Regeneration of As(V) loaded granular activated carbon through desorption in FeCl₃, CaCl₂ and MgCl₂ aqueous solutions

GAC

Granular activated carbon

PFO

Pseudo-first order

PSO

Pseudo-second order

Arsenic is widely distributed throughout the Earth's environment (Mandal & Suzuki 2002). Exposure to arsenic causes a wide array of health problems including respiratory (Milton et al. 2001), hematological (Armstrong et al. 1984) and carcinogenic (Banerjee et al. 2011). Many papers report on removal of arsenic from water and soil using a variety of methods such as filtration (Waypa et al. 1997), electrocoagulation (Nidheesh & Anantha Singh 2017), precipitation and adsorption (Gupta & Chen 1978).

Adsorption is the adhesion of an adsorbate onto the surface of an adsorbent. Common adsorbents for arsenic adsorption are iron hydroxide (Pierce & Moore 1982), aluminium hydroxide (Anderson et al. 1976) and activated carbon (Lorenzen et al. 1995), but other materials can be used, such as diatomite (Davoodi et al. 2019) and walnut shell (Ashrafi et al. 2018). The surface of activated carbon is typically porous, leading to an extraordinarily large specific surface area and pore volume. The properties of the pores and the functional groups on the surface depend on the preparation of the AC and determine its affinity to certain adsorbents. Over time the adsorption goes down as the adsorbent surface becomes saturated, at which point it is disposed of or regenerated (San Miguel et al. 2001; Ania et al. 2005). This regeneration can take place thermally, which is straightforward but expensive to operate and causes loss of adsorbent (Martin & Ng 1985; Salvador & Jiménez 1996; Hong et al. 2020). Nonthermal methods include the use of solvents, surfactants and acids or base. This last method is based on a change in pH, with the final pH being unfavorable to adsorption (McLaughlin 1995). One literature review indicated that research on desorption of cadmium and lead typically focus on regeneration through use of acids such as nitric acid, EDTA and hydrochloric acid (Fouda-Mbanga et al. 2021). Chemical regeneration is a versatile option, where desorption is improved through a change in chemical polarity, hydophobicity, solubility, molecular weight, pH, boiling point or toxicity (Larasati et al. 2021). It is important to consider the environmental fate of the desorbed contaminant, ensuring continued removal from waste streams. This means that an adsorption-desorption cycle should involve either a concentration of the contaminant or its separation. One way that separation can be achieved is through precipitation.

Precipitation is a common way to remove arsenic from aqueous waste streams. This is commonly achieved with iron (Halder et al. 2018), aluminium (Pantuzzo et al. 2014), calcium (Bothe & Brown 1999a; Palfy et al. 1999) and magnesium (Park et al. 2010). The solubility of precipitates is highly dependent on the solution pH, the species of both arsenic and the metal used for precipitation. In previous work arsenate removal from aqueous solution was tested through coagulation with FeCl₂, FeCl₃, AlCl₃, CaCl₂ and MgCl₂ (Moed & Ku 2022). For desorption and precipitation to simultaneously occur, it is important that favorable pH for these processes overlap. In this research, desorption was found to be optimal at an initial pH of 11, followed by a pH of 3. This excludes the use of FeCl₂, with optimal precipitation taking place at pH 8.5 (Johnston & Singer 2007) and AlCl₃, at pH 6 (Majzlan et al. 2018). Reported optimum precipitation of FeCl₃ varies from paper to paper but is generally shown to occur in the range of pH 2–5 (Nishimura & Tozawa 1978; Robins 1987). Precipitation with CaCl₂ increases with pH (Zhang et al. 2015), being optimal at pH over 12 (Bothe & Brown 1999a), whereas precipitation with Mg(II) was found to be optimal between pH 7.5 and 10.2 (Park et al. 2010).

The scope of this research is to investigate the use of metals to precipitate As(V) as it is desorbed through a shift in pH. The precipitation reduces the arsenic in the solution, which may lead to an increased desorption. The arsenic is simultaneously removed from the solution. Kinetic study will be performed for both adsorption and desorption of As(V) by and from the activated carbon used.

Granular activated carbon (GAC) of 8 × 30 mesh (First Chemical Works in Taiwan) was washed with 0.3% HCl solution for 2 h at a GAC concentration of 100 g/L. It was then rinsed repeatedly with dH₂O, filtered through Advantec Grade No. 2 filter paper and dried overnight at 110 °C.

Na₂HAsO₄•7H₂O (Panreac, Spain), NaCl (Fischer, UK), FeCl₃•6H₂O (Acros Organics, USA), CaCl₂•2H₂O (Showa Chemical, Japan) and anhydrous MgCl₂ (Showa Chemical, Japan) were dissolved in dH₂O to create separate stock solutions. HCl and NaOH solutions at varying concentrations were used for pH adjustment. Experiments were, unless otherwise mentioned, performed inside a thermostatic shaking bath (DK-60 by Deng Yng, Taiwan) set to a shaking speed of 150 rpm and a temperature of 25 °C. All modeling was performed nonlinearly using the application Origin Pro.

The surface characteristics were tested by Brunauer, Emmett and Teller method (BET) (BELSORP-max by BEL Japan), Field Emission Scanning Electron Microscopy with Energy Dispersive Spectrometry (FE-SEM with EDS) (JSM-7900F by JEOL) and X-ray Fluorescence (XRF) (Epsilon 1 by Malvern Panalytical).

Adsorption of As(V) on the fresh GAC used in this study was previously tested and found to be optimal in a solution with pH 6 (not reported). Different concentrations (50, 100, 150 and 200 mg/L) of As(V) were tested at a GAC concentration of 20 g/L, with samples being taken at set times to be filtered and analyzed by ICP-OES (iCAP 7000 by Thermo Scientific). The adsorption isotherm was tested at different concentrations of As(V) at pH 6 in plastic tubes of 50 mL volume.

GAC was loaded with As(V) aqueous solution at a concentration of 200 mg/L and at pH 6. After a 20-h adsorption the GAC was filtered and dried overnight at 110 °C. Desorption of As(V) from the loaded GAC was tested in dH₂O at varying initial pH, where the pH was adjusted with HCl and NaOH solution. Desorption took place induplicate in 250 mL glass laboratory flasks. Samples were taken at set intervals, filtered and analyzed by ICP-OES.

The influence of pH was tested on the regeneration of As(V) loaded GAC. The selected desorption solutions were FeCl₃, CaCl₂ and MgCl₂ at a molar concentration of Fe(III)/Ca(II)/Mg(II) equal to 1.5 times that of the As(V) loaded onto the GAC. The concentration of As(V) is calculated as the adsorption capacity after adsorption (4.00 mg/g) multiplied by the concentration of GAC in the regeneration (20 g/l) for a total of 80 mg/L. Desorption was tested in FeCl₂ at initial pH (pHi) of 2.0, 2.5 and 3.0 and in CaCl₂ and MgCl₂, separately, at pHi 11.0, 11.5 and 12.0. NaCl was added as a comparison at pHi 2.0, 3.0, 11.0 and 12.0. Desorption took place overnight after which the GAC was filtered and dried at 110 °C. Subsequent adsorption in 200 mg/L As(V) solution of pH 6 was performed overnight to determine regeneration efficiency. The desorption was performed in triplicate, to ensure accuracy, whereas subsequent adsorption was performed in duplicate, in either case using 50 mL plastic tubes. Samples of adsorption and desorption were taken which were diluted in HNO₃, filtered and measured by ICP-OES. The pH of each sample was measured after both desorption and adsorption, to test the influence of the GAC surface acidity on the desorption and adsorption media.

The influence of molar ratio was tested on the regeneration of As(V) loaded GAC was performed by adding GAC at a concentration of 20 g/L in aqueous solution of either Fe(III), Ca(II) or Mg(II) at molar ratios of 0.75:1, 1.5:1, 3:1, 6:1 and 12:1 of Fe(III)/Ca(II)/Mg(II):As(V). NaCl was found to be least effective in the previous test and therefore omitted from further tests. Due to a combination of factors (the As(V) desorbed, regeneration efficiency and perceived As(V) precipitation) the optimum pH_i were chosen to be 3 for regeneration in FeCl₃ and 11 in.CaCl₂ and MgCl₂. Desorption was performed over 4 h, as this was found to be sufficient to reach maximum desorption, in duplicate in 50 mL plastic tubes. Two types of samples were taken, with the first diluted in HNO₃ and filtered to test the total As(V) desorption. The other type of samples was filtered prior to dilution in H₂O to test As(V) precipitation. The GAC was filtered and dried at 110 °C to test subsequent adsorption, and both desorption and subsequent adsorption were performed in 50 mL plastic tubes.

Kinetic desorption studies of As(V) from loaded GAC were performed at the highest ratios of 3:1, 6:1 and 12:1 of metal to arsenic. These kinetic tests were performed in 250 mL glass laboratory flasks at 25 °C, with samples taken at timed intervals. The data was modeled using the PFO and PSO kinetic models.

The results suggest that the PSO is a more accurate fit, as confirmed in Table 1 which reveals that, at any initial As(V) concentration, the PSO is most accurate in describing the adsorption of As(V) on GAC. This is revealed from the sum of the squared residuals which are, in all cases, lower for the PSO model than for the PFO model. This is also backed by the higher adjusted R² (over 0.96 for PSO and 0.94 for PFO) and from the comparison of the equilibrium adsorption capacity q_e found experimentally to those determined by the models. In their publication of 1999, Ho and McKay concluded that for all data analyzed and previously described using other kinetic models, the PSO was more accurate. The following popularity of the model has only increased over time as the PSO has been favored in the majority of kinetic modeling performed since (Ho 2006; Sun et al. 2013; Hassan et al. 2014; Koohzad et al. 2019). It appears that this superiority is so widely accepted that some papers do not even report on the PFO kinetic model (Yao et al. 2014; Gong et al. 2015).

When desorbing in FeCl₃ solution, a higher pH_i led to a lower concentration of As(V) desorbed. The concentration of Fe(III) in the solution is also revealed in Figure 4. The pH_i shows a clear effect on concentration, which is presumed to be through precipitation. The Fe(III) concentration decreases from 54.3 mg/L at pH_i 2 to 46.7 and 38.3 mg/L at pH_i 2.5 and 3, respectively. This corresponds to the decrease in As(V) detected at these pH_i, indicating that the lower As(V) concentration is due to precipitation. The pH after desorption, as listed in Figure 5, reveals that the pH goes up slightly, likely due the GAC surface acidity. For FeCl₃, an initial pH of 2.0, 2.5 and 3.0, led to a final pH of 2.1, 2.7 and 3.3, respectively. Robins (1987) examined and compared data of four studies to model the stability of scorodite, or ferric arsenate (AsFeO₄), by pH. Their model suggested that optimum removal took place at pH 4.8. There seems to be some disparity between optimum pH among papers, with the data of Nishimura & Tozawa (1978) showing this pH to be at 2. Ferric iron and arsenic exist in a variety of compounds such as angelellite (Fe₄(AsO₄)₂(OH)₃) and kamarizaite (Fe₃(AsO₄)₂(OH)₃). Majzlan et al. (2018) tested the precipitation of these compounds and scorodite and found that, while scorodite solubility was lowest in the range of pH 3.5 ∼ 6.5, for angellelite and kamarizaite the optimum was narrower, being close to 6.5. This seems to be in line with the data found in this paper, with the final pH of 3.3 leading to a lower detection of As(V) and Fe(III), even when dissolving in 10% HNO₃ solution. Subsequent adsorption of As(V) loaded GAC is highest following desorption in FeCl₃, according to Figure 4, and is highest at a pH_i of 2, followed by 3. The average adsorption is lowest at pH_i 2.5, which may be due to a limited overlap of desorption, which favors a pH removed from that of optimum adsorption pH, and precipitation, which was highest at pH_i 3. Considering the goal is regeneration of GAC while simultaneously removing As(V) from the solution this pH_i of 3 is considered optimal.

For desorption of As(V) in CaCl₂ aqueous solution Ca(II) concentration goes down with pH_i while As(V) concentration increases. This can be explained through the work of Bothe & Brown (1999b) who analyzed the phases of calcium arsenates at increasing pH. It was found that, as pH increases, the ratio of As(V) to Ca(II) in the predominant phase gradually decreases from 0.8 to 0.5. The relatively low ratio in this research this likely results in a lower fraction of bound As(V). Figure 5 lists the final pH after desorption in CaCl₂, where a pH_i of 11.0, 11.5 and 12.0 led to a final pH of 6.4, 7.5 and 11.0, respectively. When testing precipitation of various forms of calcium arsenates, Bothe & Brown (1999a) found that, at a pH of 11.2, the arsenic concentration had dropped to 3 mg/L and a pH of 12 further reduced this less than 0.5 mg/L. Zhang et al. (2015) similarly found that within their tested pH range (3–10) arsenic content in the precipitate increased with pH, where calcium transformed from gypsum (CaSO₄) to calcium arsenate (Ca₃(AsO₄)₂). This indicates that precipitation in this experiment likely takes place at the higher pH_i of 12, but is too low to be optimal. Subsequent adsorption seems to be optimal at pH_i 11.5, followed by pH_i 11, which lead to a final pH of 7.5 and 6.4, respectively. As(V) desorption, however, is slightly lower at pH_i 11. A pH_i of 12 leads to both the highest detected As(V) desorption and the lowest subsequent adsorption.

For desorption in Mg(II) solution, pH_i within this range does not have a major effect on Mg(II) concentration, while As(V) concentration increases with pH_i. A pH_i of 11.0, 11.5 and 12.0 led to a final pH of 7.9, 8.6 and 11.2, respectively. Park et al. (2010) found that for precipitation of As(V) with Mg, the pH for removal was optimal between 7.5 and 10.2, where arsenic existed in the form of Mg₃(AsO₄)₂. A higher pH will lead to precipitation of magnesium to Mg(OH)₂, which leads to a lower removal of arsenic. Li et al. (2019) tested precipitation of arsenic with magnesium in the presence of NH₄ and determined the optimum pH to be 9.5. The samples with pH_i 11.0 and 11.5 both fall within a favorable pH range, with 11.5 ending up closest to the optimum found by Li et al. Regeneration in MgCl₂ was slightly higher at pH_i 12, measuring by subsequent As(V) adsorption. The increase in adsorption is limited, however, compared to the As(V) desorbed during regeneration. Considering that a pH_i of 11 led to an only slightly lower regeneration efficiency, a better precipitation and requires less base addition it was deemed optimal.

where Ce is the equilibrium concentration of As(V), the variable a is the multiplier of the ratio, Ratio is the molar ratio of Fe(III), Ca(II) and Mg(II) to As(V) and b is the concentration of As(V) at a ratio of 1. This model was fitted both for the complete set of ratios (labeled Log. All) and for the ratios from 0.75:1 to 6:1 (labeled Log. Excl.), due to discrepancy a a ratio of 12:1. It is observed in Figure 6 that, for desorption in Fe(III) aqueous solution, the fitting excluding the ratio of 12:1 showed a better fit. This is confirmed by the adjusted R² increasing from 0.90282 to 0.99992. For desorption in Ca(II) solution, a logarithmic fit is more accurate when using all ratios, with an R² of 0.96676 as opposed to 0.94387. For desorption in Mg(II) solution, the logarithmic fit was very accurate whether the 12:1 ratio was included or not, with an R² of 0.99820 and 0.99596, respectively.

The concentration of As(V) with samples diluted in H₂O is also revealed in Figure 6. The As(V) concentration when desorbed in FeCl₃ initially increases with molar ratio from 0.75:1 to 1.5:1, then decreases with molar ratio until 6:1. At this ratio an average concentration was found of 0.15 mg/L. When further increasing the ratio to 12:1, however, the As(V) concentration rises again. A higher molar ratio generally leads to a higher As(V) desorption, but also increases the amount of iron to coprecipitate As(V) with. This explains the initial rise in As(V) concentration, as the effect of increased As(V) desorption is stronger than that on the precipitation. At higher ratios the increase in precipitation is stronger than the increase in desorption. The mechanism behind the final rise in concentration at a ratio of 12:1 is unclear, however, as no increase was observed when diluted in HNO₃. The optimum molar ratio is 6:1 for both desorption of As(V) and its removal from the solution through precipitation, with near complete removal taking place (99.77% of the As(V) detected in HNO₃). The final pH for each molar ratio was measured between 2.6 and 3.2. No relation was found between this divergence in pH and the results. When desorbing in MgCl₂ the solubility of As(V) is unaffected by ratio in the range of 0.75:1 to 6:1, remaining nearly similar to the solubility in HNO₃. When further increasing the ratio from 6:1 to 12:1, however, solubility in water decreases, with the As(V) concentration increasing by only less than 1 mg/L. In comparison, the As(V) concentration increased by nearly 10 mg/L when diluting in HNO₃ solution when the molar ratio was increased from 6:1 to 12:1. The final pH went up greatly with molar ratio, being 6.9 at a ratio of 0.75:1 and 9.9 at 12:1. Li et al. (2019) showed that a pH of 9.5 was optimal, which is likely another cause for the increased precipitation at this highest ratio. When desorbing in CaCl₂ solution the molar ratio did not affect solubility of As(V). As(V) desorption in CaCl₂ solution behaved similarly whether dilution took place in HNO₃ solution or in H₂O. The final pH was relatively stable in a range of 6.8–7.1. A logarithmic fitting was tested again which was unsuitable for desorption in Fe(III) solution, reaching an adjusted R² of 0.35636 and 0.53932 for all ratios and excluding 12:1, respectively. This is expected, as the solubility in water was greatly diminished. For desorption in Ca(II) solution the logarithmic fitting showed good accuracy of 0.97097 and 0.95613, respectively, both of which are slightly higher as compared those found for the data in Figure 6. For desorption in Mg(II) solution the logarithmic model shows a good fitting in the range of ratio 0.75:1 to 6:1, at 0.99669, but is significantly worse when including the 12:1 ratio, at 0.94130. This is in contrast to the fitting for samples diluted in HNO₃, where an R² of over 0.99 was found. This fact reinforces the previous findings that, at this highest ratio and diluting in water, maximum solubility is reached.

The subsequent adsorption of As(V) by GAC regenerated in FeCl₃, CaCl₂ and MgCl₂ solutions is also displayed in Figure 6. While there is no significant difference between the metal chlorides at a molar ratio of 0.75:1, at all higher ratios, the regeneration in FeCl₃ aqueous solution is highest. The trend for As(V) desorbed in HNO₃ is reflected in subsequent As(V) adsorption, with the As(V) adsorbed 10% higher than the As(V) desorbed (with a maximum of 5.4% deviation). Similarly, a further increase in molar ratio from 6:1 to 12:1 has limited beneficial effect on the regeneration of As(V) loaded GAC, with the adsorption increasing from 73.9 to 75.9 mg/L (an increase of ∼2.7%). This increase indicates that the decrease in As(V) concentration for desorption in FeCl₃ solution at a molar ratio of 12:1 is caused by precipitation even when diluting in 10% HNO₃ solution. Fresh GAC reached an adsorption of 80.2 mg/L, which means that regeneration in FeCl₃ aqueous solution led to a regeneration of 92.2% and 94.7% at molar ratios of 6:1 and 12:1, respectively, in comparison to adsorption of fresh GAC. From a ratio of 1.5:1 to 6:1 regeneration in MgCl₂ solution is superior to that in CaCl₂ but, at the ratios of 0.75:1 and 12:1, there is no significant difference. At Ca(II)/Mg(II):As(V) ratios of 12:1, regeneration in CaCl₂ and MgCl₂ led to a regeneration efficiency of 83.6% and 83.4%, respectively. Comparing the three metal chlorides, FeCl₃ is superior to CaCl₂ and MgCl₂ in As(V) desorption, As(V) precipitation and As(V) adsorption following regeneration. A logarithmic model was fitted to the adsorption data in Figure 6, which was comparable to the fittings for desorption of As(V) when diluted in HNO₃ solution. Regeneration in FeCl₃ solution led to an adjusted R² 0.94458 when fitting all ratios and 0.99739 with the exclusion of 12:1, both of which were higher than the R² for desorption. Adsorption following regeneration in CaCl₂ was very accurate at 0.99034 when fitting all ratios and dropped slightly to 0.97955 with the exclusion of 12:1, an increase in accuracy over the data for desorption. Adsorption following regeneration in MgCl₂ was most accurate regardless of whether or not the ratio of 12:1 was included, with an R² of 0.99090 and 0.99917, respectively.

While other literature investigating regeneration of activated carbon loaded with heavy metals through coagulation was not found, other methods have been tested. Hamdaoui et al. (2005) employed ultrasound to regenerate granular activated carbon loaded with Hg(II), Cr(VI), Cu(II) and Mn(II), reaching regeneration efficiencies of 24%, 35%, 37% and 43%, respectively. The same technique was used by Jing et al. (2011) for regeneration of Cr(VI) from powdered activated carbon which reached an efficiency of 29%. When eluting Pt and Pd from activated carbon using NaCl solution, Snyders et al. (2015) found a high desorption of 99%, but this required a temperature of 95 °C. At 60 °C, efficiencies were markedly lower at 64% and 68%, respectively. Removal of Au was significantly lower, reaching ∼32% at 95 °C and being nonexistent at 60 °C. When adsorption is exothermic, higher temperature will aid in desorption (Van Deventer & Van der Merwe 1994). This higher temperature, while increasing regeneration efficiency, leads to a significant increase in operating costs. Whereas the previously mentioned research of Di Natale et al. (2013) reached complete desorption of As and a regenerated adsorption capacity of 93%, this was at a relatively low loading of 0.1 mg/g. In comparison, the arsenic removal by electrocoagulation in this study displays high removal while taking place at room temperature and high loading. For regeneration of cadmium loaded activated carbon, made from highly ligno-cellulosic jute stick, using 0.1 mol/L HNO₃, Ghosh et al. (2021) found that 95.4% desorption of cadmium took place. This led to a subsequent adsorption of 96.3% of the original 73,53 mg/g. The highest regeneration of 94.4% (measured as subsequent adsorption) found in this research is very comparable, proving the feasibility of coagulation in the regeneration of activated carbon.

Elemental analysis was performed by XRF, as listed in Table 4. Surface arsenic is found to increase to 6.5% after loading and decrease to 1.4–1.6% after regeneration with the exception of Fe 6:1, which is slightly higher at 2.5%. This could be due to the relatively high iron content, which is commonly doped onto AC to improve arsenic adsorption (Lorenzen et al. 1995). Regeneration with FeCl₃ leads to higher concentrations than regeneration with calcium and magnesium chloride. The same is true for calcium after regeneration with CaCl₂, with the observed 2.1% the highest among samples. This is not observed for magnesium after regeneration with MgCl₂, however, with the detected 22.0% being among the lowest found between samples. This is likely due to the magnesium content on the GAC already being among the highest of elements measured, making it more difficult to additional magnesium to bind to the surface. Chlorine concentration decreased after loading with As(V) and was found to be highest after regeneration with FeCl₃ at a molar ratio of 3:1 and 12:1. After regeneration it was found to be the lowest after regeneration with MgCl₂, which corresponds to the low adsorption found for magnesium. For future experiments, the use of EDX, CHNS or XPS may provide additional insights into deposition on the GAC.

Surface area and pore diameter were tested by BET method, also listed in Table 4. It was found that surface area increases slightly through regeneration with FeCl₃ and CaCl₂ and more effectively with MgCl₂. In contrast, the pore diameter decreases slightly during the regeneration, most visibly for CaCl₂. The effect of this seems low, however, with the regeneration of GAC as displayed in Figure 7 equally effective when using CaCl₂ and MgCl₂.

SEM analysis was performed but the regeneration solution showed no influence on the observed GAC surface.

The increase in adsorption capacity of As(V) on granular activated carbon (GAC) went down when increasing the initial As(V) concentration from 150 to 200 m/L, indicating that the GAC was near completely satisfied. The pseudo-second order (PSO) was more accurate than the pseudo-first order (PFO). The Freundlich model was most accurate followed by the Temkin model and finally the Langmuir model. Desorption of As(V) loaded GAC in dH₂O was highest at pH 11 and followed by pH 3, removed from the previously found optimum pH for adsorption of 6. PSO modeling was most accurate for desorption. Desorption of As(V) from GAC was tested. Despite NaCl leading to the highest desorbed As(V) detected it displayed the lowest As(V) adsorption following regeneration. Regeneration in FeCl₃ showed the highest subsequent As(V) adsorption, with the largest precipitation taking place at pH 3.0. Regeneration in CaCl₂ and MgCl₂ was both deemed optimal at pH 11. Regeneration of GAC and precipitation of As(V) generally increased with the molar ratio of Fe(III)/Ca(II)/Mg(II) to As(V). A logarithmic relationship between molar ratio and As(V) desorption and subsequent adsorption was found. Subsequent As(V) adsorption was highest in FeCl₃ and lowest in CaCl₂. Kinetic modeling was inaccurate for desorption in FeCl₃. For desorption in CaCl₂ the PFO was a marginally better fit at the molar ratios of 3:1 and 6:1, but the PSO was more accurate at 12:1. Desorption in MgCl₂ was best described with the PSO at every ratio. In conclusion, regeneration of As(V) loaded GAC was best achieved in FeCl₃ aqueous solution at a pH of 3 and a molar ratio of 12:1 when subsequent adsorption is the only goal. However, a molar ratio of 6:1 was nearly as effective while requiring half the amount of FeCl₃ and removing As(V) more efficiently through precipitation. This precipitation, however, results in kinetic modeling becoming more difficult to achieve in comparison to that in CaCl₂ and MgCl₂.

N. M. Moed: conceptualization; data curation; formal analysis; investigation; methodology; software; validation; visualization; writing – original draft & editing; Y. Ku: funding acquisition; project administration; resources; supervision

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.