacity has also been studied for other types of adsorbents. Recently, Soltani et al. [63] measured sorption capacities of about 7.30 mg/g for Cu using clay as the adsorbent, and of about 22 mg/g using polyaniline-clay. It should be noted that these concentrations are much higher than the ones studied in this work (between 50 and 1000 mg/L). Moreover, other adsorbents have been used successfully for the removal of various metals, such as the graphene oxide-maize amylopectin adsorbent that reached a sorption capacity of 30.56 mg/g for Cu and of 7.92 mg/g for Mn [64]. In addition, a new imprinted magnetic biosorbent reached a sorption capacity of 69.34 mg/g for lead [28].

Competitive Adsorption
The sorption experiments were carried out for the same metal concentration and pH value, using a monometallic and a multimetallic solution. The removal rates for each case are illustrated in Figure 5. In order to quantify the effect of competition between different metals in acidic waters, we calculated the removal coefficient (Kr). A metal with higher Kr implies a lower effect of competition on that metal in the presence of other metals. According to the results, the calculated values of Kr were 0.58 (Cu 2+ ), 0.51 (Mn 2+ ), and 0.03 (Zn 2+ ) for oxidized MWCNTs and 0.66 (Cu 2+ ), 0.21 (Mn 2+ ), and 0.13 (Zn 2+ ) for double-oxidized MWCNTs. Thus, Kr values for the studied metals in decreasing order are Cu > Mn > Zn, which means that the metal that is least affected by the presence of other metals is Cu, while the most affected by competition is Zn. Previously, Salam et al. [65] showed similar results in which MWCNTs had a better removal of Cu than Zn.  It can be seen that the effect of competition affects metals proportionally to their initial concentration in the solution (Cu: 20 mg/L, Mn: 6 mg/L, Zn: 3 mg/L). Therefore, the dominant factor for the preference between the different ions corresponds to the competition for the active sites of the adsorbent and the availability of ions in the aqueous solution [18]. This has been already demonstrated in several studies, where also the pH plays an important role in the ion's competition. It has been reported that the increase in pH can improve the competition of some metals [18,19,65], i.e., Stafiej and Pyrzynska [19] showed that at pH above 6, Cu 2+ greatly improves its competition against Mn 2+ . Hence, these results support the advantage of the double oxidation in enhancing the removal potential of the MWCNTs.

pH Effect on Adsorption Rate
Sorption assays under different pH values were performed to evaluate the role of the solution's pH in the adsorption rate. The pH values studied range from 3.0 to 7.0. The results show a direct relationship between the pH of the solution and the metal removal ( Figure 6). In general, the sorption capacity increases with increasing pH. At very acidic pHs (3.0-4.0), removal rates were minimal for all three metals. This can be explained because, for pH values much lower than pHPZC, the surface of the MWCNTs will be more positive, and hence it will reduce its ability to interact with divalent cations.
For Cu 2+ ions, the removal rate increased from 18% to 40% with oxidized MWCNTs and from 27% to 58% with double-oxidized MWCNTs. The sorption rates increased steadily as the pH increased. These results are in accordance with what has been observed for Cu 2+ removal with functionalized MWCNTs. Specifically, Zhang et al. [66] achieved an elimination rate close to 72% by increasing the pH up to 12. Likewise, the same effect was observed by Stafiej and Pyrzynska [19]. They found that at pH 9 the affinity order of the metal ions using CNTs as adsorbents was Cu(II) > Pb(II) > Co(II) > Zn(II) > Mn(II) [19].
In the case of Mn 2+ , the assays with oxidized MWCNTs show that sorption rates increased from 27% to 40% at a pH value around 5.2 and then it decreased to 37% at circumneutral pH. With doubleoxidized MWCNTs, the removal rates increased from 27% to 47% at a pH value of around 5.5 and then decreased to 37% at a pH of 6.5. Both types of MWCNTs had a similar behavior for different pH It can be seen that the effect of competition affects metals proportionally to their initial concentration in the solution (Cu: 20 mg/L, Mn: 6 mg/L, Zn: 3 mg/L). Therefore, the dominant factor for the preference between the different ions corresponds to the competition for the active sites of the adsorbent and the availability of ions in the aqueous solution [18]. This has been already demonstrated in several studies, where also the pH plays an important role in the ion's competition. It has been reported that the increase in pH can improve the competition of some metals [18,19,65], i.e., Stafiej and Pyrzynska [19] showed that at pH above 6, Cu 2+ greatly improves its competition against Mn 2+ . Hence, these results support the advantage of the double oxidation in enhancing the removal potential of the MWCNTs.

pH Effect on Adsorption Rate
Sorption assays under different pH values were performed to evaluate the role of the solution's pH in the adsorption rate. The pH values studied range from 3.0 to 7.0. The results show a direct relationship between the pH of the solution and the metal removal ( Figure 6). In general, the sorption capacity increases with increasing pH. At very acidic pHs (3.0-4.0), removal rates were minimal for all three metals. This can be explained because, for pH values much lower than pH PZC , the surface of the MWCNTs will be more positive, and hence it will reduce its ability to interact with divalent cations.
For Cu 2+ ions, the removal rate increased from 18% to 40% with oxidized MWCNTs and from 27% to 58% with double-oxidized MWCNTs. The sorption rates increased steadily as the pH increased. These results are in accordance with what has been observed for Cu 2+ removal with functionalized MWCNTs. Specifically, Zhang et al. [66] achieved an elimination rate close to 72% by increasing the pH up to 12. Likewise, the same effect was observed by Stafiej and Pyrzynska [19]. They found that at pH 9 the affinity order of the metal ions using CNTs as adsorbents was Cu(II) > Pb(II) > Co(II) > Zn(II) > Mn(II) [19].
In the case of Mn 2+ , the assays with oxidized MWCNTs show that sorption rates increased from 27% to 40% at a pH value around 5.2 and then it decreased to 37% at circumneutral pH. With double-oxidized MWCNTs, the removal rates increased from 27% to 47% at a pH value of around 5.5 and then decreased to 37% at a pH of 6.5. Both types of MWCNTs had a similar behavior for different pH values. The decrease in Mn removal for high values of pH has already been reported [67]. When the pH decreases, the surface of the MWCNTs becomes more positive due to the deposition of hydrogen ions on the surface. In contrast, when the pH increases, the surface becomes more negative, that is, it has a greater affinity to remove cations [33,67]. This behavior would explain what happens with the removal of Mn 2+ up to pH~5.5. From this value, a decrease in Mn removal is observed with the increase in pH, which could be related to the Mn speciation as a function of pH. Using PHREEQC (acronym of pH-Redox-Equilibrium-C + language program) speciation models (Appendix A, Table A1), we observed that at high pH (>6.0) the formation of manganese hydroxides (Mn(OH) 3 − ) occurs, which can affect the removal process of Mn 2+ because this anion is not removed by negatively charged surfaces.
values. The decrease in Mn removal for high values of pH has already been reported [67]. When the pH decreases, the surface of the MWCNTs becomes more positive due to the deposition of hydrogen ions on the surface. In contrast, when the pH increases, the surface becomes more negative, that is, it has a greater affinity to remove cations [33,67]. This behavior would explain what happens with the removal of Mn 2+ up to pH ~5.5. From this value, a decrease in Mn removal is observed with the increase in pH, which could be related to the Mn speciation as a function of pH. Using PHREEQC (acronym of pH-Redox-Equilibrium-C + language program) speciation models (Appendix A, Table  A1), we observed that at high pH (>6.0) the formation of manganese hydroxides (Mn(OH)3 − ) occurs, which can affect the removal process of Mn 2+ because this anion is not removed by negatively charged surfaces. Finally, for Zn 2+ , the maximum sorption rates are reached at a pH close to 5.5, similar to Mn 2+ , reaching a removal of 60% with oxidized MWCNTs and of 65% with double-oxidized MWCNTs. Similarly, other studies have reported that the Zn 2+ removal by CNTs becomes constant at pH~6.0, and then it decreases as the pH increases [68]. This behavior can be also explained by the increase in the formation of Zn hydroxides (Zn(OH)3 − ) as a function of pH (Appendix A, Table A1). Finally, for Zn 2+ , the maximum sorption rates are reached at a pH close to 5.5, similar to Mn 2+ , reaching a removal of 60% with oxidized MWCNTs and of 65% with double-oxidized MWCNTs. Similarly, other studies have reported that the Zn 2+ removal by CNTs becomes constant at pH~6.0, and then it decreases as the pH increases [68]. This behavior can be also explained by the increase in the formation of Zn hydroxides (Zn(OH) 3 − ) as a function of pH (Appendix A, Table A1).

Kinetics Experiments
In order to study the adsorption kinetics of the three metals by oxidized and double-oxidized MWCNTs, kinetics experiments were performed. The experimental kinetic curves for oxidized and double-oxidized MWCNTs showed that the sorption of Cu 2+ , Mn 2+ , and Zn 2+ rapidly increased within the first 10 min of contact time. The sorption continued to increase over time until it reached an equilibrium at approximately 2 h of contact time (Figure 7). Similar results have been reported for the adsorption of metals by different CNTs. In these cases the equilibrium was reached approximately between the first 30 and 120 min [69,70]. Li et al. [41] investigated the removal of Cu 2+ and observed that the adsorption rate increased rapidly during the first 15 min and reached its equilibrium point at approximately 60 min.

Kinetics Experiments
In order to study the adsorption kinetics of the three metals by oxidized and double-oxidized MWCNTs, kinetics experiments were performed. The experimental kinetic curves for oxidized and double-oxidized MWCNTs showed that the sorption of Cu 2+ , Mn 2+ , and Zn 2+ rapidly increased within the first 10 min of contact time. The sorption continued to increase over time until it reached an equilibrium at approximately 2 h of contact time (Figure 7). Similar results have been reported for the adsorption of metals by different CNTs. In these cases the equilibrium was reached approximately between the first 30 and 120 min [69,70]. Li et al. [41] investigated the removal of Cu 2+ and observed that the adsorption rate increased rapidly during the first 15 min and reached its equilibrium point at approximately 60 min.  We calculated the parameters of the model from the experimental data to describe the kinetic process. The parameters and the regression correlation coefficients (R 2 ) are presented in Table 2. Both the theoretical equilibrium sorption capacity (q e ) model and the values of the correlation coefficient show that the pseudo-second-order model is better suited to reproduce the experimental data for the three metals studied (Cu 2+ , Mn 2+ , and Zn 2+ ). These results support the idea that the sorption process is mainly controlled by the adsorption reaction at the liquid/solid interface at the adsorbent, and not by a diffusion-controlled process as the pseudo-first-order model would indicate [71].

Desorption Experiments
The desorption of heavy metals was studied as a function of pH. The desorption process is important because the recycling and regeneration of the adsorbent are necessary to reduce operating costs [62]. On the other hand, this information also reports about the stability of adsorption in the treatment of acidic waters. Figure 8 shows the Cu 2+ , Mn 2+ , and Zn 2+ recovered from oxidized and double-oxidized MWCNTs at pH values between 1 and 5.
Molecules 2020, 25, x FOR PEER REVIEW 11 of 22 We calculated the parameters of the model from the experimental data to describe the kinetic process. The parameters and the regression correlation coefficients (R 2 ) are presented in Table 2. Both the theoretical equilibrium sorption capacity (q e ) model and the values of the correlation coefficient show that the pseudo-second-order model is better suited to reproduce the experimental data for the three metals studied (Cu 2+ , Mn 2+ , and Zn 2+ ). These results support the idea that the sorption process is mainly controlled by the adsorption reaction at the liquid/solid interface at the adsorbent, and not by a diffusion-controlled process as the pseudo-first-order model would indicate [71].

Desorption Experiments
The desorption of heavy metals was studied as a function of pH. The desorption process is important because the recycling and regeneration of the adsorbent are necessary to reduce operating costs [62]. On the other hand, this information also reports about the stability of adsorption in the treatment of acidic waters. Figure 8 shows the Cu 2+ , Mn 2+ , and Zn 2+ recovered from oxidized and double-oxidized MWCNTs at pH values between 1 and 5.  It can be observed that for the three metals, the percentage of desorption at the same pHs was lower for double-oxidized MWCNTs than for oxidized ones. These results indicate that double-oxidized MWCNTs generate more stable chemical interactions at low pH, so they have a better sorption capacity for acidic waters, as it is also shown in Figure 6. However, both types of MWCNTs show almost a total desorption for a pH values below 2. Other studies with oxidized CNTs have shown similar behavior even with different metals. Liang et al. [72] used MWCNTs for the removal of Mn and obtained a desorption of~80% at pH 4 and of 100% at pH 3. Moreover, Lu et al. [73] used SWCNTs and MWCNTs for the adsorption of Zn. In both cases, they obtained a 60% of desorption at a pH 3 and more than 90% at pH 1 [73]. For other metals, a desorption greater than 80% has been observed at higher pHs than those used obtained in this study, i.e., pH 4.5 for titanium and pH 3 for Pb 2+ [70,[74][75][76].

X-ray Photoelectron Spectrometry (XPS) Analysis
The XPS analysis was used to quantify the surface elemental composition of MWCNTs and to understand the adsorption mechanism. Figure 9 summarizes the results of this analysis. Figure 9a shows the full spectrum for MWCNTs samples. The abundance percentages for the primary XPS region of carbon (C1s) and oxygen (O1s) are 95.18% and 4.82% for raw MWCNTs, 86.01% and 13.99% for oxidized MWCNTs, and 65.87 and 34.13% for double-oxidized MWCNTs, respectively. According to literature data, oxidized CNTs have oxygen percentages in the order of 8-12% in relation to carbon content [77][78][79]. On the other hand, the oxygen abundance in raw CNTs usually does not exceed 3% [77,78,80], which is consistent with the results obtained in this study. Thus, these results show a significant increase in oxygen content due to the formation of oxygen-containing functional groups. Other studies on the effect of oxygen-containing groups on the performance of adsorbents have been performed. A significant increase in the sorption capacity of biochar [81] and graphene oxides [82,83] was reported, with the increase in the content of oxygen-containing groups. Figure 9b shows the deconvoluted spectra of C1s. The analysis shows that about 40% of carbon content corresponds to chemical bonds of functional groups that contain oxygen. On the other hand, Figure 9c, and d show the deconvolution of the spectra of O1s and Cu2p, respectively. The O1s peak can be divided into two parts. The first peak, around 531 eV corresponding to O 2− , which interacts with Cu 2+ to form chemical bonds, and another peak around 533 eV, attributed to surface absorbed O 0 or oxygen vacancy (V 0 ) [84][85][86]. The Cu spectrum shows a peak at 934.8 eV, representing the copper hydroxide (Cu(OH) 2 ) [87]. This area represents about 50.82% of the total abundance. In addition, satellite peaks are observed, which according to the literature, is related to Cu 2+ /CuO [84,87]. Finally, a smaller peak corresponding to Cu2p 1/2 is observed at 944.1 eV, which is also attributed to Cu 2+ /CuO [84,85]. Therefore, Cu can be found as superficial Cu 2+ , which may explain a physical removal of copper attached to the surface of the MWCNTs. However, Cu can be observed mostly as copper oxide/hydroxide, which demonstrates that the main removal mechanism corresponds to the chemical interaction between Cu 2+ and the functional groups of oxidized MWCNTs. The Mn spectrum (Figure 9e) shows two major peaks, corresponding to Mn2p 1/2 (~652 eV) and Mn2p 3/2 (~643 eV). These peaks and their positions are typical of manganese oxides. In particular, the shape of the peak Mn2p 3/2 and the absence of satellite zones around 647 eV, indicate that manganese is mostly found as manganese(III) oxide (Mn 2 O 3 ) [88][89][90]. Likewise, the Zn spectrum (Figure 9f) also has two main peaks, corresponding to Zn2p 1/2 (~1046 eV) and Zn2p 3/2 (~1023 eV) that indicate the presence of zinc oxide (ZnO) [91,92]. Therefore, Cu, Mn, and Zn are forming oxides, which indicates that the main removal mechanism is the chemical interaction between metal ions and oxygen-containing groups, forming ionic bonds. This type of chemical bond between metallic and non-metallic atoms makes us conjecture that the main mechanism for removal is ionic exchange. This deduction was also obtained by some studies that investigated adsorption in CNTs [18,19,53]. The XPS analysis was complemented with desorption experiments as the pH value affects the speciation of the metals [93,94]. In the case of Cu, copper hydroxide is the dominant species for pH values greater than 7, and it decreases with the reduction of the pH value [95]. Finally, for pH values below ~2.5, Cu is fully dissolved as Cu 2+ , which is consistent with the desorption results shown in Figure 8 [95,96]. Furthermore, Mn and Zn are predominantly found as divalent species Mn 2+ and Zn 2+ at pH below 9 and 8, respectively [97].

MWCNT Oxidation
The MWCNTs, commercially acquired from NanoTechLabs, Inc. (Yadkinville, NC, USA) (purity > 95%, avg. length = 100 µm), were oxidized to introduce defects in their structure and increasing the number of functional groups present in it. The XPS analysis was complemented with desorption experiments as the pH value affects the speciation of the metals [93,94]. In the case of Cu, copper hydroxide is the dominant species for pH values greater than 7, and it decreases with the reduction of the pH value [95]. Finally, for pH values below~2.5, Cu is fully dissolved as Cu 2+ , which is consistent with the desorption results shown in Figure 8 [95,96]. Furthermore, Mn and Zn are predominantly found as divalent species Mn 2+ and Zn 2+ at pH below 9 and 8, respectively [97].

MWCNT Oxidation
The MWCNTs, commercially acquired from NanoTechLabs, Inc. (Yadkinville, NC, USA) (purity > 95%, avg. length = 100 µm), were oxidized to introduce defects in their structure and increasing the number of functional groups present in it.

Oxidized MWCNTs
For this carboxylation method, 500 mg of MWCNTs were oxidized with 250 mL of 65% nitric acid (HNO 3 ) by refluxing at 120 • C for 4 h. Then, the solution was filtered using a 0.45 µm membrane filter. Subsequently, the MWCNTs were washed with deionized water until reaching a neutral pH. The obtained solid was dried at 60 • C for 12 h.

Double-Oxidized MWCNTs
This method consists of two consecutive oxidations. First, MWCNTs are oxidized with 65% HNO 3 as described above. Subsequently, a second oxidation was carried out using a proportion of nitric acid and sulfuric acid (H 2 SO 4 ) [23]. For this, 500 mg of oxidized MWCNTs were treated with a (v/v 1:3) mixture of 65% HNO 3 and 97% H 2 SO 4 . The mixture was sonicated at 40-50 • C for 3 h in an ultrasonic bath (Isolab Laborgerate GmbH, Wertheim, Germany). Then, the resulting solution was filtered using a 0.45 µm membrane filter and was washed with deionized water until the pH was neutral. The obtained solid was dried at 60 • C for 12 h.
The schematic diagram of both oxidation processes is presented in Figure 10.
Molecules 2020, 25, x FOR PEER REVIEW 14 of 22 For this carboxylation method, 500 mg of MWCNTs were oxidized with 250 mL of 65% nitric acid (HNO3) by refluxing at 120 °C for 4 h. Then, the solution was filtered using a 0.45 µm membrane filter. Subsequently, the MWCNTs were washed with deionized water until reaching a neutral pH. The obtained solid was dried at 60 °C for 12 h.

Double-Oxidized MWCNTs
This method consists of two consecutive oxidations. First, MWCNTs are oxidized with 65% HNO3 as described above. Subsequently, a second oxidation was carried out using a proportion of nitric acid and sulfuric acid (H2SO4) [23]. For this, 500 mg of oxidized MWCNTs were treated with a (v/v 1:3) mixture of 65% HNO3 and 97% H2SO4. The mixture was sonicated at 40-50 °C for 3 h in an ultrasonic bath (Isolab Laborgerate GmbH, Wertheim, Germany). Then, the resulting solution was filtered using a 0.45 µm membrane filter and was washed with deionized water until the pH was neutral. The obtained solid was dried at 60 °C for 12 h.
The schematic diagram of both oxidation processes is presented in Figure 10.

Point of Zero Charge (PZC)
To measure the PZC, we prepared 0.01 M solutions in 10 mL tubes at different pH values (2 to 10) adjusted with NaOH 0.1 M and HCl 0.1 M [33]. Then, 10 mg of raw, oxidized and double-oxidized MWCNTs were added to these solutions and shaken at 350 rpm at room temperature for 24 h. The final pH was measured using a pH meter (PHC301, HACH). Initial pH was plotted against final pH, and the intersection of this curve with the line that represents pH initial = pH final, determine the PZC value.

Point of Zero Charge (PZC)
To measure the PZC, we prepared 0.01 M solutions in 10 mL tubes at different pH values (2 to 10) adjusted with NaOH 0.1 M and HCl 0.1 M [33]. Then, 10 mg of raw, oxidized and double-oxidized MWCNTs were added to these solutions and shaken at 350 rpm at room temperature for 24 h. The final pH was measured using a pH meter (PHC301, HACH). Initial pH was plotted against final pH, and the intersection of this curve with the line that represents pH initial = pH final, determine the PZC value.

Adsorption Experiments
Adsorption experiments were performed for raw, oxidized, and double-oxidized MWCNTs. For this, 20 mg of MWCNTs were mixed with 30 mL of metal ion solution. The concentration was increased for each metal: Cu (5 to 25 mg/L), Mn (2 to 10 mg/L), and Zn (1 to 5 mg/L). The solution with MWCNTs was shaken for 14 h at 350 rpm. The resulting metal concentration was measured colorimetrically in a Hach spectrophotometer DR/2010. The experiment was performed in triplicate to obtain the average and standard deviation of the data.

Adsorption Isotherms
Experimental results were compared with Langmuir and Freundlich isothermal models. The sorption capacity q (mg/g sorbent) was obtained using the equation where:

Langmuir Model
The sorption capacity q (mg/g sorbent) was obtained with the equation [98] q = q L K L C e 1 + K L Ce (2) where: q L : amount of adsorption corresponding to a monolayer coverage. -K L : Langmuir constant related to the energy of adsorption. -C e : equilibrium concentration of metal in aqueous solution (mg/L).

Freundlich Model
The sorption capacity q (mg/g sorbent) was determined as [98]: where: -K F and n are Freundlich constants related to adsorption capacity and adsorption intensity, respectively. -C e : equilibrium concentration of metal in aqueous solution (mg/L).

Competitive Adsorption Experiments with Synthetic Acid Mine Drainage Waters
To study the interference between metals, adsorption experiments, as described in Section 2.4, were carried out using a synthetic acid mine drainage water. The synthetic water was designed, based on samples from typical rivers impacted by AMD in the central-northern zone of Chile. The synthetic water was prepared with distilled water, and it contained (per liter): 20 mg of Cu +2 (added as CuSO 4 ·5H 2 O); 6 mg of Mn +2 (added as MnSO 4 ·H 2 O); and 3 mg of Zn +2 (added as ZnSO 4 ·7H 2 O).
The pH was adjusted to 4.2 with 0.5 M HCl. Also, the removal coefficient (K r ) was calculated to determine the metal sorption changes for monometallic and multimetallic solutions. K r = % Removal in multimetallic water % Removal in monometallic water (4) The experiment was performed in triplicate to obtain the average and standard deviation of the data.

Effect of pH
Adsorption experiments were repeated at fixed pH values according to the PZC results. For oxidized and double-oxidized MWCNTs, assays were carried out at pH 3, 4, 5.5, and 7 approximately. The experiment was performed in triplicate to calculate a mean and standard deviation for the data.

Kinetic Studies
The kinetic studies were carried out using 20 mg of MWCNTs and 30 mL of metal solution. The metal concentration for the assays was the same used in interference experiments. The samples were taken at 10 min, 30 min, 1 h, 2 h, 4 h, and 24 h. The removal rate was represented as a function of time. The kinetic process was investigated according to a pseudo-first-order and pseudo-second-order kinetic equations. The pseudo-first-order kinetic model is presented by Lagergen as [99,100]: log q e −q t = logq e − k 1 2.303 t (5) where q e and q t are the adsorption capacity in mg/g at equilibrium and at time t, respectively, and k 1 is the constant of first-order adsorption in min −1 . The pseudo-second-order rate equation is expressed as [100]: t q t = 1 k 2 ·q 2 e + t q e (6) where k 2 is the constant of second-order adsorption in g mg −1 min −1 .

Desorption Experiments
Adsorption experiments were carried out for oxidized and double-oxidized MWCNTs using monometallic solutions of Cu, Mn, and Zn. Assays were conducted by shaking 20 mg of MWCNTs with 30 mL of solution at room temperature until reach equilibrium. Then, the solution was filtered, and the filtrate was transferred to 30 mL of deionized water adjusted at different pH values (1 to 5) and was shaken for 14 h. The percentage of desorption was calculated using the equation

X-ray Photoelectron Spectrometry (XPS) Analysis
The chemical composition of the MWCNTs surface, as well as the interaction with metals, was further characterized by X-ray photoelectron spectrometry (XPS, SPECS FlexMod, Berlin, Germany) equipped with PHOIBOS 150 1D-DLD analyzer and with monochromatic source Focus 500. Casa XPS software was used for detailed data processing.

Conclusions
In summary, our results showed that double-oxidized MWCNTs have a high capacity to remove metals from acidic waters. The oxidation treatment was useful to incorporate functional groups onto the MWCNTs surface. The chemical interactions between the functional groups and Cu 2+ , Mn 2+ , and Zn 2+ were evidenced from the XPS analysis. This interaction may be the main removal mechanism. Indeed, the Cu 2+ removal was more favored due its higher concentration in the acidic solution (simulated) compared with Mn 2+ and Zn 2+ .
Double-oxidized MWCNTs were more efficient than oxidized MWCNTs in the removal of Cu 2+ , Mn 2+ , and Zn 2+ . SEM-EDX analyzes support this idea, because the metals weight percentage was higher in double-oxidized MWCNTs samples. Also, the metals were removed more efficiently at circumneutral values of pH, but Cu 2+ tightly stuck to double-oxidized MWCNTs with a less desorption percentage compared to Mn 2+ and Zn 2+ . Furthermore, kinetic experiments suggest that the pseudo-second-order model is a better representation of the Cu 2+ , Mn 2+ , and Zn 2+ adsorption compared to the pseudo-first-order model.
In general, double-oxidized MWCNTs showed better performance in all cases, which supports the idea of its use as an alternative sorbent for metals from acidic waters. These findings contribute to a better understanding of the potential of MWCNTs for water treatment.
Author Contributions: The manuscript was written by C.R., but all the authors contributed to its preparation and review. Experiments were performed by C.R.; Data analyses were carried out by C.R. in discussion with E.L. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by PIA-UC Eduardo Leiva/DDA-VRA, FIC-R Fondo de Innovación para la Competitividad Gore Coquimbo BIP nos. 30485965-0 and CORFO-L2 L2 ISV93456. also extended to the reviewers for their corrections and suggestions, who contributed significantly to improving the quality of this manuscript. Finally, we also thank to Camila Tapia and Maria Estela Cayazzo for their analytical assistance in IR spectroscopy analysis and other measurements.

Conflicts of Interest:
The authors declare no conflict of interest. A   Table A1. Molality of manganese and zinc hydroxides as function of pH. The speciation data corresponds to a simulation with PHREEQC using experimental concentrations of the analysis.