E ﬃ cient Mercury Removal at Ultralow Metal Concentrations by Cysteine Functionalized Carbon-Coated Magnetite

Featured Application: This study highlights the utility of L-cysteine, a thiol-containing amino acid, as a functionalizing reagent for enhancing the adsorption capacity of Hg(II) ions by carbon-coated magnetite sorbents. Abstract: This work reports the preparation and utility of cysteine-functionalized carbon-coated Fe 3 O 4 materials (Cys-C@Fe 3 O 4 ) as e ﬃ cient sorbents for remediation of Hg(II)-contaminated water. E ﬃ cient removal (90%) of Hg(II) from 1000 ppb aqueous solutions is possible, at very low Cys-C@Fe 3 O 4 sorbent loadings (0.01 g sorbent per liter of Hg(II) solution). At low metal concentrations (5–100 ppb Hg(II)), where adsorption is typically slow, Hg(II) removal e ﬃ ciencies of 94–99.4% were achievable, resulting in ﬁnal Hg(II) levels of < 1.0 ppb. From adsorption isotherms, the Hg(II) adsorption capacity for Cys-C@Fe 3 O 4 is 94.33 mg g − 1 , around three times that of carbon-coated Fe 3 O 4 material. The highest partition coe ﬃ cient (PC) of 2312.5 mgg − 1 µ M − 1 was achieved at the initial Hg (II) concentration of 100 ppb, while signiﬁcantly high PC values of 300 mgg − 1 µ M − 1 and above were also obtained in the ultralow concentration range ( ≤ 20 ppb). Cys-C@Fe 3 O 4 exhibits excellent selectivity for Hg(II) when tested in the presence of Pb(II), Ni(II), and Cu(II) ions, is easily separable from aqueous media by application of an external magnet, and can be regenerated for three subsequent uses without compromising Hg(II) uptake. Derived from commercially available raw materials, it is highly possible to achieve large-scale production of the functional sorbent for practical applications.

adsorption kinetics and thermodynamics, in addition to probing the selectivity of Hg(II) adsorption in the coexistence of other contaminant metal ions.

Sample Preparation
Carbon-coated (Carbon@Fe 3 O 4 ) nanoparticles were prepared using a solvothermal method reported by Zheng et al. [34]. Briefly, FeCl 3 •6H 2 O (0.8 g) was dissolved in 60 mL of ethylene glycol with stirring. After a clear solution was obtained, urea (3.0 g) and D-glucose (0.36 g) were added, and the mixture was stirred at room temperature for 30 min. Subsequently, the solution was transferred to a Teflon-lined stainless-steel autoclave and solvothermally treated at 200 • C for 12 h. Notably, previous works suggested that solvothermal temperatures of 160 • C and above are required to obtain a single phase of magnetite phase (Fe 3 O 4 ) [29,38,39], and the boiling point of ethylene glycol is elevated due to the elevated pressure in the reactor and the colligative property of solutions [40][41][42]. After cooling down to room temperature and washing sequentially with ultrapure water and ethanol (each three times), the resulting carbon-coated Fe 3 O 4 nanoparticles were magnetically separated from the solution using a neodymium magnet and dried at 60 • C in a vacuum oven (SHEL LAB, Shelton Manufacturing, Inc., Cornelius, Oregon) for 12 h. Uncoated Fe 3 O 4 nanoparticles were synthesized from a solvothermal treatment of a FeCl 3 /EG/urea mixture (without addition of D-glucose). Cysteine-functionalized (Cys-C@Fe 3 O 4 ) particles were synthesized using a modified version of the protocol previously utilized in functionalization of iron oxide materials [43,44]. In a typical synthesis, an aqueous suspension of C@Fe 3 O 4 nanoparticles (1.0 g in 100 mL) was ultrasonically treated for 30 min, and then mixed with an aqueous solution of L-cysteine (250-1000 mg dissolved in 100 mL). The pH of the suspension was adjusted to 5.8 using 0.1 M HCl or 0.1 M NaOH, followed by stirring for 12 h under an N 2 atmosphere at room temperature. The solid products were then separated using an external magnet, and cleaned by washing sequentially with water and ethanol (each three times). The particles were then vacuum dried at 60 • C for 12 h and were denoted as either 250Cys-C@Fe 3 O 4 , 500Cys-C@Fe 3 O 4 , or 1000Cys-C@Fe 3 O 4 , with the number reflecting the weight of L-cysteine added in the preparation. Scheme S1 (Supporting Information) summarizes sample preparation methods used in this study.

Sample Characterization
Powder X-ray diffraction (PXRD) measurements were conducted on a Bruker powder X-ray diffractometer (AXS model D8 advance) with Cu Kα radiation, λ = 1.5418 Å, 3-s step time and 0.075 • step size. Fourier-transform infrared spectroscopy (FTIR) was measured over a range of 480-4000 cm −1 using a Perkin Elmer Fourier-transform infrared spectrometer (Spectrum GX FT-IR System). Thermogravimetric analyses (TGA) were performed on a TA Instruments thermogravimetric analyzer (SDT 2960 Simultaneous DSC-TGA), with samples heated from 25 to 800 • C (heating rate 20 • C/min) under an N 2 atmosphere. Magnetic properties of samples were investigated using a Lakeshore vibrating sample magnetometer (VSM 7404) at room temperature (applied field −6000 to 6000 Oersted). Transmission electron microscopy (TEM) and high-resolution TEM were carried out on a JEOL transmission electron microscope (2100), operated at 200 keV. Zeta potential measurements were performed on a Malvern Instruments zeta potential analyzer (Zetasizer, Nano ZS), in water at Appl. Sci. 2020, 10, 8262 4 of 18 25 • C. Surface properties of sorbents were examined using N 2 sorption on a Quantachrome Instrument (Nova 2000e). All samples were degassed at 200 • C for 10 h prior to measurements. Quantification of cysteine loading in Cys-C@Fe 3 O 4 particles was obtained through combustion sulfur analysis using a LECO (CHNS-628S) elemental analyzer.

Hg(II) Adsorption Test
The Hg(II) adsorption capacity of sorbents was examined using the following procedure. Typically, 1 mg of the sorbent was dispersed into 100 mL of aqueous 1000 ppb Hg(II) (synthetic wastewater) in a polypropylene bottle. The suspension was shaken with a thermostatic shaker at 25 • C and aliquots were removed at regular time intervals, with dispersed sorbent removed from the aliquot using a magnet. The liquid sample was then stabilized by addition of 5% (w/v) KMnO 4 solution, then mixing with 1% (v/v) HCl solution. Prior to determination of the Hg(II) content, addition of sufficient amount of 12% (w/v) hydroxylamine aqueous solution is required to remove the excess KMnO 4 and to acquire clear solutions. Atomic absorption spectroscopy (Perkin Elmer, AAnalyst 200 equipped with FIAS 100 system) was utilized to determine the concentration of Hg(II) in the aliquot sample, through coupling with flow injection analysis system-hydride generation (FIAS-HG) (reductant 0.2% (w/v) NaBH4 in 0.05% (w/v) NaOH, carrier solution 3% (v/v) HCl). The Hg(II) removal efficiency (%) [10,12] or breakthrough (BT, %) [45] was calculated according to the formula where C 0 (mg L −1 ) is the initial concentration of Hg(II) and C t (mg L −1 ) is the concentration of Hg(II) ions after adsorption The mercury uptake (q e , mg g −1 ), and removal efficiency, was calculated according to the equation: where C 0 is the initial mercury concentration in solution (mg L −1 ), C e is the equilibrium concentration of mercury (mg L −1 ) after adsorption, V is the volume of the solution (0.1 L; in this study), and m is the weight of the sorbent used (g). The 500Cys-C@Fe 3 O 4 sample was employed to study pH dependence of Hg(II) removal, while HCl and NaOH were added for pH adjustment. In addition, partition coefficients (PC, mg g −1 µM −1 ) were evaluated using the adsorption capacity or the mercury uptake (q e , mg g −1 ) divided by the equilibrium concentration of mercury (µM) after adsorption [4,46]. PC = (q e /C e ). (3)

Sample Characterization
TEM images of uncoated Fe 3 O 4 and C@Fe 3 O 4 (Figure 1a,b) show individual particles to be spherical, although these undergo aggregation. Uncoated Fe 3 O 4 particles are larger (diameter 40-70 nm) than those of C@Fe 3 O 4 . This finding is consistent with previous work reported by Zheng et al. [34] and Pham et al. [47]. With the addition of D-glucose in the reaction media, a thin carbon shell coating clusters of the magnetite particles are formed (cluster size 50-80 nm, Figure 1b). HRTEM (Figure 1c) reveals nanostructural detail of these C@Fe 3 O 4 clusters, highlighting the amorphous carbon layer and crystalline nature of the magnetite core, as indicated by the lattice spacing (0.29 nm) being consistent with the (002) plane of magnetite [48]. The smaller size of C@Fe 3 O 4 particles relative to uncoated Fe 3 O 4 suggests that glucose acts as both a precursor of carbon shell formation and crystallization inhibitor, resulting in slow Fe 3 O 4 crystal growth. L-Cysteine loading appears to have no effect on the Cys-C@Fe 3 O 4 cluster size and the size of Fe 3 O 4 particles (Figure 2), reflecting the stability of carbon-coated magnetite towards surface functionalization. carbon layer and crystalline nature of the magnetite core, as indicated by the lattice spacing (0.29 nm) being consistent with the (002) plane of magnetite [48]. The smaller size of C@Fe3O4 particles relative to uncoated Fe3O4 suggests that glucose acts as both a precursor of carbon shell formation and crystallization inhibitor, resulting in slow Fe3O4 crystal growth. L-Cysteine loading appears to have no effect on the Cys-C@Fe3O4 cluster size and the size of Fe3O4 particles (Figure 2), reflecting the stability of carbon-coated magnetite towards surface functionalization.   Figure 3a shows PXRD patterns of uncoated Fe3O4, C@Fe3O4, and Cys-C@Fe3O4 particles and that of bulk L-cysteine for comparison. The characteristic peaks of magnetite (Fe3O4, cubic phase, JCPDS file No. 75-1609) [36,48] appear in the diffraction patterns of all particles. However, the intensity of these magnetite peaks in C@Fe3O4 and Cys-C@Fe3O4 samples is significantly lower than those in uncoated Fe3O4, possibly related to shielding by the surface layers [49]. XRD peaks   [36,48] appear in the diffraction patterns of all particles. However, the intensity of these magnetite peaks in C@Fe 3 O 4 and Cys-C@Fe 3 O 4 samples is significantly lower than those in uncoated Fe 3 O 4 , possibly related to shielding by the surface layers [49]. XRD peaks corresponding to graphitic carbon were not observed in C@Fe 3 O 4 and Cys-C@Fe 3 O 4 , suggestive of amorphous carbon coatings in agreement with HRTEM observations. Additional XRD peaks corresponding to crystalline cysteine (2-theta: 18 suggest that cysteine content increases with cysteine loading, as found to be 0.0563, 0.0963, and 0.1828 mmol per gram of solid samples of the 250Cys-C@Fe3O4, 500Cys-C@Fe3O4, and 1000Cys-C@Fe3O4, respectively. Thermogravimetric (TG) analysis plots for all particle systems and that of bulk cysteine are illustrated in Figure 3c. The TGA profile of uncoated Fe3O4 shows around 2% weight loss between 100 and 400 °C, possibly due to loss of residual water and surface hydroxyl groups. On the other hand, three distinct weight loss steps are evident for C@Fe3O4: 200, and >500 °C, corresponding to the loss of surface hydroxyl groups and water molecules and the combustion of amorphous carbon shells [34]. In addition, the TGA curve of bulk cysteine highlights decomposition occurring between 200 and 450 °C, while those of Cys-C@Fe3O4 particles show three weight loss steps, below 230, 230-270, and >500 °C, which coincide with the loss of residual water and hydroxyl groups, cysteine decomposition, and the decomposition of the carbon shell, respectively [33,50]. With no change in weight after 680 °C, the total weight loss results are indicative of 81% wt./wt. Fe3O4 in C@Fe3O4 material. On the other hand, the steep slopes at around 800 °C in in TG plots of 250Cys-C@Fe3O4, 500Cys-C@Fe3O4, and 1000Cys-C@Fe3O4 systems indicate slower combustion (compared with C@Fe3O4), suggesting superior thermal stability arising from the presence of the cysteine layer. Furthermore, the residual mass of solid samples at 800 °C suggests the content of Fe3O4 in 250Cys- The surface functionality of prepared sorbents was probed using FTIR. As shown in Figure 3b, the characteristic Fe-O stretching vibration appears at 570 cm −1 in all Fe 3 O 4 -based samples [36]. Peaks corresponding to C-OH stretching and O-H bending (1069 cm −1 ) occur in C@Fe 3 O 4 , with C=C vibrations at 1620 cm −1 reflecting the successful carbonization of glucose to the coating layer [29]. Moreover, the relative intensity of the O-H vibration (3600-3100 cm −1 ) is significantly higher in C@Fe 3 O 4 samples relative to uncoated magnetite, indicating the greater abundance of these groups on the carbonized surface. The existence of these surface groups facilitates appendage of metal binding functionalities, in this case L-cysteine, to the particle periphery [34]. For Cys-C@Fe 3 O 4 samples, the presence of L-cysteine was confirmed by the N-H vibration overlapping with the O-H stretching frequency at 3600-3100 cm −1 , and two bands at 1600 cm −1 and 1240 cm −1 attributed to C=O and C-O stretching of the amino acid, respectively [29]. Moreover, further peaks at~1500 cm −1 (N-H bending) and 2590 cm −1 (S-H stretching) are evident. These results confirm the grafting of L-cysteine to the particle surface, which most likely occurs via condensation between surface hydroxyl groups and the carboxylic acid moiety in L-cysteine [50]. CHNS analysis results were employed to confirm the presence of cysteine and to determine the cysteine content in Cys-C@Fe 3 O 4 samples. These results suggest that cysteine content increases with cysteine loading, as found to be 0.0563, 0.0963, and 0.1828 mmol per gram of solid samples of the 250Cys-C@Fe 3 O 4 , 500Cys-C@Fe 3 O 4 , and 1000Cys-C@Fe 3 O 4 , respectively.
Thermogravimetric (TG) analysis plots for all particle systems and that of bulk cysteine are illustrated in Figure 3c. The TGA profile of uncoated Fe 3 O 4 shows around 2% weight loss between 100 and 400 • C, possibly due to loss of residual water and surface hydroxyl groups. On the other hand, three distinct weight loss steps are evident for C@Fe 3 O 4 : 40-200, 200, and >500 • C, corresponding to the loss of surface hydroxyl groups and water molecules and the combustion of amorphous carbon shells [34]. In addition, the TGA curve of bulk cysteine highlights decomposition occurring between 200 and 450 • C, while those of Cys-C@Fe 3 O 4 particles show three weight loss steps, below 230, 230-270, and >500 • C, which coincide with the loss of residual water and hydroxyl groups, cysteine decomposition, and the decomposition of the carbon shell, respectively [33,50]. With no change in weight after 680 • C, the total weight loss results are indicative of 81% wt./wt. Fe 3 O 4 in C@Fe 3 O 4 material. On the other hand, the steep slopes at around 800 • C in in TG plots of 250Cys-C@Fe 3 O 4 , 500Cys-C@Fe 3 O 4 , and 1000Cys-C@Fe 3 O 4 systems indicate slower combustion (compared with C@Fe 3 O 4 ), suggesting superior thermal stability arising from the presence of the cysteine layer. Furthermore, the residual mass of solid samples at 800 • C suggests the content of Fe 3 O 4 in 250Cys-C@Fe 3 O 4 , 500Cys-C@Fe 3 O 4 , and 1000Cys-C@Fe 3 O 4 to be not more than 50%, 49%, and 40% wt./wt., respectively, and reflect the increase in cysteine content with initial loading level, as indicated from elemental sulfur analysis. Magnetic properties of all Fe 3 O 4 -based samples were examined using VSM at room temperature. Results depicted in Figure 3d show that the saturation magnetization of uncoated Fe 3 O 4 , C@Fe 3 O 4 , 250Cys-C@Fe 3 O 4 , 500Cys-C@Fe 3 O 4 , and 1000Cys-C@Fe 3 O 4 samples were about 78, 51, 50, 49, and 45 emu.g −1 , respectively. Reductions in saturation magnetization after carbon coating and cysteine functionalization can be attributed to the decreasing fraction of magnetic material in these particles relative to diamagnetic coating materials (carbon and cysteine) [50,51]. Figure S1 shows N 2 adsorption-desorption isotherms and pore size distributions of all samples, with porous structural data listed in Table 1. It can be concluded that isotherms of all materials are of type-IV with a type-H3 hysteresis loop, typical of mesoporous materials [51]. The greater specific area of C@Fe 3 O 4 (90 m 2 .g −1 ) over that of uncoated Fe 3 O 4 (55 m 2 .g −1 ) may be a consequence of the presence of pores in the carbon coating. Cysteine-modified C@Fe 3 O 4 samples show lower specific surface areas and total pore volumes than C@Fe 3 O 4 due to grafting of the surface L-cysteine ligands, which results in blockage of surface channels and reduced porosity [52,53]. Table 1. Porous structural data for as-synthesized samples (S BET , specific surface area by BET; V t , total pore volume; D a , average pore diameter).  Figure 4 reports that at the initial Hg(II) concentration of 1000 ppb, 100% breakthrough (BT) was not achievable for all of the studied sorbents after 60 min treatments. Low %BT values (<30%) for C@ Fe 3 O 4 and uncoated Fe 3 O 4 were obtained even after 60 min treatments. On the other hand, 90% BT (q e = 90.03 mg g −1 ) for 500 Cys-C@ Fe 3 O 4 were obtained after treatments for 45 min and above, while about 40%BT (q e = 42.0 mg g −1 ) was achieved after 5 min. At BT time of 60 min, the %BT follows the order of 500Cys-C@Fe 3 O 4 > 1000 Cys-C@Fe 3 O 4 > 250 Cys-C@ Fe 3 O 4 >> C@ Fe 3 O 4 > uncoated Fe 3 O 4 , being consistent with the specific surface area, surface functional groups, and adsorption capacity relationship previously discussed. Notably, times to reach 10%BT, 50%BT, and 90%BT of the prepared magnetic sorbents, interpolated from Figure 4, are reported in Table S1. After 60 min of 500 Cys-C@ Fe 3 O 4 treatments, 50%BT and 90%BT were achievable, respectively, at pH = 2 and pH ≥ 6 ( Figure S2 and Table S2).  Figure 4 reports that at the initial Hg(II) concentration of 1000 ppb, 100% breakthrough (BT) was not achievable for all of the studied sorbents after 60 min treatments. Low %BT values (<30%) for C@ Fe3O4 and uncoated Fe3O4 were obtained even after 60 min treatments. On the other hand, 90% BT (qe = 90.03 mg g −1 ) for 500 Cys-C@ Fe3O4 were obtained after treatments for 45 min and above, while about 40%BT (qe = 42.0 mg g −1 ) was achieved after 5 min. At BT time of 60 min, the %BT follows the order of 500Cys-C@Fe3O4 > 1000 Cys-C@Fe3O4 > 250 Cys-C@ Fe3O4 >> C@ Fe3O4 > uncoated Fe3O4, being consistent with the specific surface area, surface functional groups, and adsorption capacity relationship previously discussed. Notably, times to reach 10%BT, 50%BT, and 90%BT of the prepared magnetic sorbents, interpolated from Figure 4, are reported in Table S1. After 60 min of 500 Cys-C@ Fe3O4 treatments, 50%BT and 90%BT were achievable, respectively, at pH = 2 and pH ≥ 6 ( Figure S2 and Table S2). Kinetic results for Hg(II) adsorption, measured using 0.01 mg L −1 sorbent suspended in 1000 ppb Hg(II) aqueous solution at pH 6, are illustrated in Figure 5a. For all samples, high Hg adsorption rate was observed during the first 10 min and reached an equilibrium after 20 min. The greater adsorption efficiency of C@Fe 3 O 4 relative to uncoated magnetite is likely due to the presence of surface hydroxyl groups, and its higher surface area. For Cys-C@Fe 3 O 4 samples, which exhibit lower surface area than C@Fe 3 O 4 , the relatively high adsorption capacity is related to the metal-chelating ability of the -NH 2 and -SH functional groups in the appended L-cysteine. Notably, the highest Hg uptake was 90 mg.g −1 by 500Cys-C@Fe 3 O 4 . Further increases in cysteine content (1000Cys-C@Fe 3 O 4 ) resulted in lower Hg(II) removal efficiency, likely due to the deposition of bulk cysteine on the carbon shell (XRD result, Figure 3a) that could inhibit effective adsorption by internal pores. Mercury removal experiments were further conducted for longer periods (60 min) to ensure adsorption saturation. To study the adsorption mechanism, 500Cys-C@Fe 3 O 4 , which exhibits the highest level of Hg(II) uptake, was selected as representative of Cys-C@Fe 3 O 4 , along with uncoated Fe 3 O 4 and C@Fe 3 O 4 for comparison. Pseudo-second-order kinetic models (Figure 5b) were then applied to probe the mechanism [33,54]. The correlation coefficients for the pseudo-second-order kinetic model are very close to 1, and the calculated equilibrium adsorption capacities (q e,cal ) derived from this model are also much closer to the experimentally determined adsorption capacities (q e,exp ) than that derived using the pseudo first order kinetic model. These results also support the assumption that sorption should occur through chemical adsorption rather than mass transport [55].  Figure 5a,d implied the high data reproducibility.

Adsorption Isotherm
The relationship between the equilibrated mercury uptake (qe) of sorbent particles and the equilibrated concentration of Hg(II) in solution (Ce) was used to quantitatively describe the sorption, by way of adsorption isotherms. In this case, experiments were performed under conditions as used in the kinetic studies, but initial Hg(II) concentrations were varied from 60 to 1000 ppb. The adsorption time of 60 min is sufficient for adsorption-desorption equilibrium, as the difference in Hg(II) uptake from 60 min treatments and from longer treatment time was found negligible. As depicted in Figure 5c, Hg(II) uptake increased with initial Hg(II) concentration until it reached a saturation point. According to these profiles, the saturated adsorption capacities of uncoated Fe3O4, C@Fe3O4, and 500Cys-C@Fe3O4 were 12, 28, and 90 mg g −1 , respectively. The equilibrium data were

Adsorption Isotherm
The relationship between the equilibrated mercury uptake (q e ) of sorbent particles and the equilibrated concentration of Hg(II) in solution (C e ) was used to quantitatively describe the sorption, by way of adsorption isotherms. In this case, experiments were performed under conditions as used in the kinetic studies, but initial Hg(II) concentrations were varied from 60 to 1000 ppb. The adsorption time of 60 min is sufficient for adsorption-desorption equilibrium, as the difference in Hg(II) uptake from 60 min treatments and from longer treatment time was found negligible. As depicted in Figure 5c, Hg(II) uptake increased with initial Hg(II) concentration until it reached a saturation point. According to these profiles, the saturated adsorption capacities of uncoated Fe 3 O 4 , C@Fe 3 O 4 , and 500Cys-C@Fe 3 O 4 were 12, 28, and 90 mg g −1 , respectively. The equilibrium data were then fitted using the Langmuir model ( Figure S3a), allowing prediction of equilibrium parameters [56].
In Equation (4), S ad is the number of ions adsorbed per unit area of sorbent surface (mg g −1 ), S aq is the equilibrium concentration of ions in solution (mg L −1 ), N max is the maximum adsorption capacity (mg g −1 ) corresponding to complete monolayer coverage on the sorbent surface, and K is the Langmuir adsorption constant (L mol −1 ) related to binding strength. From fitting results shown in Table 2 and comparisons with results calculated using the Freundlich model [15] (data shown in Figure S3b), adsorption clearly follows Langmuir behavior indicating formation of a uniform monolayer of metal ions adsorbed on the particle surface [50]. Furthermore, it is important to note that the calculated maximum adsorption capacity of uncoated Fe 3 O 4 (12.56 mg g −1 ) is enhanced by carbon coating (29.85 mg g −1 ) and more so, by cysteine functionalization (94.33 mg g −1 ).

Effect of pH on Hg(II) Adsorption
Figure 5a,d both showed that the Hg uptake reached maximum Hg(II) uptake values after 20 min. Figure 5d shows the effect of pH on Hg(II) adsorption by 500Cys-C@Fe 3 O 4 , revealing that removal capacity is highly pH dependent. The aqueous formation of charged Hg(II) species (Hg 2+ , HgOH + ) and neutral Hg(OH) 2 can result at different pH values, although these are all capable of reacting with L-cysteine thiol groups due to the high affinity of Hg for sulfur [31]. As shown in Figure 5e, speciation curves calculated using the Visual MINTEQ software [26] indicate that free Hg(II) ions predominate at pH ≤ 2, which in the presence of thiol groups may form positively charged complexes (S-Hg + ). The presence of large numbers of such complexes on the sorbent surface may, at some point, inhibit further binding based on electrostatic repulsion. At higher pH values mercuric hydroxide species (HgOH + and Hg(OH) 2 )) become more prevalent, which are able to bind to cysteine thiol groups forming neutral complexes (−S−HgOH) [57]. The buildup of uncharged Hg(II) complexes on the sorbent surface does not retard further Hg(II) adsorption in the same way as accumulation of positive charge. Accordingly, 500Cys-C@Fe 3 O 4 exhibits greater Hg(II) removal performance at higher pH values, as shown in Figure 5d. These findings are in agreement with zeta potential measurements with 500Cys-C@Fe 3 O 4 (Figure 5f), which indicate that the particle surface is positively charged at a solution pH lower than the Point Zero Charge (PZC,~5.4) and becomes negatively charged only above this value.

Adsorption of Hg(II) at Ultralow Concentrations
To further demonstrate the efficiency of 500Cys-C@Fe 3 O 4 for Hg(II) removal, adsorption studies were conducted using ultralow Hg(II) concentrations. Adsorption experiments using Hg(II) concentrations ranging from 100 to 5 ppb (pH 6) demonstrated the performance of 500Cys-C@Fe 3 O 4 , with >99% Hg(II) removal being possible in 1 h over the entire concentration range, from 100, 50, 20, 10, and 5 ppb to 0.86, 0.45, 0.36, 0.34, and 0.31 ppb, respectively. These final Hg(II) concentrations were less than 1 ppb, meeting acceptable limits for drinking water set down by the WHO and EU [8].

Selectivity of 500Cys-C@Fe 3 O 4 Sorbent
Many types of metal ions coexist in real water samples; competitive binding of these with Hg(II) to sorbents is, therefore, a serious issue potentially governing sorbent performance. In this work, three contaminant metal ions commonly found in industrial wastewater [58,59], Cu(II), Pb(II), and Ni(II), were selected as the basis for competitive adsorption studies against Hg(II) using 500Cys-C@Fe 3 O 4 . As seen in Table 3, these ions have negligible effect towards the adsorption of Hg(II). This can be rationalized using the Hard-Soft Acid-Base (HSAB) theory: Hg(II) is considered as a soft metal (relatively large ionic size, low electronegativity, and high polarizability) with higher affinity for soft donor atoms (such as sulfur) than relatively hard metal ions (Cu(II), Pb(II), and Ni(II)) [60]. Table 3. Concentrations of Hg(II) Pb(II), Ni(II), and Cu(II) ions in a simulated wastewater before and after treatment with 500Cys-C@Fe 3 O 4 sorbent. The concentrations of each metal ion in the treated wastewater shows the amount of residual metal contaminants after adsorption.

Determination of Thermodynamic Parameters
To study the thermodynamics of Hg(II) adsorption, a series of adsorption tests were performed, each using 1.0 mg of 500Cys-C@Fe 3 O 4 with 100 mL of 1000 ppb Hg(II) solution at different temperatures (25, 35, or 40 • C). The thermodynamic feasibility and spontaneous nature of the adsorption process were assessed based on thermodynamic parameters including the standard Gibbs free energy (∆G • ), enthalpy change (∆H • ), and entropy change (∆S • ), as determined from the following equations [54,55,57].
where K d is the distribution coefficient of adsorption, R is the universal gas constant (8.314 J.mol −1 .K −1 ), and T is the absolute temperature (K). q e and C e are the mercury uptake and equilibrium concentration of mercury (mg.L −1 ) after adsorption, respectively (see Experimental section). Values of ∆H • (−24.1 kJ mol −1 ) and ∆S • (−72.7 J mol −1 K −1 ) were quantified from the slope and intercept of the linear van't Hoff plot, as seen in Figure S4. The values of ∆G • (−2.93 kJ mol −1 at 298 K, −1.70 kJ mol −1 at 308 K, and −1.27 kJ mol −1 at 313 K) suggest that Hg(II) adsorption on the cysteine functionalized carbon-coated magnetite is thermodynamically feasible. The decrease in −∆G • with increasing temperature reveals that the sorption process is exothermic and more favorable at lower temperatures, similar to previous work [15,55]. Joen et al. [61] has suggested that escape of mercury ions from surface sites is more facile at higher temperature. The negative value of ∆H • indicates that Hg(II) sorption is exothermic at 25-40 • C. The value of ∆S • is also negative, suggesting a decrease in randomness at the solid/solution interface during Hg(II) sorption.

Sorbent Recyclability
As shown in the Figure 6 inset, 500Cys-C@Fe 3 O 4 sorbent can be well dispersed in aqueous media and can be harvested easily by magnetic separation. Desorption of Hg(II) from the sorbent surface was examined to provide a measure of sorbent stability and recyclability, both being important parameters for industrial applications. For 500Cys-C@Fe 3 O 4 , experiments were conducted using an eluent solution (10 mL) composed of 0.1 M HCl and 5% (w/v) thiourea. Prior to measurements, the used sorbent was washed three times with deionized water. After adding the eluent solution to the used sorbent, desorption was allowed to proceed by shaking the suspension for 1 h at a constant temperature. It is possible that dissolution of magnetite particles could occur during desorption under the experimental conditions. Atomic absorption spectroscopy of the eluent solution after the first desorption cycle indicated an Fe concentration of 0.34 ppm for 500Cys-C@Fe 3 O 4 , and 0.51 ppm and 5.94 ppm for C@Fe 3 O 4 and uncoated Fe 3 O 4 nanoparticles, respectively, highlighting the importance of the surface coating on maintaining core integrity. Results of desorption and adsorption for five cycles using the 500Cys-C@Fe 3 O 4 sorbent are shown in Figure 6. These reveal that the sorbent maintains high desorption efficiency (90%) and Hg uptake (90 mg.g −1 ) performance over the first three cycles. However, adsorption efficiencies significantly decreased for the fourth and fifth cycles, while the desorption efficiency only slightly decreased. This behavior could result from saturation of the binding sites and/or degradation of bound cysteine moieties at low pH [62,63].
Appl. Sci. 2020, 10, x FOR PEER REVIEW 13 of 19 Hg(II) sorption is exothermic at 25-40 °C. The value of ΔS° is also negative, suggesting a decrease in randomness at the solid/solution interface during Hg(II) sorption.

Sorbent Recyclability
As shown in the Figure 6 inset, 500Cys-C@Fe3O4 sorbent can be well dispersed in aqueous media and can be harvested easily by magnetic separation. Desorption of Hg(II) from the sorbent surface was examined to provide a measure of sorbent stability and recyclability, both being important parameters for industrial applications. For 500Cys-C@Fe3O4, experiments were conducted using an eluent solution (10 mL) composed of 0.1 M HCl and 5% (w/v) thiourea. Prior to measurements, the used sorbent was washed three times with deionized water. After adding the eluent solution to the used sorbent, desorption was allowed to proceed by shaking the suspension for 1 h at a constant temperature. It is possible that dissolution of magnetite particles could occur during desorption under the experimental conditions. Atomic absorption spectroscopy of the eluent solution after the first desorption cycle indicated an Fe concentration of 0.34 ppm for 500Cys-C@Fe3O4, and 0.51 ppm and 5.94 ppm for C@Fe3O4 and uncoated Fe3O4 nanoparticles, respectively, highlighting the importance of the surface coating on maintaining core integrity. Results of desorption and adsorption for five cycles using the 500Cys-C@Fe3O4 sorbent are shown in Figure 6. These reveal that the sorbent maintains high desorption efficiency (90%) and Hg uptake (90 mg.g −1 ) performance over the first three cycles. However, adsorption efficiencies significantly decreased for the fourth and fifth cycles, while the desorption efficiency only slightly decreased. This behavior could result from saturation of the binding sites and/or degradation of bound cysteine moieties at low pH [62,63].  Table 4 highlights examples of sorbents applied to removal of ultralow concentrations (<20 ppb of Hg(II)), in comparison to 500Cys-C@Fe3O4 material. From these, 500Cys-C@Fe3O4 shows comparable performance at much lower sorbent loadings, although longer contact times were employed in this work. Thio-silica-coated magnetite required very low sorbent loadings for Hg(II) removal, but its preparation is more complex, requiring more steps than that needed for Cys-C@Fe3O4 [15]. Acrylamide-hydroxyl-MOFs [10] and MOF-74-Zn [16] show lower removal efficiencies using higher sorbent loadings and are nonmagnetic, thus cannot be separated for reuse by an application  Table 4 highlights examples of sorbents applied to removal of ultralow concentrations (<20 ppb of Hg(II)), in comparison to 500Cys-C@Fe 3 O 4 material. From these, 500Cys-C@Fe 3 O 4 shows comparable performance at much lower sorbent loadings, although longer contact times were employed in this work. Thio-silica-coated magnetite required very low sorbent loadings for Hg(II) removal, but its preparation is more complex, requiring more steps than that needed for Cys-C@Fe 3 O 4 [15]. Acrylamide-hydroxyl-MOFs [10] and MOF-74-Zn [16] show lower removal efficiencies using higher sorbent loadings and are nonmagnetic, thus cannot be separated for reuse by an application of an external magnet. Furthermore, the adsorption capacity and satisfactory level of sorbent performance in removal of ultralow concentrations of Hg(II) based on the WHO and EU drinking water standards is reported in Table 5, in which 500Cys-C@Fe 3 O 4 material shows superior comparative performance at low sorbent loadings. Comparative PC values of 500 Cys-C@Fe 3 O 4 material, at varying initial Hg(II) concentrations (5-1000 ppb), were evaluated using Equation (2). They are reported in Table 5. The obtained PC values of the developed magnetic material at the studied Hg(II) concentration range were found being much higher than other sorbents. In particular, the significantly high PC values of the Cys-C@Fe 3 O 4 material at ultralow Hg(II) initial concentrations (≤20 ppb) should be attributed to the high chelating ability of Hg(II) ions and the -NH 2 and -SH functional groups present on the surface of magnetic material. Table 5. Comparative sorption capacities, Hg concentrations before and after treatment, Hg uptake, and partition coefficient values for various sorbents for Hg(II) removal at ultralow metal concentrations. The numbers in bold indicate the Hg(II) concentration lower than the World Health Organization (WHO) and European Union (EU) permissible limits of total mercury in drinking water. Note that n.d. is denoted as not determined as the volume of treated Hg(II) solution is unknown, while n.a is denoted as not applied due to the partition coefficient (PC) value is infinity using Equation (3).

Conclusions
Efficient Hg(II) sorbents based on magnetite particles have been prepared using a simple synthetic procedure from low-cost materials; cost per kilogram [64] was urea (USD $0.2), L-cysteine (USD $10), D-Glucose (USD $0.4), ethylene glycol (USD $0.5), and hydrated ferric chloride (USD $0.65). Utilizing cysteine as a functionalization reagent for magnetic sorbents is much more economic and simpler, compared with productions of functionalized magnetic sorbents via multistep material synthesis [12,13,15], expensive functionalization reagent, or high-purity chemicals [10,12,13,15,16]. These sorbents are easy to separate from aqueous media and demonstrate excellent recyclability metrics. Comparative studies were undertaken for uncoated Fe 3 O 4 , carbon-coated Fe 3 O 4 (C@Fe 3 O 4 ), and cysteine-functionalized carbon-coated Fe 3 O 4 (Cys-C@Fe 3 O 4 ) sorbent particles. The 500Cys-C@Fe 3 O 4 sorbent, having optimum L-cysteine loadings, are highly selective for Hg(II) adsorption, are capable of removing 90% of Hg(II) within 10 min, and are effective even at ultralow metal concentrations. The carbon shell enhances the Hg(II) sorption capacity, preserves the integrity of the Fe 3 O 4 core, and acts as a scaffold, allowing appendage of metal chelating L-cysteine to the particle surface. Surface modification of C@Fe 3 O 4 by cysteine boosts the performance and selectivity of the materials, resulting in enhanced levels of Hg(II) removal with fast adsorption rates, at ultralow Hg(II) concentrations. Moreover, factors potentially influencing adsorption efficiency, such as pH and temperature, were comprehensively studied in respect to both kinetics and thermodynamics. The facile preparation, fast and impressive adsorption performance, good selectivity, ease of separation, and recyclability alludes to the high potential of Cys-C@Fe 3 O 4 as a candidate for selective removal of Hg(II) contaminants from water sources and for its possible adoption in mainstream water treatment processes.