Adsorption of Zerovalent Iron Nanoparticles in the Inorganic Fraction of Volcanic Soils

Zerovalent (ZVI) iron has been used in environmental remediation, but the environmental fate of its nanoparticles (NZVI) is hard to find. The present study aimed to characterise the adsorption processes of commercial ZVI in the inorganic fraction of volcanic soils, contributing to the knowledge of the adsorption process on the whole soil. Volcanic soil samples were obtained by calcination method and characterised. A nano-ZVI stable suspension is prepared and characterised to know its size-particle distribution. Adsorption points (kinetical and isothermal) were obtained by adapting the procedure for homogeneous analyte adsorption. Kinetical, solute transport and isothermal adsorption and desorption models were adjusted to obtain more conclusive information about the process. Analysis of the results revealed that adsorption is a fast process. The inorganic fraction of an Andisol presented pH-dependant surface charge, requiring highly consistent control of the working pH to obtain proper adsorption points. Isothermal adsorption indicates that a combination of physical and chemical mechanisms contributes to the NZVI adsorption process. Desorption studies suggested chemisorption on specific surface-active sites of adsorbents, with the inorganic fraction of soil irreversibly retaining most of the NZVI particles. NZVI become irreversibly adsorbed over the inorganic fraction of soils, extending particle reactivity through time, affecting the stability of organic matter and microbial communities. In turn, this increases the possibility for plant uptake. Nevertheless, there is a low risk for the transport of NZVI through soil profiles or co-transport of other components that becomes adsorbed over iron, lowering the risk of aquifer pollution.


Introduction
Zerovalent iron (ZVI), including zerovalent iron nanoparticles (NZVI), has been used for environmental remediation of different pollutants present in soils (Luna et al. 2015;Orozco et al. 2015;Guo et al. 2021;Qiao et al. 2020). When ZVI is placed in contact with soil, pollutants become oxidised through a Fenton reaction mechanism leaving iron oxide as an innocuous by-product. ZVI slurry comprised of micro and nanoparticles in liquid suspension represents a new generation of environmental remediation technologies (Brumovsky et al. 2021;Hosseini and Tosco 2015;Luna et al. 2015;Mondino et al. 2020;Tosco et al. 2014). Because of its reduced size, when NZVI are injected into soils have higher reactivity and mobility than conventional iron powder, contributing to their broad use for the decontamination of groundwater and soil remediation (Yang et al. 2018). After remediation, the excess of NZVI may generate problems in the ecosystem. For instance, some studies have reported the negative impact of NZVI on soil microbial community structures (Lefevre et al. 2016). Due to the complexity of soils, there is a significant knowledge gap regarding the fate of NZVI in soils; specifically, more information is needed regarding the adsorption and mobility processes (Simonin & Richaume 2015).
In the literature, the few studies of NZVI adsorption or transport processes generally employ soil models. Zhang et al. reported outcomes of NZVI over quartz sand and iron oxide coated quartz sand (M. Zhang et al. 2017). Others studied NZVI transport through model adsorbents of simple composition (Ibrahim et al. 2019;Zhang et al. 2017). Moreover, Micic et al. were able to describe the effect of Nahumate on the mobility enhancement of poly(acrylic acid)coated ZVI nanoparticles through standard sand mediums, but this effect was not observed when using natural minerals as adsorbents due to the surface heterogeneity (Micic et al., 2017). While these studies have set the ground for understanding the environmental fate of NZVI on simple-composition adsorbent, none has still analysed the adsorption process of NZVI on complex soils. NZVI has been employed to remediate volcanic soils as Nitisols with banana plantations (Mouvet et al. 2020); however, there are no studies regarding the fate of the NZVI excess in these matrixes.
Regarding the effect of NZVI on plants, results are contradictory depending mostly on particle concentration and plant species. For instance, Martinez-Fernandez and Komarek tested the distinct effect of Fe 2 O 3 and ZVI nanoparticles on tomatoes plants, with no harmful effects from zerovalent particles found (Martinez-Fernandez & Komarek 2016). In addition, Teodoro et al. reported on the effect of NZVI on growth of Agrostis capillaris and Festuca rubra, evidencing that the amendments had increased shoot growth and elongation rate (Teodoro et al. 2020). On the other hand, Wang et al. reported the effect of NZVI on rice seedling with growth inhibition at high nanoparticle concentration (Wang et al. 2016). Also, Sun et al. presented that starch-stabilised NZVI decreased germination rate, increased iron uptake rate and altered nutritional balance of mung bean seeds up to phytotoxic effects depending on nanoparticle concentration (Sun et al. 2019). Nevertheless, it is important to point out that all these studies are made on hydroponical conditions, which does not represent the effect of ZVI nanoparticles when immersed in soil-water matrix. The main reason for the lack of studies regarding the effects and fate of NZVI particles using full complex soils, including volcanic soil, is the high complexity of these systems acting as adsorbent surfaces hindering the obtainment of consented conclusions. For instance, some volcanic soil orders possess considerable organic matter percentages and important contents of shortly ordered minerals (Escudey et al. 2004) with particular adsorption behaviour, hindering repeatable results. Moreover, studying the adsorption of nanoparticles is complicated because of solution parameters on aggregation and reactivity of suspended nanoparticles, such as pH and ionic strength (Flory et al. 2013;Mpinda Tushiminine 2014). Thus, further adsorption studies using NZVI and natural volcanic soils are required to fully understand their fate and consequences in the environment. A strategic approach to assess the complete full soil adsorption process could be to determine the relative contribution from volcanic soil constituents to the adsorption process, employing soil fractions.
In the present report, we present the adsorption of a commercial zerovalent iron particle slurry, containing NZVI, over the inorganic fraction of two Chilean volcanic soils. The use soil fractions in NZVI adsorption experiments would permit to establish the relative importance of soil components on the NZVI-interaction until reaching to an appropriate interpretation of NZVI behaviour in full soils. This information is also important to eventually assess the potential environmental fate ZVI in nature. For this study, the inorganic fraction of soil was obtained through the calcination method, removing the organic layer and exposing the mineral fraction (Escudey et al. 1999). Adsorption kinetics, solute transport mechanism, adsorption and desorption isotherm curves were obtained, modelled and discussed, permitting evaluation of the NZVI fate in volcanic soils.
Nitrogen contents and organic carbon (OC%) was quantified using the Walkley-Black method (Allison 1965). Organic phosphorus (OP) was determined using sequential extraction method adapted for volcanic soils (Steward & Oades 1972). The amounts of Fe 2 O 3 , Al 2 O 3 and SiO 2 (wt%) were obtained after chemical analysis by complete dissolution of samples and quantification using ICP-OES spectrometer. Soil pH was measured at the supernatant of the mixture soil:solution with ratio equal to 1:2.5 in mass. The solution corresponded to distilled water or 1.0 mol L −1 KCl. Conductivity was measured at the supernatant in the same soil:solution mixture with distilled water after a 10 min centrifugation (8000 rpm) step (Sadzawka 1991). Cation exchange capacity was calculated as the sum of exchangeable bases (CH 3 COONH 4 extract and determination by atomic absorption spectroscopy) and extractable acidity (BaCl 2 -triethanolammine extract and determination by HCl-titration) (Sadzawka, 1991).
Surface charge was determined through electrophoretic mobility measurements. Samples were suspended in 10 −3 M NaCl to perform electrophoretic measurements in a zeta metre to establish the isoelectric point (IEP) and pH was adjusted with 10 −2 M HCl or NaOH. The zeta potential (ZP) was calculated using the Helmholtz-Smoluchowski equation (Hunter 1981).
External specific surface area (ESSA) was determined by the adsorption-desorption isotherms of N 2 at − 195 °C (77 K), changing the relative pressure (P/Po) of the gas and recording the volume adsorbed on the solid's surface. ESSA was calculated from the amount of N 2 adsorbed employing the Brunauer-Emmett-Teller or BET equation (Brunauer et al. 1938).
Prior to adsorption experiments, soil samples were sieved to collect the < 2 mm size particles and calcined at 550 °C for 3 h. Properties of soil samples are presented in Table 1.
Collipulli and Ralun exhibit differences in soil texture. Both soils possess comparable amounts of silt at about 40%. But the sand-and clay-size fractions behave in an inverse fashion on both soils, with ca. 15% and 45% for Collipulli soil and vice versa for Ralun. Nitrogen (N%), organic carbon (OC%) and organic phosphorus contents (OP) decreased dramatically after heat treatment on both soils. Metallic oxides contents (wt%) Fe 2 O 3 , Al 2 O 3 and SiO 2 increased in proportion after calcination. Soil pH (in water or KCl) got elevated after organic matter destruction especially for Ralun, the Andisol. This soil possesses higher amounts of organic matter, and its acid pH is highly dependent on the acid-base reactions from organic moieties located at the surface. Therefore, calcination would intensely shift Ralun pH, while for Collipulli this effect is less important. Electrical conductivity (EC) increases on both soil samples after calcination, due to concentration of soluble salts from the mass-loss generated by the heat treatment. Determination of the isoelectric point (IEP) of the resulting calcined soils is presented in Figure S1. Extrapolation of the curve until zeta potential is 0 yields an IEP around pH 2.2 for Collipulli. On the other hand, calcined Ralun samples an IEP of 7.4. Permanent and variable surface charge calcined soil samples are mainly explained by their mineralogy (Table 1). Thus, permanent and variable surface charge on Collipulli and Ralun calcined samples, respectively, are mainly explained by their mineralogy (Table 1); while permanent charge mineral kaolinite is highly abundant on Collipulli soils, a variable surface charge mineral as allophane dominates Ralun mineralogy. The IEP for Ralun got dramatically elevated after calcination, a behaviour not observed on Collipulli sample. Regarding external specific surface area (ESSA), there are great differences between soil series, up to ca. 30 times greater for calcined Collipulli samples over Ralun, which can be explained by its higher clay content. Therefore, calcined soil samples display different features that could determine distinct NZVI adsorption behaviours.

Zerovalent Iron
A sample of zerovalent iron (ZVI) slurry for remediation of contaminated soil and ground water was provided by Regenesis (Regenesis 2020). This product corresponds to a colloidal suspension of metallic sulphidated ZVI, dark grey colour and with slight odour. Its pH is typically between 7 and 9 and it has a density of 1.79 kg L −1 . The iron concentration was determined as 425.5 mg mL −1 . The ZVI slurry was used directly from original container to prepare a stable suspension of NZVI particles as explained on Fig. 1a. As presented in Fig. 1a, a stable NZVI suspension was prepared using 500 μL from original slurry and diluted with 50 mL of double distilled water. To enhance particle dispersion, this solution was sonicated by means of an ultrasound probe at 100% amplitude for 10 min. Afterwards, the suspension was vacuum filtered using Whatman® nitrocellulose membranes with 5.0 and 1.0 µm pore size. Then, the filtered solution was diluted to the desired concentration by checking its absorbance at 500 nm wavelength by means of a Thermo Fisher Scientific Genesys 30 vis/UV spectrophotometer. In this report, absorbance values provided by spectrophotometer are interpreted as scattered light (S) by the nanoparticle suspension. This solution was prepared daily to ensure its stability and repeatability of results. Particle size distribution of this solution was controlled using a Malvern Zetasizer Nano ZS dynamic light scattering (DLS) system. This device uses classical single-scattering DLS methods to calculate particle sizes in a range from 0.3 nm to 10 µm with 633 nm laser and detects the backscattered light at 173°. This device also features a temperature-controlled holder for a cuvette. According to the DLS report ( Figure S2), 97.6% and 2.4% of particles have 47.4 and 272.2 nm of diameter, respectively. The concentration of the NZVI suspension was determined by gravimetric analysis. Solutions were prepared to fixed S values (0.300, 0.600, 0.900 and 1.200) at 500 nm wavelength by means of a vis/UV spectrophotometer. Aliquots of adequate volumes for each solution were transferred in triplicate to tared beakers and placed in lab-oven at 50 °C until reaching constant mass. Beakers were weighted to obtain the mass from dry NZVI particles, and a calibration curve was constructed to check the dependence between S and mass concentration (mg L −1 ) of NZVI suspension. A linear relationship between ZVI concentration and scattered light at 500 nm is observed ( Figure S2). Thus, prepared NZVI suspension possesses most of its particles below 50 nm and there is a linear relationship between ZVI nanoparticle concentration and scattered light at 500 nm.

Experimental Points Construction
A NZVI stable suspension of appropriate concentration was employed for adsorption experiments following the procedure illustrated in Fig. 1b. A set of five 50 mL centrifuge tubes was prepared for each point to assure repeatability: three with NZVI suspension and soil; one with NZVI suspension and one with soil and water to control particle liberation from adsorbent. After reaching equilibrium, S was measured at 500 nm. The S values for each experimental point correspond to the average between the three soil + NZVI tubes, minus the S value from the soil + H 2 O Adsorption Kinetics Experimental points were made at different times (0-10 min) until reaching stable S values through time. Equilibration time was determined for each adsorbent at the conclusion of the experiment. Constructed curves were adjusted using adsorption kinetic models (Table 2).

Adsorption Isotherms
For Collipulli calcined soil, points were constructed using an increasing concentration of ZVI suspension (0-500 mg mL −1 ) using the procedure outlined in 'Section 2.3.1'. Experimental points were made using the equilibration time determined after adsorption kinetics. Constructed curves were adjusted using adsorption isotherm models (Table 2).
Due to significant pH variations upon mixing Ralun calcined soil samples and NZVI suspensions of different concentrations, pH adjustment was required. To adjust the pH, prior to isotherm adsorption experiments Ralun soil was treated with HCl 0.01 M. Three portions of 0.2 mL of the acid were added to 5 mL of water and 1.2 g of Ralun calcined soil. The tubes were placed on an orbital shaker for 30 min at 50 rpm after every addition. Then NZVI suspensions were added to reach a total volume of 12 mL using same concentration range (0-500 mg mL −1 ) as the procedure from 'Section 2.3.1'. The pH of the mixtures was measured at the conclusion of the procedure.

Desorption Isotherms
Once the adsorption process had ended and all supernatant was removed, 12 mL of water was added to remnant adsorbent. The mixtures were subsequently agitated in an orbital shaker for 10 min at 65 rpm. The supernatants were filtered through a 1.0 µm pore-size filter and the quantification of NZVI was carried out in a UV/vis spectrophotometer at 500 nm.

Adsorption Kinetics
Sorption of NZVI nanoparticles as a function of time is presented in Fig. 2. Saturation was reached during the first 10 min of the experiment. After each kinetic experimental point, the pH of the supernatant in contact with Ralun and Collipulli calcined samples was measured, with average values of 5.2 ± 0.1 and 6.3 ± 0.2, respectively. Figure 2 shows sorption kinetics and kinetic model adjustment. The maximum experimental NZVI absorbed at the end of the kinetic adsorption process (C m ) was 0.729 and 0.797 mg g −1 for Ralun and Collipulli, respectively. Experimental data were fitted to three simple kinetic models: hyperbolic, pseudo-first (PFO) and pseudo-second order (PSO) models; the resulting kinetic parameters are presented in Table 3. The PFO model did not fit either of the experimental data sets with R 2 values of 0.448 and 0.479 for Collipulli and Ralun, respectively. Calculated maximum concentration (C m-cal ) for the PSO model agrees with C m and this model yielded the highest R 2 parameter, indicating that PSO best describes the adsorption kinetics.
Data were fitted to several solute transport mechanism models: Elovich, intraparticle diffusion and their dimensionless versions. The resulting parameters are presented in Table 3.

Solute Transport Mechanism
Values obtained from R 2 for the Elovich equation are close to 1, indicating that this model describes the experimental data for both soils (Fig. 3a). The initial rate constant α is higher for Collipulli than for Ralun. The number of surface sites given by β parameter was similar on both soils.
From the Elovich linear equation (Figure S3), the slope (1/β) is considered to be the sorption rate as a function of time during the slow stage of the reaction, while the intercept (1/β)ln(αβ) corresponds to the amount absorbed during the initial equilibrium stage (fast phase reaction). As established by the analysis of the Elovich parameters described above, intercepts 0.461 and 0.357 are calculated for Collipulli and Ralun soil, respectively. Figure 3b shows the plot for the intraparticle diffusion model for the adsorption of NZVI on both soils. The intraparticle diffusion model developed by Weber and Morris could present multilinearity indicating simultaneous occurrence of several adsorption stages; in this adsorption process, two stages can be identified. The intraparticle diffusion constants can be calculated using Eq. (5). Table 3 shows the intraparticle diffusion constants (k int1 , k int2 ), the Table 2 Kinetic and adsorption models Name Equation

Parameters
Theoretical and empirical description

Kinetic models
Pseudo-second order model k 2 is the rate constant of second order adsorption (mg g −1 s −1 ) The pseudo-second order kinetic model postulates that solute adsorption occurs on two available surface sites of the adsorbent (Evangelou 1998) Hyperbolic model

B (s) in an empirical constant
This model uses hyperbola function to fit kinetic adsorption data (Fernández-Bayo et al., 2008) Kinetic models involving solute transport mechanisms Elovich equation α is the initial sorption rate (mg g −1 s −1 ), and β (g mg −1 ) is related to the extent of surface coverage and activation energy for chemisorption This model considers that the adsorption kinetic takes place in two stages: a fast initial reaction associated with the movement of the adsorbate to the most accessible parts of the adsorbent followed by a slower reaction due to particle diffusion in and out of the adsorbent micropores (Önal 2006) Dimensionless Elovich obtained. This allows to describe adsorption process mechanism in stages defined by zones (Wu et al. 2009a)  Theoretical and empirical description Intraparticle diffusion dimensionless The Langmuir model assumes that the adsorption of the solute occurs at specific homogenous sites on the adsorbent (Escudey et al. 2004) Sips or Langmuir-Freundlich model C constants and the correlation coefficient (R 2 ). Finally, Figure S4 presents dimensionless Elovich and intraparticle diffusion adjustment on experimental data.

Adsorption Isotherms
Sorption isotherms were carried out at the equilibration time determined after the kinetics adsorption studies. When mixing soils with increasing concentrations of NZVI, the mixtures produced extremely high C s variations between on Ralun calcined samples (Fig. 4).
The above results demonstrated that greater, on Ralun calcined soil samples, control of the working pH was necessary to obtain adsorption isotherms that can be described by models. On the other hand, pH pre-treatment procedure was not necessary to obtain adsorption isotherms on Collipulli calcined samples. Figure 5 shows the adsorption isotherms for both soils. After measuring S from adsorption curve points, an average pH = 5.4 ± 0.4 for Collipulli and 6.3 ± 0.3 for Ralun was determined for each curve.
For Collipulli and Ralun soil, the maximum experimental amount of adsorbed NZVI corresponded to 2.127 and 2.205 mg mL −1 , respectively. Data analysis by three isotherm models is presented in Table 4.
The best fit was obtained with the Langmuir-Freundlich model for both soils, indicating that NZVI adsorption process is formed by the contribution of chemical and physical interactions. Calculated maximum concentration was higher over Collipulli than Ralun soil, while K d values for Collipulli and Ralun are 121.4 and 5996, respectively.

Desorption Isotherms
The corresponding desorption isotherms were obtained, and data were analysed with the Langmuir-Freundlich model ( Figure S5). After measuring S from desorption curve points, an average pH = 5.9 ± 0.2 for Collipulli and 6.6 ± 0.1 for Ralun was determined for each curve, which are consistent with the pH values of adsorption isotherms.
Desorption data was fitted using Langmuir-Freundlich model adjustment. Table 4 displays desorption percentage (D%) and hysteresis index (HI), calculated using C m-cal values for adsorption and desorption processes (Ma et al. 1993). Parameters indicate that NZVI adsorption is an irreversible process, more pronounced over Collipulli than Ralun soil. Desorption equilibrium constant, K d , is about 2000 times higher for Ralun. This agrees well with the D% for both soils. Finally, both soils present negative hysteresis index (HI) suggesting a chemically irreversible process.

Discussion
Adsorption kinetics results determine the equilibrium time, necessary for the construction of isothermal adsorption points, and are used to fit the kinetic and solute transport mechanism models to provide a more detailed description of the process. Equilibrium during absorption kinetics experiments was a fast process (Fig. 2), reaching maximum adsorption after the first 10 min on both adsorbents. Regarding kinetic model adjustment, the PFO model did not fit either of the experimental data sets (Fig. 2) evidenced by low correlation coefficients (Table 3). On the other hand, calculated maximum concentration (C m-cal ) for the PSO model agrees with C m and this model yielded the highest R 2 parameter, indicating that PSO best describes the adsorption kinetics. Accordingly, this implies that the interaction occurs at two sites of union and might suggest that NZVI interacts by chemisorption with the calcined soil. Solute transport mechanism model adjustment of kinetic data enables to identify the steps involved in the NZVI transport process. For instance, experimental data for calcined Ralun soils are well-described by the Elovich equation (Fig. 3a), including linear regression analysis ( Figure S3), meaning heterogeneous surface characteristics. Cáceres-Jensen et al. (2013) observed that high organic matter content can confer some characteristics of energetically homogeneous surfaces on Andisol full soil samples, which is lost after Ralun calcination due to degradation of organic matter (Table 1). The higher α for Collipulli over Ralun (Table 3) implies that, if part of the process is controlled by chemisorption, this possibly occurs much faster over Collipulli than Ralun surfaces at the beginning of the process. The β parameter was similar on both soils; thus, despite the process initially occurring faster over Collipulli, the amount of surface covered by the adsorbate was most likely similar on both calcined soils. From the Elovich linear equation fitting ( Figure S3), intercepts indicate that while adsorption is relatively rapid, it is not instantaneous and takes place in two consecutive stages. Accordingly, as Wu et al. (2009a) proposed after dimensionless Elovich analysis ( Figure S4a), both soils are found in zone III, which indicates that the adsorption increases rapidly at the beginning of the process. Going deep in NZVI transport mechanism description, the intraparticle diffusion model permits adsorption stages definition which has been applied for modelling nanoparticle adsorption processes before (Hamouda & Abhishek 2019). Accordingly, NZVI adsorption process takes place in two stages (Fig. 3b). Model adjustment indicates that another process besides intraparticle diffusion takes place in the adsorption of NZVI. As has been described for other solute-sorbents systems (Cáceres-Jensen et al. 2013), if the external resistance to mass transfer surrounding is discarded, the first line will describe the gradual adsorption stage and depicts macropore diffusion, while the second line represents micropore diffusion until equilibrium achievement (Önal 2006). The calculated intraparticle diffusion constants reveal a noticeable change in slope for both soils (Table 3). Initially, the NZVI were adsorbed by the external macropore structure of soil, and consequently the sorption rate was very high. After saturation of the macroporous structure, the NZVI diffused in the internal surface pores within the soil particle (micropore structure) and were adsorbed. Later, the diffusion resistance increased, which caused the diffusion rate to decrease. When the NZVI concentration was decreased, the diffusion rate became lower, and the diffusion processes reached the final equilibrium stage. The intercept C (Table 3) is proportional to the extent of the boundary layer thickness; that is, the larger the intercept the greater boundary layer effect. Positive intercepts indicate a rapid sorption in a short time on adsorbents with a wide distribution of pore sizes (Wu et al. 2009b). This intercept decreases with increasing surface heterogeneity of soils (Cáceres-Jensen et al. 2013). The intraparticle diffusion dimensionless model ( Figure S4b) displays R i that falls in zone III for both soils, indicating strong initial adsorption. This agrees with assumptions of the pseudo-second order, Elovich and intraparticle diffusion models, describing a fast adsorption process comprised of an initial fast equilibrium that takes place at the external micropore structure of the calcined soil. Equilibrium between solved solute solutions with increasing concentration and adsorbents is a common procedure used to evaluate isothermal adsorption in soils or its fractions (Evangelou 1998), achieving a thermodynamic description of the process. However, equilibria to evaluate adsorption of suspended particles in solution, as nanoparticles, are still a complex aspect for two main reasons: experimental and results-modelling. First, experimentally, compared to a solved solute system, nanoparticle suspension stability depends on their size (Cornelis et al. 2013), concentration (Ibrahim et al. 2019), pH (Flory et al. 2013; Mpinda Tushiminine 2014), among other parameters. Thus, special experimental procedures to ensure particle stability during the duration of the adsorption experiments are required ('Section 2.3.1'). Second, results-modelling for nanoparticle adsorption isotherms is a controversial topic. Two major theories are commonly used to model soil adsorption or sorption equilibrium processes: the Freundlich and the Langmuir approach (Evangelou 1998). While Freundlich model has no mechanistic interpretation, the Langmuir model was developed to describe vapour adsorption on a homogeneous surface. Considering the contrast between the complexity of nanoparticle suspension adsorption system, the simplicity of the Freundlich model interpretation and the specificity of the Langmuir model assumptions, the validity of the interpretation of adjusting these models to NZVI-adsorption should be matter of discussion. At a minimum, acknowledgement, and discussion of differences between systems is essential if model adjustments and their parameters are compared between different systems (particle composition, size and adsorbents characteristics, from same or different authors). For instance, at present report, during isotherm experimental point obtainment, when mixing soils with increasing concentrations of NZVI at first,  the mixtures produced extremely high C s variations between repetitions-particularly for calcined Ralun soil-precluding attainment of conclusive results (Fig. 4). Later, it was discovered that calcined Ralun soil-NZVI mixture provoked high system-pH variations. Considering that pH values were different at each experimental point of the isotherm and the variable surface charge of calcined Ralun soil ( Figure S1), the imprecise C s results are not surprising, preventing conventional isothermal construction. Such an unpredictable behaviour of variable surface charge adsorbents for NZVI has been reported by other authors. Micic et al. observed that the mobility poly(acrylic acid)-coated ZVI nanoparticles through columns filled with variable surface charge natural minerals (dolomite, muscovite) (Nosrati et al. 2009;Zhou et al. 2020) did not behaved as predictable than when using sand and glass beads as adsorbents (Micic et al. 2017). Therefore, a greater control of the working pH when using variable surface charge adsorbents as Ralun calcined samples is necessary to obtain isotherm points adequate for fitting; this may be carried out by the addition of HCl solution before adsorption point obtainment (Fig. 5). The effect of pH on the adsorption process is important as it could vary the number of binding sites on the adsorbent, especially on a variable surface charged adsorbent at calcined Ralun soil. Additionally, pH could influence loading of the NZVI and cause agglomeration between particles, which can be mistakenly interpreted as an adsorption phenomenon. Collipulli soil does not present this pH-mediated effect due to its marked permanent charge character ( Figure S1). Therefore, the pre-treatment HCl addition procedure was not necessary to obtain adsorption isotherms with Collipulli calcined soil.
Regarding isothermal model adjustment of experimental data, Langmuir-Freundlich model fitted best for both soils ( Table 4), suggesting that NZVI adsorption process does not correspond to a single type of process and consist of a combination of both chemical and physical process. Zhang et al. (2017) described that the adsorption of carboxymethyl cellulose stabilised NZVI over glass beads, sand, aluminiumoxide-and iron-oxide-coated sands obeyed the Langmuir model. A later work by Zhang et al. (2019) reported that adsorption of macromolecule-stabilised NZVI over Argosols and Cambosols was described by the Langmuir isothermal model. Thus, as anticipated, adsorption depends highly on particle and adsorbent nature. The isotherm produced by the Langmuir-Freundlich model for Collipulli soil corresponds to L-type according to the classification of isotherms by Giles et al. (1974). The curve slope steadily falls with rise in NZVI concentration as vacant sites become more difficult to find with the progressive covering of the soil surface. This describes a high affinity between NZVI and calcined Collipulli soil. On the other hand, the isotherm obtained for the Ralun soil is a S-type, which indicates that adsorption can occur through interactions between the NZVI particles and then over calcined Ralun soil surface adsorption sites (Giles et al. 1974). Calculated maximum concentration was higher over Collipulli than Ralun soil (Table 4), which may be in part explained by the different ESSA values measured (Table 1). K d values suggests that adsorption equilibrium is thermodynamically more stable over Ralun calcined soil, attributed to the lower crystalline degree of dominant minerals present this Andisol sample (Table 1). Therefore, NZVI isothermal adsorption over the inorganic fraction of the volcanic soils under study is a complex system that, depending on soil surface characteristics, requires monitoring and control of parameters such as pH to obtain repetitive results. The interaction between NZVI and calcined soils consist of a combination between physical and chemical processes that most likely depends on adsorbents surface area and mineralogy. To establish reversibility degree of the process more definitively, desorption isotherm analysis should be considered ( Figure S5). The difference between adsorbed and desorbed NZVI amounts indicates the existence of an irreversible adsorption attributed to specific interaction in high energy sites on calcined soils. This agrees with the pseudo-second order model that fitted best kinetic adsorption data which is typically associated with dominant chemical interactions. Moreover, according to the calculated maximum concentration desorbed (Table 4), NZVI are mostly retained by the absorbents; a more pronounced behaviour on Collipulli than Ralun calcined soil. The desorption equilibrium constant (Table 4), K d , is about 2000 times higher for Ralun, highlighting the strength of the bond formed between NZVI and Collipulli calcined soil. This behaviour is also corroborated by the D% for both soils. Finally, the highly negative hysteresis index (HI) suggests that adsorption over both calcined soils is essentially a chemically irreversible process and possibly dominated by electrostatic interactions. This observation is more prominent in Collipulli than Ralun soil. Thus, in mild conditions NZVI adsorbed will not be removed over either surface. Taken together, the outcomes from this work suggest that the NZVI adsorption process over the inorganic fraction of volcanic soil is a kinetically fast process described by pseudo-second order model, implying that it takes places over two different bonding sites on the adsorbent. This is consistent with information obtained by the Elovich and intraparticle diffusion models: adsorption takes place at two consecutive stages, begging with fast adsorption at macropores followed by diffusion to micropores until reaching equilibrium. Adsorption isotherms demonstrate that the inorganic fraction of Collipulli soil reaches higher maximum amounts of adsorbed NZVI, possibly due to its higher surface area. Modelling allowed thermodynamic description of the adsorption process and was better adjusted by the Langmuir-Freundlich model, indicating that the process is a combination of physical and chemical interactions. The modelling parameters reveal that the adsorption process is almost 50 times stronger over the inorganic fraction of Ralun than that of Collipulli. Finally, desorption demonstrated that while both physical and chemical processes are at play, chemical adsorption dominates markedly, as evidenced by the desorption percentages and negative hysteresis indexes. These outcomes elucidate the adsorption process of commercially available ZVI particles over the inorganic fraction of volcanic soils. Therefore, despite the effect of NZVI has been thoroughly described for hydroponical systems, it is reasonable to induce that the irreversible adsorption of NZVI onto the inorganic fraction of Chilean volcanic soils determines low concentration of nanoparticles on the water soil phase; thus, probably low plant uptake is forecast in this type of soils.

Conclusions
Zerovalent iron slurry is a widespread remediation tool for organic pollutant degradation in soils, but there is a lack of knowledge on the environmental fate of its leftovers due in part to the difficult description of the adsorption process of particle suspension over full soil. A way to shorten this information gap could be to previously study the adsorption process of zerovalent iron particle suspension on soil fractions by composition. Thus, the system comprised of commercial zerovalent iron, that includes nanoparticles (NZVI), and the inorganic fraction of two volcanic soils was fully characterised regarding the adsorption processes. Kinetically, the interaction of NZVI is a fast process that takes place in two stages and in two active sites over the adsorbents. Isothermal adsorption proved to require pH adjustment of the system that considered the Andisol as adsorbent, due the significant influence of pH on adsorbent surface charge. Experimental adsorption isotherms were better fitted by the Langmuir-Freundlich; thus, NZVI adsorption is a mix of physical and chemical interactions. Finally, the low amount of NZVI desorbed suggests an irreversible chemical adsorption on specific active sites of the adsorbents.
This reveals that NZVI adsorption over the inorganic fraction of volcanic soils is a complex process that requires strict pH monitoring and control during adsorption to prevent surface charge variation between measurement points.
Collectively, these results suggest that NZVI will tend to become irreversibly adsorbed over the inorganic fraction of soils. Regarding environmental implications, these results likely indicate that retention of NZVI particles by inorganic components of volcanic soils will extend their reactivity. This implies that there are overall low risks of NZVI transport through soil profiles, co-transport of iron-adsorbed compounds (pollutants or nutrients) into aquifers and low possibility of plant uptake.