Arsenic removal by magnetite-loaded amino modified nano/microcellulose adsorbents: Effect of functionalization and media size

3.2.1 Textural properties and pH_PZC

Multi-step synthesis of adsorbents was applied in order to design the material with an appropriate geometry, surface properties and pore structure. The textural properties and zero point of charge (pH_PZC) before and after adsorption are summarized in Table 1.

Due to dimensional differences of parent materials, the higher value of specific surface area and pore volume was in favor of NC-MA/L-MG, while higher pore diameter was found for MC-O/L-MG compared to the raw MG. Literature data of unmodified nano- and micro fibrillated cellulose (Dufresne, 2012), showed that differences in surface area and pore diameter are generally influenced by the textural properties of substrate, applied modification method as well as post processing of obtained material. In contrast to mild thermal treatment applied for NC-MA/L-MG and MC-O/L-MG production, high temperature processing exerts significant change in textural properties of final material. The surface area and pore volume of thermally treated silica, after ligand immobilization, showed significant decrease in textural properties (Awual, 2015; Awual et al., 2016a, 2016b; Awual, 2016a, 2016b, 2016c). Regardless of obtained results both methodologies could find application in production of adsorbents of different properties. Precipitation of MG on NC-MA/L and MC-O/L provided such conditions applicable for preservation of organic branched structures with controllable deposition of nanoscale materials. The shift of pH_PZC value (Table 1) indicates specific arsenate adsorption rather than electrostatic interaction, and formation of complexed/precipitated arsenic species at surface of magnetite based adsorbent is a main adsorption mechanism (Markovski et al., 2014a, 2014b).

3.2.2 XRD analysis

XRD diffractograms of NC and NC-MA, presented in Fig. 3(a), provide insight into structural changes in the course of NC modification. An increase in crystallinity of NC-MA was observed. Both diffractograms display two well-defined peak doublets around 2θ = 20° and 22.0° which correspond to (2 0 0) plane of cellulose crystal structure. Those doublets indicate the existence of cellulose I and cellulose II allomorphs (Peng et al., 2013). Modification with MA causes sharper signal, and increases contribution of cellulose II crystallites, and overall degree of crystallinity. In the sample of NC-MA signal at 12.24° corresponds to (−1 1 0) crystal plane of cellulose I (Klemm et al., 2011). The signal at 35.0° showed higher intensity for NC-MA, corresponds to (0 0 4) atomic plane of cellulose I (Peng et al., 2013). Also, it was shown that appearance of new crystalline peaks at 2θ ∼ 35° as well as at 41.0° indicates the increase in the crystallinity of NC-MA (Maiti et al., 2013).

Close

Figure 3

XRD diffractograms of (a) NC and NC-MA, and (b) MC-O/L-MG and NC-MA/L-MG.

Export to PPT

Phase and structural analysis of NC-MA/L-MG and MC-O/L-MG confirmed the presence of MG (Fe₃O₄), as it was indicated by peak at 30.16, 35.52, 37.16, 43.17, 54.22, 57.10, 62.70° (JCPDS 19-629) (Fig. 3(b)). Differences in peak broadness and intensity indicate lower degree of crystalline phase of embedded MG in MC-O/L-MG. The mean crystallite size, calculated by using Scherrer equation based on the half width of the most intense peaks in XRD patterns, indicates 22 and 19 nm of grain sizes of MG present on NC-MA/L-MG and MC-O/L-MG surface, respectively. These values are somewhat lower than ones observed by transmission electron microscopy (TEM) (Fig. 7). A somewhat lower crystallite size, found for MC-O/L-MG, is due to lower number of amino groups (crystallization centers) accessible for iron coordination. Obtained results indicated the high coverage of NC-MA/L and somewhat lower for MC-O/L by MG. Due to the known fact that MG and maghemite (γ-Fe₂O₃) have similar XRD pattern it was necessary to apply Raman spectroscopy in order to clearly distinguish nanocrystalline form of deposited iron oxide.

3.2.3 Raman spectroscopy

Raman spectra of the NC, NC-MA/L and NC-MA/L-MG, and MC and MC-O are shown in Fig. 4. The region below 1750 cm⁻¹, observed at Raman spectra of NC and NC-MA/L, originates from the conformation of cellulose skeleton, while the region above 2888 cm⁻¹ and higher than 3000 cm⁻¹ originates from hydrogen bonding of OH group (Szymańska-Chargot et al., 2011). Comparing the Raman spectra of the NC and NC-MA/L, the overlapped bands, observed in the region 3215–3405 cm⁻¹, are due to OH and NH₂ stretching vibration (Gabelica, 1976). Also, the signals at 1456 cm⁻¹ and 1473 cm⁻¹, a characteristic for methylene bending modes of two stereo-chemically nonequivalent CH₂OH groups of the crystalline NC skeletons, merge into one single signal at 1456 cm⁻¹ in NC-MA/L. This change explained the differences between the vibrational behavior of crystalline and amorphous domain in NC (Schenzel et al., 2005), and indicated transition from crystalline to amorphous form of these two CH₂ bending modes during surface modification (Schenzel and Fischer, 2001; Wiley and Atalla, 1987). Also, the intensity of the characteristic peak of highly crystalline NC (1473 cm⁻¹), and other signals decrease in comparison with amorphous NC-MA/L domain.

Close

Figure 4

Raman spectra of (a) NC, NC-MA/L and NC-MA/L-MG, and (b) MC and MC-O.

Export to PPT

Weak bands, observed in region from 964 cm⁻¹ to 1113 cm⁻¹, originate from C6 atom bending vibrations (H—C—C and H—C—O), while the strong stretching vibration of unhydroglucose skeleton is observed at 1145 cm⁻¹ (C—C, C—O and C—O—C) (Schenzel et al., 2005; Wiley and Atalla, 1987). The NC backbone deformation vibration bands, δ(CH₂), δ(HCC), δ(HCO) and δ(COH), are observed at 1374 cm⁻¹ (Zhang et al., 2011). Wide bands observed in the region between 1565 cm⁻¹–1992 cm⁻¹ and 900 cm⁻¹–1500 cm⁻¹ originate from the MA and EDTA moieties in NC-MA/L, respectively.

The non-polarized spectrum of NC-MA/L-MG shows magnetite predicted phonon bands as characteristic peak at 670 cm⁻¹ (Shebanova and Lazor, 2003). Due to susceptibility of MG to transform into maghemite, under exposure to laser, two weak peaks appeared at 350 and 500 cm⁻¹. The absence of NC characteristic bands in NC-MA/L-MG spectrum indicates homogenous distribution/entire surface coverage by deposited MG on amino functionalized NC.

The Raman spectra of the MC and MC-O showed similar peak structure and position, except that cellulosic structure of MC basis participates at higher extent in response to laser impulse. Regardless of adjustment of laser power the Raman spectrum of MC-O/L-MG sample was not obtained.

3.2.4 FTIR spectroscopy

The FTIR spectra of intermediates and adsorbents, before and after arsenate adsorption, are presented in Fig. 5. From the FTIR spectra of all intermediates, a broad band about 3423 cm⁻¹ ascribed to OH and NH stretching vibrations can be observed. Adsorption of arsenate oxyanions causes shifting of the stretching vibration of OH band to 3451 cm⁻¹. Differences in the peak intensity, peak shifting and appearance of peak after adsorption indicate arsenate bonding by appropriate functional groups. Carbonyl group absorption at 1714 cm⁻¹, found in NC-MA, is missing in the spectrum after MG modification. Absorption peak, at 1635 cm⁻¹, originates from deformation vibrations of OH bond, and is negligibly shifted to 1637 cm⁻¹ after MG modification. Broad bands at ≈1714 and ≈1635 cm⁻¹, are shifted to ≈1637 cm⁻¹, in NC-MA/EDA and NC-MA/L, a band corresponding to amide stretching vibration (amide I) coupled with in-plane N—H deformation vibration. The bands in the region 750–1000 cm⁻¹, and one at 1373 cm⁻¹, show the presence of sulfonates and amino groups (Socrates, 2004). The strong bands at 582 cm⁻¹, assigned to Fe—OH vibrations present at MG surface, after arsenate adsorption appeared at 617 cm⁻¹. This region corresponds to vibration of As—O—Fe complexed/precipitated arsenate species, and at different surface coverages, bidentate binuclear and mononuclear monodentate complex could be formed, where As—OH free bonds are present as unprotonated and/or protonated depending on pH (Markovski et al., 2014a, 2014b; Grossl et al., 1997).

Close

Figure 5

FTIR spectra of (a) NC, NC-MA, NC-MA/EDA and NC-MA/L and (b) NC-MA/L-MG, MC-O/L-MG, NC-MA/L-MG/As and MC-O/L-MG/As.

Export to PPT

3.2.5 Thermal analysis (TGA)

The success of nano/microcellulose modification and thermal stability of composites constituent was evaluated by the use of TGA. TGA and differential thermal analysis (DTG) curves, recorded for NC, MC, NC-MA/L, NC-MA/L-MG and MC-O/L-MG, are shown in Fig. 6. The TGA curve and DTG peaks, observed at 258 °C for NC and 260.5 °C for MC, indicate similar thermal behavior of both materials in whole temperature range. As it can be observed, the thermal decomposition of NC and MC occurred in three steps: (i) 30–225 °C, (ii) 225–325 °C, and (iii) 325–700 °C range. In the first stage, ∼5% and ∼11% weight loss were observed due to dehydration of NC and MC, respectively. In the second stage, almost 61% mass loss occurred due to the splitting/thermal decomposition of cellulose structure, i.e. chain scission which evolves gaseous products leaving condensation/carbonaceous residues. After 325 °C, 20% and 18% weight loss were observed due to the degradation of the NC and MC condensed carbonaceous material, respectively. The end of first region is shifted to the lower temperature (145 °C) for the NC-MA/L sample. Also, the weight loss difference, ∼12%, between NC and NC-MA/L is similar to evaluated organic fraction, obtained according to amino groups calculation ∼18% (Section 3.1). DTG peak for NC-MA/L occured at 158 °C refers to dehydration and evaporation of residual solvents, while DTG peak at 273.4 °C refers to cellulose backbone decomposition.

Close

Figure 6

TGA (a) and DTG (b) curves of NC, MC, NC-MA/L, NC-MA/L-MG and MC-O/L-MG.

Export to PPT

Close

Figure 7

FEG-SEM images of NC-MA/L (a), NC-MA/L-MG (b), MC-O/L-MG (d) and (e); and TEM images of NC-MA/L-MG (c) and MC-O/L-MG (f).

Export to PPT

The higher stability, comparing to NC and MC, and similar thermal behavior of NC-MA/L-MG and MC-O/L-MG were noticed (Fig. 6(a)). Higher thermal stability is due to contribution of MG protective layer which forms, by sintering of MG nanocrystallite, a stable core shell-like structure. The first stage of mass loss (<300 °C) is due to removal of physically adsorbed water, residual reactants and amorphous cellulose domain. The weight loss in the range from 300 to 450 °C is ascribed to the rapture of weak head to head acetal linkage between cellobiose units in an amorphous/lower crystallinity domain of cellulose. Corresponding DTG peaks could be observed at 306.8 °C, 324.6 °C and 379 °C for MC-O/L-MG and 394.2 °C for NC-L-MG, respectively. The third stage, at >400 °C, takes place by random scission/gradual decomposition of main cellulose polymeric chain. A residual inorganic material of 42% and 28% in an inert atmosphere, and comparative experiments in air showed that 33% and 17% residuals in NC-L-MG and MC-O/L-MG were obtained, respectively. On the DTG curves of NC-L-MG and MC-O/L-MG peaks observed at temperatures >470 °C refer to thermal oxidation and dehydroxylation of precipitated magnetite (Przepiera and Przepiera, 2001).

3.2.6 Morphological characterization

The morphology of the synthesized adsorbents was analyzed by using SEM/FEG–SEM techniques, results given in Fig. 7(a), (b), (d) and (e), while TEM results are given in Fig. 7(c) and (f).

Morphology of NC-MA/L (Fig. 7a) indicated surface coverage by organic material due to chemical modification, and subsequent MG deposition affects the morphological structure of the surface of NC-MA/L-MG (Fig. 7b). SEM micrograph of NC-MA/L-MG adsorbent obtained at lower magnification is shown in Fig. S1 (Supplementary material).

SEM micrograph of MC-O/L-MG adsorbent is given in Fig. 7(d) and (e). The evaluated average diameter of MC, MC-O and MC-O/L-MG was in the range 30 ± 1.2 μm. The modified MC fibers showed small diameter changes of parent cylindrical strands due to organic modification and MG precipitation at MC-O/L surface. Evolution of the MC-O/L-MG morphology can be explained by heterogeneous reaction which takes place by MG precipitation on terminal amino branched structure with incorporated NC grains on the fiber surface.

TEM analysis (Fig. 7(c) and (f)) confirmed nanoscale dimension and irregular distribution of cubic MG nanoparticles onto irregular/porous morphology of NC-MA/L and MC-O/L. It can be concluded that morphology of NC-MA/L-MG and MC-O/L-MG adsorbents is the result of both physical structure of nano/microscale material, and chemical structure/properties of functional groups present at NC-MA/L and MC-O/L surface.

3.3.1 pH-dependent, equilibrium, thermodynamic and competitive adsorption study

Considering significance of pH influences on arsenic speciation and ionization state of adsorbent surface, it was important to study influence of pH (in the range 3–10) on effectiveness of As removal. The results of the pH-dependent adsorption study, i.e. arsenate removal versus initial pH (pH_i) for both NC-MA/L-MG and MC-O/L-MG adsorbents, and change of final pH (pH_f) in relation to initial pH (pH_i) are shown in Fig. 8(a). The extent of adsorption pH_f/pH_i relation is affected by both release/consumption of hydrogen ions, i.e. protonation/deprotonation of MG surface, and pH-dependent arsenic speciation. These factors determine that effective As(V) adsorption was obtained in pH range 3.0–7.0 (Fig. 8(a)), while pH_f/pH_i relationship indicates similar arsenic adsorption mechanism with respect to both adsorbents. Similar behavior was found for As(III) pH-dependent adsorption (Fig. S2).

Close

Figure 8

Effect of pH on adsorption of As(V) on NC-MA/L-MG and MC-O/L-MG, and pH_f versus pH_i dependence (a); Freundlich fitting of adsorption data for NC-MA/L-MG (b) and MC-O/L-MG adsorbents (c).

Export to PPT

The pKa values of arsenic acid: pKa₁ ∼ 2.3, pKa₂ ∼ 7.0 and pKa₃ ∼ 11.5, indicate that molecular form mainly exists in a solution at pH < 2.0, and pH-dependent equilibrium of anionic species, H₂AsO₄⁻ or HAsO₄²⁻, exists at pH in the range 2–10 (Fig. S3) (Issa et al., 2011). At pH_i < pH_PZC (Table 1), the protonation of surface functional generates positive charge which contributes to favorable bonding of negatively charged arsenate ions (Fig. S3). Both surface state and As speciation play significant contribution to electrostatic interaction (attraction/repulsion) between surface/ions causing intensity of As flux toward the specific adsorption sites. Enhancement of electrostatic attraction is thus highly feasible for As(V) species, while it is of minor importance for the neutral form of arsenous acid (Fig. S3). At pH_i > pH_PZC, deprotonation of surface functional groups, increased concentration of negatively charged arsenic species and competition of hydroxyl ion at higher pH contributed to enhanced repulsion at the boundary layer of the adsorbent interface, i.e. lower adsorption efficiency is a consequence.

Simultaneous determination of arsenic and iron by ICP-MS showed that at pH < 5 significant dissolution of MG precipitate was noticed. Good adsorption capacity and noticeable iron dissolution at pH < 5 dictated that selection of pH 6.0 was the best choice, and this pH was used in subsequent adsorption experiments.

The state of adsorption equilibrium was analyzed by fitting experimental data with various isotherm models (Foo and Hameed, 2010) by applying nonlinear least squares fitting to adsorption data. Statistical criteria (Section 2.11) were used to evaluate the model fitting of adsorption data. Results indicated that the good fitting was obtained by using Langmuir, Eq. (S2), and Freundlich, Eq. (S3), isotherm models.

Adsorption results presented in Fig. 8(b) and (c) and Table 2 showed that good fitting of the experimental data with Freundlich isotherm indicates multilayer adsorption on heterogeneous surfaces. Significantly higher adsorption capacity for NC-MA/L-MG (85.3 mg g⁻¹), in comparison with MC-O/L-MG (18.5 mg g⁻¹), was obtained by using Langmuir model (Table S2). The favorable textural parameters of NC-MA/L-MG contribute to higher arsenate uptake, i.e. larger number of available adsorptions sites at nanocomposite surface (Table 1). Somewhat lower capacity was obtained with NC-MA/L-MG and MC-O/L-MG in a processes of As(III) removal (Table S3), 68.2 and 17.8 mg g⁻¹, respectively. This result is indicated on a wider applicability of synthesized adsorbents. Also, comparative adsorption study of arsenic removal by using NC-MA/L showed that 15.4 mg g⁻¹ and 4.8 mg g⁻¹ were found for As(V) and As(III) at pH 6.0 ± 0.1, respectively. Even lower capacities were obtained for MC-O/L (<6 mg g⁻¹ for As(V) removal). These results clearly indicate that validity of the modification of NC-MA/L and MC-O/L with MG was justified by production of high performance NC-MA/L-MG and MC-O/L-MG adsorbents. Thermodynamic parameters are calculated, according to Eqs. (S4) and (S5) (Vuković et al., 2011), based on adsorption studies performed at different temperatures and calculated K_L Langmuir constant (Table S2), and results are given in Table 2.

The negative value of Gibbs free energy together with positive enthalpy and entropy changes indicates spontaneity and endothermicity of adsorption processes (Table 2). The positive ΔH° confirms favorable adsorption at higher temperatures. It means that disruption of water hydration shells and transport of arsenic anionic species through the bulk solution, within the pores and through the surface boundary layer are more intensive process at higher temperatures. The transport of exchangeable arsenic oxyanions to adsorption site, and number of (exchangeable) water molecules were released into bulk solution contributing to the increase in entropy change (Markovski et al., 2014a, 2014b). At steady-state condition randomness at adsorbent/solution interface is increased due to different intermolecular interactions which contribute to positive entropy change, i.e. adsorption is an entropy-driven process. All of these elementary processes during equilibration contribute to positive enthalpy change. Thermodynamic parameters of both As(V) (Table 2) and As(III) (Table S3) were similar to ones given by Liu et al. (2015).

Preliminary evaluation of the potential of use of both adsorbents NC-MA/L-MG and MC-O/L-MG for arsenic removal from real water samples, contaminated natural water from the area of the city of Zrenjanin was used (located in Vojvodina, Serbia). Sample used in this work was collected at the point before any purification treatment. Total arsenic concentration, determined by ICP-MS, was found to be 106 μg L⁻¹. Other ionic species of interest present in real water sample are shown in Table 3.

The total arsenic concentration includes following arsenic species: inorganic, As(III) and As(V), and methylated organic arsenic species (monomethylarsenic acid, MMAs(V), and dimethylarsenic acid, DMAs(V)). Due to this it was necessary to perform preliminary determination of present arsenic species. In a previous report we developed simple and efficient method for separation and determination of inorganic arsenic (iAs) and organic arsenic (oAs) in natural water (Issa et al., 2011). Two types of resins were used for determination of arsenic species: a strong base anion exchange (SBAE) resin and HY–AgCl based on the activity of silver-chloride loaded on SBAE resin. The SBAE resin has been chosen as a selective resin for the separation of As(V) from As(III), and HY–AgCl for quantitative separation of molecular and ionic forms of iAs and oAs. Molecular form of As(III) that exists in the water at pH < 8.0 was not bonded with SBAE, which was convenient for direct determination of As(III) concentration in the effluent. Experiments performed in a column system filled with HY–AgCl showed that oAs species are not presented in a sample water used. Additional experiment performed in a column system filled with SBAE resin showed that As(III) was present at the level less than 5 μg L⁻¹. In order to perform As(III) oxidation air bubbling for two hours was performed, and determined As(V) in oxidized solution was found to be 104 μg L⁻¹.

Preliminary adsorption study performed with real water sample (Table 3), spiked with 10 mg L⁻¹ of As(V), showed 19% decrease in adsorption capacity for NC-MA/L-MG (69.1 mg g⁻¹), and 8% for MC-O/L-MG (17.0 mg g⁻¹) with respect to experiments performed under non-competitive condition. These results showed possible application of synthesized adsorbents, even in the presence of significant concentration of interfering ions in natural water, in a processes of As(V) removal from groundwater.

3.3.2 Adsorption kinetics in a non-competitive condition

Due to complexity of adsorption processes, i.e. contribution of mass transfer and chemical reaction, different kinetic models were used to fit experimental data: pseudo-first order, pseudo-second order (PSO), Roginsky-Zeldovich-Elovich, and second-order rate equation. Also diffusion models: liquid film linear driving force rate equation, liquid film diffusion mass transfer rate equation, homogeneous solid diffusion model, Weber-Morris model (WM), Dunwald-Wagner model, and double exponential model (Markovski et al., 2014a, 2014b) were used. PSO and W–M models showed the best fitting (Fig. 9 and Table 4). Additionally, result of comparative kinetic experiments with As(III), and PSO fitting showed higher rate constant, 0.762 ± 0.022 g mg⁻¹ min⁻¹ for MC-O/L-MG at 25 °C, which is an additional evidence of high performances of obtained material.

Close

Figure 9

Plot of arsenate adsorption on MG by using PSO (a) and WM (b) model at 25 °C (m/V = 100 mg L⁻¹, C_i = 0.1 mg L⁻¹, pH 6.0 ± 0.1), and diagram of activation energies E_a vs α c).

Export to PPT

The results obtained following the W–M model showed three successive linear steps for MC-O/L-MG, while two-line correlation lines were obtained for NC-MA/L-MG (Fig. 9). Fast kinetics in the first step was followed by medium adsorption rate in the second, and very slow attainment of equilibrium in the third step was found for MC-O/L-MG. Higher values of the rate constant, found for MC-O/L-MG (Table 4), reflect better diffusibility/easier accessibility of surface active sites by sorbate due to shortening the path of ions inside the adsorbent body.

It is well known that multi-step nature of adsorption processes could be affected by bulk diffusion, boundary layer diffusion, intra-particle diffusion and surface adsorption. The first linear steep part demonstrates external mass transfer from bulk solution to the most available adsorptive sites at outer adsorbent surface. The second and the third parts of the adsorption process significantly depend on adsorbent porosity, i.e. pore structure, geometry and network density. Due to the concentration gradient, the ions diffuse through bulk solution and treelike pore system extending into adsorbent interior to reach all available surface active sites (Sun et al., 2014; Yousif et al., 2016). The intra-particle resistance slows down adsorbate transport, i.e. net transport in a direction of variable time-dependent concentration gradient. At the final stage, the adsorption takes place at low rate until the saturation of all available surface sites is achieved.

The activation parameters: entropy of activation ΔS^#, Gibbs free energy of activation ΔG^# and enthalpy of activation ΔH^# were calculated from kinetic data, obtained at 25, 35, and 45 °C, by using Eqs. (S6)–(S9). The results are presented in Table 4.

The activation energies of 13.28 and 10.87 kJ mol⁻¹ for As(V) removal with NC-MA/L-MG and MC-O/L-MG, respectively, were obtained. These results are in agreement with similar result obtained for synthetic goethite (Lakshmipathiraj et al., 2006). Adsorbate transport is generally controlled by diffusive mass transfer, and thus rates of approaching to equilibrium usually increase with increase in temperature. The results show that the intra-particle diffusion is a rate-controlling step since the activation energy is low and within the range of 8–22 kJ mol⁻¹, characteristic value for diffusion-controlled processes such as ion-exchange/complexation (Haring, 1942).

The improvement of the analysis of kinetic data for whole period of adsorption was performed by applying model free kinetics of adsorption, i.e. determination of activation parameters by iso-conversional method. Model free kinetics was selected in order to get the most reliable system responses at appropriate period of time, i.e. at the beginning, middle and plateau (approaching to steady state condition) parts of the process. The adsorption rate can be described as the amount of adsorbed pollutant per unit of time by using dimensionless number α. It denotes time-dependent fraction of bounded arsenic, and could be described by Eqs. (S10)–(S13). A plot of −ln t versus 1/T at appropriates α yield E_a from the slope regardless of the kinetic/adsorption model was used. The reason for choosing the ‘model free’ method of kinetic analysis is based on the fact that any misidentification of the kinetic model has got a profound effect on the values obtained for the Arrhenius parameters. Sewry and Brown (2002) and Galwey (2003) have critically reviewed the use of various kinetic models and asserted that caution has to be exercised in choosing the right kinetic model. By applying the fitting of conversion-time data, using predetermined point for calculation of α value, obtained results are presented in Fig. 9(c).

From the slope of correlations Eq. (S12), at appropriate T and α, a series of activation energies data (E_a) were calculated, and results are presented in Fig. 9(c). The plots show that overall adsorption mechanism progressively changed from the initial point of the process to attainment of equilibrium. In general, E_a varies from 15 to 18 kJ mol⁻¹ for NC-MA/L-MG, and from 12 to 14 kJ mol⁻¹ for MC-O/L-MG. In the first part of adsorption (0 < α < 0.3), the favorable process was noticed for MG due to synergetic effect of geometry/surface group density. Afterward, a steady region (0.3 < α < 0.7) was attained, and followed by steep increases in the course of equilibration. Similar trend of low change of E_a, for both NC-MA/L-MG and MC-O/L-MG, was found for α < 0.7, and α > 0.7, and somewhat higher slope was found for NC-MA/L-MG. Generally, the change of Ea values (Fig. 9) indicates that adsorption is diffusion controlled process, and it is in accordance with the concept of variation of E_a in a real system due to multi-step nature of most processes encountered in practice. The time-dependent change in kinetics is attributed to textural properties, material structure/defects at exposed plane, steric hindrances and electrostatic attraction/repulsion between adsorbed/released molecules/ionic species. The significance of external/internal mass transport and diffusion inside the pores was indicated (Markovski et al., 2014a, 2014b). The obtained activation energy for the As(V) removal with vacuum and freeze dried goethite modified macroporous resin was 18.63 and 20.60 kJ mol⁻¹ (Taleb et al., 2015), which indicates that similar phenomena dictate diffusional transport inside morphologically/chemically similar porous system.

3.3.3 Comparison of kinetic/adsorption data of magnetite based adsorbent

Additionally, performance of synthesized adsorbents was evaluated from comparison with other magnetite based adsorbents. Among the large numbers of cellulose based adsorbents synthesized up to date, most of them are claimed and considered to be the best candidates to feasible and applicable for effective arsenic removal. The adequate properties of cellulose based adsorbents, with respect to established criteria in the areas of environmental protection, mainly arise from their low cost and adsorption performances. The literature survey, shown in Table 5, of the data on adsorption kinetics and maximum adsorption capacity of the materials of interest is presented.

The diversity of experimental condition makes impossible convenient overview/comparison of the literature and presented adsorption data. General consideration of the relation of adsorbent performances/textural properties indicates positive dependence of increased surface area with respect to obtained adsorption capacity. According to the results presented (Table 5), the q_max was found in a wide range between 2.46 mg g⁻¹ for Magnetite/Cellulose fiber and 108.7 mg g⁻¹ for Amino/G3MA/TiO₂/Cellulose. The q_max obtained from the Langmuir model for the NC-MA/L-MG was found to be 85.3 mg g⁻¹ at 25 °C, and lower for the MC-O/L-MG, 18.5 mg g⁻¹ was found, which is comparable to the literature data. It means that synthesized high performance adsorbents offer an acceptable alternative to those published in the literature or eventually used in a real processes, and production of adsorbents based on presented material offers many benefits to existing ones. Fast attainment of adsorption equilibrium, i.e., kinetic rate of arsenic adsorption in the presence of the MC-O/L-MG mainly exceeds the values found in the literature (Table 5), and capacity of NC-MA/L-MG is similar to those found in the literature. Obtained results are in accordance with general experienced rules on relationship between textural data and adsorbent performance, i.e. kinetic and adsorption data. The developed fibers integrated mechanical integrity (stability), hydraulic properties and arsenic selectivity. Despite lower adsorption capacity, another factor contributes to substantial improvement of its performance, and namely the results of adsorption kinetics open an alternative for the use of MC-O/L-MG exchange fibers relative to NC-MA/L-MG counterparts. The results presented point out to the advantage of application of micro-sized adsorbent which contributes to adsorption performances of chemically identical but physically different materials NC-MA/L-MG and MC-O/L-MG.

3.3.4 Competitive kinetic study

Except high capacity and fast adsorption rate of arsenic removal, the synthesized adsorbents should possess high affinity/selectivity with respect to selected pollutant. Similar to non-competitive kinetic study, the data obtained in competitive study of arsenate removal in the presence of single interfering ion: SiO₄⁴⁻, CrO₄²⁻ and PO₄³⁻ ion were processed by using PSO equation. The other kinetic models were of limited validity to be applied for description of the competitive kinetic study. Therefore, PSO model was used and results of fitting, obtained from plots q_t − t and t/q_t − t, are given in Table 6.

On the other hand, calculated values show good agreement with the experimental data for PSO kinetics which means that sorption depends on the both concentration of arsenic and adsorbent surface functionalities. Single competitive As(V) adsorption showed the highest influence of phosphate (11.5% decrease of K₂) for MG, 25.9% for NC-MA/L-MG and 9.6% for MC-O/L-MG. Similar trend but of lower extent was found for chromate and the lowest influence of silicate was noticed. Largest adverse effect showed phosphate due to similar chemistry with arsenate. The acid dissociation constants of phosphoric acid are pKa₁ = 2.16, pKa₂ = 7.21 and pKa₃ = 12.32, which indicates that at pH ∼ 6 monovalent phosphate ion, H₂PO₄⁻, is a dominant competitive species (Awual et al., 2011). Selection of pH 6 in adsorption study was optimal due to an increase in pH, hydroxide concentration increases so that interference of hydroxide with the hardest Lewis acids anion in water increases resulting in decrease in arsenate and phosphate adsorption and phosphate showed lower adsorption capacity at pH 6 than at higher pH on magnetite nanoparticles (Abo Markeb et al., 2016). Among the other interfering ions commonly present in natural water, sulfate, as divalent ions, shows greater competition through enhanced electrostatic interaction. Selection of pH 6, i.e. pH ≈ pH_PZC, contributes to lower extent of electrostatic interaction of all ions with MG surface. Therefore, the main benefit of the designed system adsorbent/adsorption condition arises from achievement of desired levels of arsenate removal in the presence of interfering ions. Silicate showed low effect which was less than 8%.

Except this, experiments performed with NC-MA/L-MG in the presence of different concentrations of KCl, KNO₃ (1 and 5 mg L⁻¹), arsenic removal were slightly enhanced. Due to bonding of negatively charged arsenic species, concentration of negative charge is increased in a electrical double layer (Grossl et al., 1997), and at higher ionic strength of solution, increased concentration of counter cations could compensate the surface negative charges. Observed fact that increased ionic strength causes increased arsenate uptake indicating that formation of inner-sphere complexes (Markovski et al., 2014a, 2014b) is the main adsorption mechanism. Additionally, the low influence of interfering ions (Table 6) and pH_PZC decreases after adsorption (Table 1) are the properties indicative for the formation of inner-sphere surface complexes. Ligand exchange is a final/crucial step at surface which occurs through formation of monodentate complex, while the next slow step represents a second ligand exchange resulting in the formation of an inner-sphere bidentate binuclear complex (Grossl et al., 1997). Formation of inner-sphere complexes is a process of highest probability, and other types of interaction between As species and protonated hydroxyl groups by forming outer-sphere surface complexation could be operative at lower pH (Liu et al., 2015).

Additionally, competitive kinetic experiments, performed with NC-MA/L-MG and MC-O/L-MG and natural water sample (Table 3), were carried out to find out the level of the efficiency of total arsenic removal from natural water sample. At 100 mg L⁻¹ of NC-MA/L-MG 70% of arsenic was removed for 30 min and more than 99% after 60 min. Fast arsenic uptake was obtained with MC-O/L-MG adsorbent: 70% removal was achieved for 6 min and after 14 min the total arsenic concentration was below the WHO limit prescribed for arsenic in drinking water. This results showed that at even high concentration of interfering ion (Table 3) fast removal of As(V), present at low level, could be efficiently removed from natural water.

3.3.5 Regeneration versus safety

Experiments related to regeneration study with respect to both NC-MA/L-MG and MC-O/L-MG adsorbents, and related disposal issue of spent alkali desorption solution are given in Supplementary data.

3.3.6 Future challenges

The adsorption performances, comparable/higher than that of the most common referenced material, observed for well-dispersed/manipulative NC-MA/L-MG and MC-O/L-MG adsorbents result from the contribution of two factors: (i) design of surface modification with branched organic structure with number of amino terminal groups and (ii) development of controllable method for magnetite precipitation. The main goal of presented study was devoted to understanding of scientific principles related to influences of the parent media size, NC versus MC, on the adsorption properties of produced adsorbents. Also, investigation on the influences of chemical structure/properties of the modified precursor surface on the optimization of the method for magnetite precipitation was an additional goal of this study. Obviously, crystal structure of precipitated magnetite, exposed surface and crystal faces, had pronounced influence on adsorbent properties and strongly depends on experimental condition and NC/MC surface properties. It was shown that in a moderately acidic/neutral condition arsenic species are strongly bonded to iron oxide surface. EXAFS analysis suggested predominant formation of bidentate binuclear corner-sharing complexes (²C) for As(V), with dominant complexes formed between AsO₄³⁻ tetrahedron and two FeO₆ octahedra on {1 0 0} surfaces of magnetite (Liu et al., 2015). Also, calculation using density functional theory (DFT) predicted the relative energies and geometries of surface complexes using density functional theory showed higher stability of ²C complex for 55 kJ mol⁻¹ in relation to ²E edge-sharing bidentate complex (Sherman and Randall, 2003). According to this, good adsorption performances of the novel adsorbents were related to proper NC and MC modification, and controllable MG precipitation which causes beneficial facet evolution at MG surface. As a result, high affinity and efficient arsenate removal from water, as well as acceptable kinetics with possibility for adsorbent use in the processes of natural water remediation even in the presence of competing ions were obtained.

Also, exceptional adsorption performances and magnetic properties of MG nanoparticles, precipitated on modified NC and MC surfaces, provide an opportunity for practical application and easy separation from aqueous solution by applying external magnetic field. Except this, results of reusability study showed low decrease in adsorption efficiency after five adsorption/desorption cycles (Supplementary data; Section 3.3.5). No measurable iron ion was detected in water during repeated cycles of adsorption. Therefore, MG loaded NC-MA/L and MC-O/L adsorbents could be effectively used for long time of exploitation. Using a proper selection of technology for treatment of spent waste alkali concentrated arsenic provides safe technology for arsenic removal (Supplementary data; Section 3.3.5). In that way, the three main goals were realized by understanding/explaining of scientific principles related to the influences of parent media size, NC versus MC, chemical structure/properties of organic functionalities bonded on NC and MC surface, and MG precipitation method on adsorption performances of novel synthesized adsorbents. Additionally, a number of idea/strategy were designed based on presented results/conclusions, and future work will be focused on development/production of woven nano/-micro cellulose MG modified hybrid textile materials.