Iron nanoparticles to recover a co-contaminated soil with Cr and PCBs

Little attention has been given to the development of remediation strategies for soils polluted with mixture of pollution (metal(loid)s and organic compounds). The present study evaluates the effectiveness of different types of commercial iron nanoparticles (nanoscale zero valent iron (nZVI), bimetallic nZVI-Pd, and nano-magnetite (nFe3O4)), for the remediation of an industrial soil co-contaminated with Cr and PCBs. Soil samples were mixed with nZVI, nZVI-Pd, or nFe3O4 at doses selected according to their reactivity with PCBs, homogenized, saturated with water and incubated at controlled conditions for 15, 45 and 70 days. For each sampling time, PCBs and chromium were analyzed in aqueous and soil fractions. Cr(VI) and Cr leachability (TCLP test) were determined in the soil samples. The treatment with the three types of iron nanoparticles showed significant reduction in Cr concentration in aqueous extracts at the three sampling times (> 98%), compared to the control samples. The leachability of Cr in treated soil samples also decreased and was stable throughout the experiment. Results suggested that nZVI and nZVI-Pd immobilized Cr through adsorption of Cr(VI) on the shell and reduction to Cr(III). The mechanism of interaction of nFe3O4 and Cr(VI) included adsorption and reduction although its reducing character was lower than those of ZVI nanoparticles. PCBs significantly decreased in soil samples (up to 68%), after 15 days of treatment with the three types of nanoparticles. However, nFe3O4 evidenced reversible adsorption of PCBs after 45 days. In general, nZVI-Pd reduced PCB concentration in soil faster than nZVI. Control soils showed a similar reduction in PCBs concentration as those obtained with nZVI and nZVI-Pd after a longer time (45 days). This is likely due to natural bioremediation, although it was not effective for Cr remediation. Results suggest that the addition of nZVI or nZVI-Pd and pseudo-anaerobic conditions could be used for the recovery of soil co-contaminated with Cr and PCBs.

Soil. Soil samples from an industrial area located in Oviedo (North of Spain) historically polluted with PCBs were collected from the surface layer (0-30 cm), air-dried and sieved (< 2 mm). Physico-chemical soil properties were determined using the Spanish official methodology for soil analysis 70 and are shown in Table 2. Briefly, pH and electrical conductivity (EC) were measured in a 1:2.5 soil-to-water ratio; available phosphorous was determined after extraction with sodium bicarbonate at pH 8.5; organic matter and total nitrogen were analyzed by the Walkley-Black and the Kjeldahl methods, respectively; available Na, Mg, Ca and K were extracted with ammonium acetate 0.1 N and quantified using a flame atomic absorption spectrometer (AA240FS, Varian); the particle size distribution of the soil was determined by the Pipette method; the concentration of Cd, Cr, Cu, Ni, Pb, Pd and Zn was measured by acid digestion with a mixture of HNO 3 (6 mL, 69%) and HCl (2 mL, 37%) in a microwave reaction system (Multiwave Go, Anton Paar GmbH), followed by the analysis of the metals in the digestion extracts by flame atomic absorption spectrometry. The limits of quantitation were in the range of 0.1 to 1.0 µg/g. The results of blank analysis were always below the detection limit, and soil reference material (SQC001, Sigma-Aldrich) recoveries were within the certified value (88-104%).
The mean concentration of Cd, Cu, Ni, Pb and Zn were below the current Regional Screening Levels for industrial uses in Asturias 71 (Table 1). In contrast, the concentration of Cr(VI) was above the allowed value (2-50 mg/kg, depending on the land uses). Regarding the PCBs, the sum of the analyzed PCBs exceeded almost three times the maximum level allowed for industrial use in Spanish legsilation 72 . Thus, given these concentrations of Cr(VI) and PCBs and according to the legislation limits, 71, 72 this soil has a contamination level which presents ecological and/or health risks, and needs to be remediated.
After nanoparticle treatment, Cr availability in soil samples was evaluated using the TCLP (Toxicity Characteristic Leaching Procedure) test (USEPA 1311). In brief, 20 mL of sodium acetate buffer (0.1 M, pH 4.93 ± 0.05) was mixed with 1 g of soil, shaken overnight and the concentration of Cr in the extract was quantified by flame atomic absorption spectrometry.
Batch experiment. Fifteen grams of soil sample was weighed in a plastic tube (50 mL Falcon) and mixed with the iron nanoparticles at 10% (20 g Fe/kg) dose for nZVI and nZVI-Pd, and 5% for nFe 3 O 4 (36.2 g Fe/kg). The doses were selected based on previous assays. Lower doses of iron nanoparticles did not induce significant degradation of PCBs in this soil (data not published). Then, the mixture was homogenized, and 35 mL of water was added to completely fill the tube and minimize the presence of oxygen. Tubes were capped. These pseudo- Chromium speciation. The concentration of Cr(VI) in soil samples was determined using ion chromatography with an UV-VIS detector at 365 nm according to the method described by Phesatcha et al. 73 . Briefly, 0.1 g of soil sample was mixed with nitric acid (5 mL, 50 mM), 10 mL eluent stock (2 mM pyridine-2,6-dicarboxylic acid, 2 mM disodium hydrogen phosphate anhydrous, 10 mM sodium iodide, 50 mM ammonium acetate and 2.8 mM lithium hydroxide) and 10 μL nitric acid (69%). The mixture was heated for 30 min, cooled to room temperature and filtered before injection in the chromatograph.
PCBs extraction in soil aqueous extract. PCBs in aqueous extract were extracted by liquid-liquid extraction. Soil leachate (10 mL) was shaken with hexane (2 × 10 mL) and EtAc (2 × 10 mL) using a separatory funnel. After shaking, the combined organic extracts were concentrated to 1 mL using a Genevac EZ-2 evaporator (NET Interlab, Spain). A small spatula-tip full of anhydrous sodium sulfate was then added to the organic extract to remove any water it may contain prior to its GC-MS analysis.
PCBs extraction in soil samples. For PCBs determination in soil samples, 1 g of soil was placed in a 20 mL glass column with a cellulose paper filter at the bottom and 1 g of Florisil. The columns were closed with one-way stopcocks before adding the extraction solvent (4 mL of EtAc). They were placed in an ultrasonic water bath at room temperature for a 15 min sonication cycle. They were then placed on a multiport vacuum manifold and the eluates were collected in conical tubes. This procedure was repeated with another 4 mL of the extraction solvent. The combined extracts were adjusted to 3 mL prior to analysis by GC-MS.
GC-MS analysis. The analysis was performed with an Agilent 6890 gas chromatograph equipped with an automatic injector, HP 7683, coupled to a quadrupole mass spectrometer model 5977. Separations were car- www.nature.com/scientificreports/ ried out using a fused silica capillary column ZB-5MS, 5% phenyl polysiloxane as nonpolar stationary phase (30 m × 0.25 mm i.d. and 0.25 µm film thickness), from Phenomenex (Torrance, CA). Helium (purity 99.995%) was used as carrier gas at a constant flow rate of 1 mL/min. Injections were carried out with 2 µL of extract in the pulsed splitless mode with the injection port at 285 °C, pulsed pressure 45 psi for 1.5 min, with the splitless injector purge valve activated 1.5 min after sample injection, in a glass liner with deactivated glass wool. The column temperature was initially set at 80 °C (held for 0.5 min), increased at 20 °C/ min to 280 °C (held for 4 min). The total time of the analysis was 14.7 min.
A stock solution of PCBs was made up at 2 µg/mL level in EtAc and the standard solutions were stored in glass flasks at 4 °C. The quantification of the PCB isomers was based on their relative response to external standards. The linear range was established by a five-point calibration curve in the range 12.5-250 ng/mL.
The target and qualifier abundances were determined by injecting standards under the same chromatographic conditions, using full-scan with the mass/charge ratio ranging from 100 to 800 m/z. The chromatographic method was divided in five time segments. Table S1 lists the mass spectrometry parameters of the PCBs for the GC-MS method. Retention times must be within ± 0.2 min of the expected time and qualifier-to-target ratios within a 20% range for positive confirmation.
Statistical analysis. Data were statistically analyzed using the IBM SPSS package for Windows, release 19.0.0.1. The normal distribution of all variables was checked by a Kolmogorov-Smirnov test. Difference among treatments were determined by one-way analysis of variance at significance level of p < 0.05, followed by a Tukey post-hoc test.

Results and discussion
Characterization of iron nanoparticles. SEM images of the three iron nanoparticles are shown in Fig. 1.
Regarding nZVI, most of the nanoparticles are spherical with a regular size and forming chain like aggregates. Bimetallic nanoparticles showed a similar surface topography to nZVI formed by spheres and chain-like aggregates. Magnetite nanoparticles were made of spherical particles with irregular sizes. Figure 2 shows the FTIR spectra and the powder XRD patterns of the three types of iron nanoparticles. The FTIR spectra of nZVI and nZVI-Pd are similar, with an intense absorption peak near 3400 cm −1 related to O−H stretch from water. The band at 1600 cm −1 can be due to O-H-O scissors-bending or to the -C=O stretch of the carboxylic acid of the stabilizer (sodium salt of polyacrylic acid, Table 1) 74 . The bands between 570 and 400 cm −1 , characteristic of the Fe-O bond, confirm the presence of iron oxides 38,65,75 . FTIR spectrum of nFe 3 O 4 showed two strong absorption bands at 540 and 430 cm −1 , which can be assigned to the Fe-O stretching mode of the tetrahedral and octahedral sites 76 .
As can be noted, nZVI and nZVI-Pd exhibit rather similar XRD patterns with diffraction peaks at 2θ = 44.64°, 65.01° and 82.32°, characteristic of α-Fe 0 (ICDD 04-013-5208, 04-006-3633), and a weaker peak at 35.48° which can be attributed to iron oxide (ICDD 00-003-0863, 01-084-2782). These results verified that nZVI has a core-shell structure, a core of Fe 0 and a thin shell constituted by iron oxides. Regarding the XRD pattern of nFe 3 Figure 3 collects the XPS spectra obtained for the three nanoparticles. nZVI and nZVI-Pd showed a similar pattern, being iron and oxygen quantitatively the most important elements; sodium and carbon come from the organic stabilizer (Table 1) and Si may be an impurity of the nanoparticles as concluded in a previous study 38 . The spectra of O 1s region (Fig. 3C,F) included oxides (529 eV), hydroxides (531 eV) and chemically or physically adsorbed water (532 eV), characteristics of the iron oxide from the nZVI surface 34 . On the basis of the binding energy and the form of the Fe 2p peak of nZVI (Fig. 3B) we can concluded that αFe 2 O 3 predominates on the surface of nZVI. In the case of nZVI-Pd, αFe 2 O 3 and FeOOH were found on the surface. No metallic iron was identified in either nZVI or nZVI-Pd probably due to XPS technique analyzes the surface of the nanoparticles and metallic iron is present on the core of nZVI 38,77 . A low proportion of Pd was also detected on bimetallic nanoparticle surface (Fig. 3G). The Pd 3d spectrum showed Pd 3d 5/2 signals at 334.7 eV and 336.5 eV, characteristic of metallic Pd and PdO, respectively 77-79 . Huang et al. 77 also detected metallic Pd and Pd 2+ on nZVI-Pd surface. Regarding the nFe 3 O 4 , the Fe 2p 3/2 spectrum (Fig. 3I) showed a binding energy of 711 eV and a satellite peak located at 719 eV, characteristic of Fe 3+ . This observation, together with the narrow shape of the main Fe 2p 3/2 peak, suggests that iron is present mainly as maghemite (γFe 2 O 3 ) 80 , which can be formed through the oxidation of magnetite. The region of O 1s showed that oxygen is present mainly as oxide and hydroxide at the magnetite surface. The magnitude of zeta potential indicates the colloidal stability of the nanoparticles; values more positive than 30 mV or more negative than − 30 mV, are considered stable, with maximum instability (i.e., aggregation) occurring at a zeta potential of 0 30 . In the present study, the colloidal stability in decreasing order was nZVI > nZVI-Pd > nFe 3 O 4 ( Table 1).

Method validation of PCBs analysis.
After optimization, the developed method was evaluated in terms of linearity, accuracy precision and detection limits before it was used to determine the concentrations of PCBs at the different times in the soil samples. Detailed information about its validation is provided in the Supplementary section.
Cr in soil aqueous fraction. The addition of nZVI, nZVI-Pd and nFe 3 O 4 significantly reduced Cr concentration in the extracts at the three sampling times (Fig. 4A). In contrast, control samples showed a higher Cr concentration (close to 22 mg/L) which was similar throughout the experiment. Most of the Cr present in aqueous www.nature.com/scientificreports/ fraction is likely Cr(VI) since Cr(III) it is almost completely precipitated at pH 5.5 or higher 2, 3 . No differences were detected among the nanoparticle treatments and after 15 days of interaction, the Cr concentration in aqueous extract of treated samples was below 0.4 mg/L, after 45 days it was below 0.2 mg/L and after 70 days below the quantitation limit for the three iron nanoparticles (0.06 mg/L). In this regard, Wang et al. 81 did not find Cr in the aqueous phase after 72 h of interaction between Cr-polluted soil and nZVI stabilized with carboxymethyl cellulose (CMC) in aqueous medium.
Cr availability in soil samples. Cr availability in soil samples was evaluated by considering the potential Cr leachability using the TCLP test (Fig. 4B). The addition of iron nanoparticles significantly reduced the TCLP-Cr which agrees with the previous results in water extracts. In this regard, TCLP-Cr in control samples was between 5.9 and 7.5 mg/kg, significantly higher than those found in treated soils which ranged from 1.1 to 1.7 mg/kg. No significant differences were observed in Cr immobilization among the three types of iron nanoparticles. In relation to the sampling times, no differences were found among them for each treatment, for both the treated and control soils. In summary, the addition of nZVI, nZVI-Pd or nFe 3 O 4 to Cr-polluted soil  Fig. 4C. The soil samples treated with nZVI and nZVI-Pd showed similar concentrations of Cr(VI) at the three sampling times (15, 45 and 70 days), with a mean value close to 6 mg/kg, significantly lower than the initial concentration (65 mg/kg). However, the TCLP values for these soil samples were below 1.6 mg/kg (likely as Cr(VI)). These results demonstrate that Cr was immobilized as Cr(III) and Cr(VI). Thus, the interaction mechanism between nZVI and Cr(VI) in the soil samples implied the combined process of reduction to Cr(III) and adsorption of Cr(VI) on the shell of the nanoparticle as other authors have concluded [81][82][83][84] .
In this regard, Wang et al. 84 applied CMC-nZVI to a Cr(VI)-polluted soil and they concluded that the Cr(VI) was initially absorbed into the shell of nZVI, then, most of it was gradually reduced to Cr(III) whereas iron and chromium oxidation products (Fe(III) and Cr(III) oxides/hydroxides) as well as the products of the hydrolyzation of the CMC covered the nZVI, limiting contact of the Cr(VI) with the reductant Fe(0). According to the reduction potential (E 0 (Fe 2+ ) = − 0.41 V, E 0 (Cr 6+ ) = 1.36 V), the reduction of Cr(VI) to Cr(III) is thermodynamically favorable forming Cr(OH) 3 and Cr-Fe (oxy)hydroxide 34,82,83,85,86 . The same interaction mechanism exists for nZVI and nZVI-Pd; the main difference between both iron nanoparticles is that Pd acts as a catalyst  showed similar concentration of Cr(VI) as the controls and they were higher than those found in nZVI and nZVI-Pd treatments for all the sampling times. However, it should be noted that the control samples exhibited much of the mobile chromium in the aqueous fraction (Fig. 4A). No significant changes among sampling times were detected. The concentrations of Cr(VI) found in soil samples treated with nFe 3 O 4 were in the range of 8.5 and 10.5 mg/kg, higher than Cr-TCLP (Fig. 4B). Thus, reduction and adsorption were employed for Cr(VI) removal. In this case, Cr(VI) reduction was favored by the presence of Fe(II) from the magnetite (Fe(II)Fe(III) 2 O 4 ) 91-93 . However, the reducing character of nFe 3 O 4 was lower than nZVI particles which contains Fe 0 and Fe 2+ as reducing agents. The application of nFe 3 O 4 with different modifications has been effectively used as an adsorbent for the decontamination of Cr-polluted waters or inert materials like sand 65,92,94 . However, to the best of our knowledge, little data is available on the effectiveness of nano-magnetite for Cr immobilization in polluted soils. Similar results were found in the three sampling times.
Control soil samples presented mean values lower than that found in the original soil. This may be because most of the Cr present in aqueous phase was probably Cr(VI), the most mobile form of Cr. The decrease of the concentration of Cr(VI) over time can be due to its reduction to Cr(III) by abiotic and biotic processes 5,95 . The presence of Fe(II), reduced sulfur compounds and organic matter (humic and fulvic acids) can be a source of electrons for reducing Cr(VI) 5 4D). Thus, a low percentage of the total Pd was available (a maximum of 0.95% at the first sampling time), and this parameter decreased significantly over time (p < 0.05).

PCBs in soil aqueous fraction and soil.
PCBs were not detected in aqueous fraction for any of the treatments at any sampling time, probably due to their high hydrophobicity. The PCBs mean concentration in the soil samples at the different sampling times is shown in Fig. 5. After 15 days of interaction between polluted soil and iron nanoparticles, a significant decrease of PCBs was observed for all the studied PCBs. The three types of iron nanoparticles significantly reduced the PCBs concentration under the experimental conditions after 15 days; nFe 3 O 4 and nZVI-Pd showed similar reduction results, and they were significantly more effective than nZVI for PCB101, PCB153 and PCB138. The PCB28 was the least relevant quantitatively, and the reduction rate was also the lowest (close to 30%). The decrease of PCB52 was between 57 and 64%. For the PCB101, decreases of 66%, 47% and 64% were observed for nFe 3 O 4 , nZVI and nZVI-Pd, respectively. Percentages of decrease of 60%, 42% and 58% for PCB153, of 68%, 38% and 58% for PCB138, of 48%, 29% and 48% for PCB180 were obtained for nFe 3 O 4 , nZVI and nZVI-Pd, respectively after 15 days of interaction. Thus, the present results show the capacity of magnetite nanoparticles to reduce PCBs content in soil samples. However, the soil samples treated with nFe 3 O 4 over a longer time period, 45 and 70 days, showed an increase in PCB content. We hypothesize that magnetite retains the PCBs by adsorption, but the interaction is reversible, and the PCBs are released over time. Thus, the PCBs concentration in soil samples treated with nFe 3 O 4 was higher after 70 days of contact than those collected after 45 and 15 days. Thus, further studies are necessary to determine the optimum conditions of contact time, so as to evaluate the effectiveness of their use for reducing PCBs availability in soil, as well as the potential regeneration and reusability of the nFe 3 O 4 . In addition, due to the magnetic properties of nFe 3 O 4, they could be easily separated from the soil under a magnetic field, regenerated and reused several times 67 .
In the 15-day sampling, the bimetallic nanoparticles proved to be more effective and reduced PCB concentration faster than nZVI, especially in the case of PCB101, PCB153 and PCB 138. In the later samplings (45 and   97 concluded that the removal efficiency of the PCB101 in spiked soils improved with increasing concentration of bimetallic ZVI-Pd nanoparticles (both Fe and Pd dosage). The authors found an increase in the removal percentage from 20.3% at dose of 0.005% of Pd to 57.8% at 0.1% of Pd. This is probably due to two main reasons: i) the galvanic effects of the Fe-Pd system, in which Fe acts as anode and Pd as cathode, and the electrons released from Fe contribute to form Cl -; ii) Pd can strongly bond to Cl thereby accelerating the dissociation of chlorinated hydrocarbons; the H 2 (g) produced during the iron oxidation is adsorbed on Pd and dissociated into atomic H, one of the strongest reductants for the dechlorination reactions [97][98][99][100] . The comparison over time showed that the PCB concentrations did not change in soils treated with nZVI-Pd, no significant degradation was observed after 45 and 70 days. In those treated with nZVI, only PCB101, PCB153 and PCB138 significantly decreased after 45 days, however no more reductions were observed at longer time. The decrease in effectiveness of nZVI and nZVI-Pd over time is probably due to two main reasons. Firstly, the reducing power of ZVI decreases with time, due to oxidation of Fe(0) and aggregation of ZVI nanoparticles in  www.nature.com/scientificreports/ the soil matrix, leading to a decrease of nanoparticle reactivity 99 . Although both nanoparticles contain an organic stabilizer to prevent the aggregation (Table 1), soil is a very complex matrix, and many different interactions can occur. Secondly, the different availability of the PCBs in soil matrix, i.e., the degraded PCBs were likely in the most available fraction of the soil and those most recalcitrant PCBs were not available to react with ZVI nanoparticles. PCBs can be strongly adsorbed to soil particles, especially organic matter which makes their interaction with nanoparticles difficult. In this regard, Varanasi et al. 54 effectively reduced the concentration of PCBs in a soil after treatment with nZVI together with a nitrogen stream at 300 `C to allow the desorption of PCBs from the soil. In addition, as the soil is also polluted with Cr, this can induce competitive phenomena among pollutants to react with ZVI nanoparticles. Control samples suffered a bioremediation process, showing the higher decrease of PCBs at 45 days remaining stable up to 70 days. The best results were observed for the PCB52, PCB101, PCB153 and PCB138, with removal rates between 49 and 60%. These reductions are likely associated with anaerobic biodegradation processes because of the pseudo-anaerobic conditions. Other authors have observed the microbial degradation of highly chlorinated PCB congeners under anaerobic conditions 48,101 . Consequently, the concentrations of PCBs in control and treated samples were similar after 45 days. Thus, the application of nZVI-Pd and nZVI reduced the concentration of PCBs in soil in a similar range to those obtained by the bioremediation process in control samples but in less time. Similar PCBs concentration were found at 45 and 70 days; thus, no biodegradation processes occurred after 45 days of interaction at the experimental conditions. As previously explained, this can be due to biodegradation occurs with the most available fraction of PCBs, and the less available fraction is not easily accessible for microbiota. Thus, the PCBs availability could be the limiting factor for both, nanoremediation and bioremediation of PCBs in the polluted soil. Soil properties and degree of pollution would affect the remediation efficiency. According to the present results, bioremediation would be feasible for soil polluted exclusively with PCBs but not when soil also includes metals such as Cr.

Conclusions
The addition of nZVI, nZVI-Pd or nFe 3 O 4 to a soil co-contaminated with Cr and PCBs significantly reduced the leachability of Cr in soil and the immobilization was stable for at least 70 days under the experimental conditions. The nZVI and nZVI-Pd showed higher effectiveness for the reduction of Cr(VI) to Cr(III) compared to that of nFe 3 O 4 . After 15 days of interaction between soil-nanoparticles, the PCBs concentration significantly decreased in soils treated for the three types of iron nanoparticle. Magnetite nanoparticles exhibited a reversible process for PCBs adsorption. Soils treated with nZVI and nZVI-Pd showed a similar PCB degradation rate at 45 days of treatment. The latter approach required less time to degrade these compounds and was more effective even at 15 days; however, the use of nZVI-Pd implies the incorporation of Pd into the soil, although we observed that the available content was lower 1% of the total and it decreased over time. Due to bioremediation processes, the control soils showed a reduction in PCBs concentration in the 45-day sampling time, reaching similar values to those found in soils treated with nZVI and nZVI-Pd. In this regard, bioremediation would be feasible for soil polluted exclusively with PCBs but not when soil also includes metals such as Cr. Thus, nZVI based nanoparticles evidence a moderate efficacy for the remediation of PCB-polluted soils. In contrast, both nZVI nanoparticles exhibited successful results for the immobilization of Cr in soils. The results suggest that the addition of nZVI or nZVI-Pd and pseudo-anaerobic conditions could be used for the recovery of soils co-contaminated with Cr and PCBs.