The effect of aqueous corrosion on the structure and reactivity of zero-valent iron

(cid:1) nZVI retain a Fe 0 core for over 4 w when aged in pure water. (cid:1) Goethite is proposed to be the ﬁnal aqueous corrosion product of nZVI. (cid:1) Aqueous corrosion of nZVI has a signiﬁcant negative impact on remediation reactions. (cid:1) Loss of the Fe 0 core led to metal remobilisation from nZVI surfaces.


Iron in the environment
Zero-valent iron nanoparticles are an emerging technology for the treatment of contaminated groundwater.The bulk form of zero-valent iron has been employed for remediation purpose for over twenty years [1], in the form of permeable reactive barriers with reactive materials such as cast iron, steel wool, and amorphous ferric oxide [2].The corrosion products of these 'filler' materials are fundamentally bulk, static iron oxides, and are ubiquitous in nature (the term iron oxides here refers to both iron oxides, such as wüstite -FeO and iron oxide hydroxides, such as goethite -c-FeOOH).Despite the ubiquity of these iron phases, concerns have been raised as to the corrosion products of nZVI [3,4], as to whether they are similar to their bulk counterparts.

The effect of ageing on the structure of nZVI
Due to its high surface area and greater availability of surface reactive sites, nZVI will corrode faster than bulk-Fe 0 [5], but Fe 0 in any form will already have a film of surface oxide acquired directly after synthesis [6][7][8], typically 3 nm thick [9], but the exact structure depends on the synthesis process, particle size, and storage conditions [10,11].On aqueous exposure, it has been suggested that at the shell-water interface, water will disrupt the oxide shell by hydration or autoreduction [12,13], yielding Fe 2+ species, the stability of which is mostly dependent on local geochemical conditions [14,15].For waters encountered in nature, Fe 2+ species are of limited stability and are very sensitive to rapid oxidation by dissolved oxygen (DO) [12], yielding a large array of porous and highly adsorptive iron hydroxides (e.g., Fe(OH) 2 , Fe(OH) 3 , ferrihydrites) [16].The resulting Fe 3+ species readily hydrolyse, precipitate and transform to various less porous and less adsorptive oxides in the following way: [Fe(H 2 O) 6 ] 3+ (aq) ?Fe(OH) 3(aq) ?Fe(OH) 3(s) ?oxides [17], which includes wüstite, magnetite (Fe 3 O 4 ), maghemite (c-Fe 2 O 3 ) and FeOOH species [16].

The effect of ageing on the reactivity of nZVI
A small number of studies have examined the effect of corrosion on the reactivity of nZVI.These are summarised in Table 1.However, it has been reported that reactivity (degradation of C 2 HCl 3 ) Another study reported that air-exposure did not significantly effect nZVI reactivity (degradation of C 2 HCl 3 ), although all Fe 0 had been consumed within 11 d [25], although this has been contradicted by other works [11,22].
From the above, it can be stated that there has been limited work concerning the progressive environmental oxidation of nZVI, an idea highlighted in a recent review paper [27], which noted that although ageing has been identified as a significant issue impacting the reactivity of nZVI, limited work has been completed.Therefore, the specific objectives of this study are to map the structural and surface changes to nZVI as it corrodes in a simple solution, and then expand this to investigate how aqueous corrosion effects nZVI reactivity in relation to the uptake to three metal contaminants.The overall aim is to provide underpinning evidence and demonstration that the corrosion products of zero-valent iron nanoparticles are similar to the corrosion products of bulk zero-valent iron, and in this way, allay some of the concerns raised [3,4] as to the use of zero-valent iron nanoparticles for the remediation of contaminated groundwater.

Synthesis of nanoparticles
Zero-valent iron nanoparticles were synthesised using borohydride to reduce ferric to a metallic state [28,29].Briefly, 30.6 g of Fe(SO 4 )Á6H 2 O was dissolved in 200 mL of water that had been purged with oxygen free nitrogen gas for 10 min.By slowly adding 4 M NaOH dropwise to prevent the formation of hydroxocarbonyl complexes, the solution pH was adjusted to 6.8.To this, 12.0 g of NaBH 4 was added ($0.2 g per min), to reduce the Fe 2+ to Fe 0 .The resultant solid material was washed through centrifugation and sequential rinsing in 50 mL of water, acetone and then ethanol.The nanoparticles were dried under low vacuum in a vacuum desiccator ($10 À2 mbar), and then stored in an argon filled (BOC, 99.998%) MBraun glovebox to prevent further oxidation until required.

Effect of aqueous corrosion on structure and surface chemistry
To assess bulk and crystallinity changes to nZVI, a 5 L aspirator bottle was filled with 5 L of water and 5.0 g of nZVI (leaving $100 mL headspace).The nZVI were dispersed by agitating the aspirator bottle for 5 min.From this, aliquots of 500 mL of nZVIliquid were removed periodically after agitation: 2 h, 1 d, 3 d, 1 w, 2 w, 4 w, 8 w, and 16 w.These aliquots were centrifuged at 5000 rpm for 10 min using an Eppendorf 5810 bench-top centrifuge, and then the supernatant decanted.The remaining solid material was cleaned using the same sequential washing method used in Section 2.2.The oxidised nZVI were stored under clean ultra-pure ethanol before analysis (typically 24 h).The nZVI for BET analysis were dried and stored using the same method used in Section 2.2.Headspace was not controlled within the aspirator bottle during the experiment and so comprised of ambient air.

Effect of aqueous corrosion on reactivity
The effect of aqueous corrosion on the reactivity of nZVI was assessed using three metal contaminants, divalent copper (Cu) and zinc (Zn) and hexavalent chromium (Cr(VI)) chosen for their differing uptake inter-reactions with zero-valent iron nanoparticles (reduction, adsorption and reduction-precipitation respectively).[30] For each metal contaminant, eight 250 mL Pyrex beakers were filled with 190 mL of pH 4 (adjusted using 0.1 M HCl) Milli-Q water.To each, except one blank (nZVI free) solution, 0.5 g L À1 of nZVI were added.All batches were sonicated in a VWR USC100T ultrasonic bath for 5 min after nZVI addition to aid dispersion, and then covered loosely with thin, clear, laboratory plastic film to allow gas exchange but to prevent evaporation and cross contamination.
After the required ageing period: unaged, 2 h, 1 d, 3 d, 1 w, 2 w and 4 w (the last time point which Fe 0 was detected), an aqueous Bromate (KBrO 3 ) A decrease in the reactivity for nZVI stabilized with inert gas and then exposed to air and humid air up to 1 month [11] metal contaminant that had been adjusted to pH 4 using either 0.1 M HNO 3 or 0.1 M NaOH was added to each batch.For divalent Cu(II) and Zn(II), initial concentrations were 50 mg L À1 , while Cr (VI) was 20 mg L À1 .Each batch system was sampled at 0, 1, 2.5, 5, 10, 20, 40, 80, 160, 320 and 640 min, as well as 1 d, 3 d, and 7 d.At the same time point, geochemical parameters (pH, Eh and DO) were measured with a Hach HQ40d Multimeter.To ensure sample homogeneity, the beakers were gently agitated prior to sampling and geochemical measurement.A solid sample for analysis was retained at the completion of the experiment for further analyses.From each beaker, 10 mL aliquots were withdrawn, and the solid and liquid phases were separated by centrifugation at 6500 rpm for 1 min in a Hamilton Bell Vanguard V6500 desktop centrifuge.The resultant liquid was syringe-push filtered (0.22 lm cellulose filter), and then acidified using 1% v/v concentrated nitric acid to prevent metal precipitation or adsorption to the sample vessel.The remaining solid material was cleaned using the same sequential washing method as outlined in Section 2.2.The samples were stored under ultra-pure ethanol until analysis (<24 h).

Sample analysis methods
Specific surface area (SSA) was determined by Brunauer-Emmett-Teller (BET) analysis using a Quantachrome NOVA 1200 instrument.A known weight of the dried, vacuum degassed ($10 À2 mbar for 12 h at 75 °C) material was analysed using nitrogen as the adsorbent and following a 7 point BET method.Transmission electron microscopy (TEM) images of nZVI mounted on 200 mesh holey carbon coated copper grids were obtained with a JEOL JEM 1200 EX Mk 2 TEM, operating at 120 keV.A Phillips Xpert Pro diffractometer with a Cu radiation source (CuK a 1.5406 A°) was used for X-ray diffraction (XRD) analysis (generator voltage of 40 keV; tube current of 30 mA), and plots were acquired between 10 and 70°2h, with a step size of 0.02°and a 2 s dwell time.A Thermo Scientific K-Alpha + spectrometer employing a monochromatic source (AlK a 1486.6 eV) was used for X-ray photoelectron spectroscopy (XPS) analysis.Data analysis was performed using the CasaXPS package [31].Fe 2p 3/2 peak was fitted according to the method of Grosvenor et al., (2004) [32], and the O 1 s peak according to the method of Li and Zhang (2004) and Abdel-Samad and Watson (1997).[30,33] X-ray absorption spectroscopy (XAS) spectra were collected at Beamline B18 at the Diamond Light Source, Harwell, UK from pelletised samples in transmission mode at room temperature.The storage ring was operating in a 10 min top-up mode for a ring current of 250 mA and an energy of 3 GeV.The radiation was monochromated with a Si(1 1 1) double crystal, and harmonic rejection was achieved through the use of two platinum-coated mirrors operating at an incidence angle of 8.0 mrad.The monochromator was calibrated using the K-edge of an iron foil, taking the first inflection point in the Fe-edge as 7112 eV.Samples for XAS were prepared in an MBraun argon filled glove box by grinding 3 mg of nZVI with 50 mg of cellulose in an agate mortar for 3 min.The subsequent homogenous mixture was pressed into an 8 mm pellet and then sealed with Kapton tape.Obtained spectra were summed and background-subtracted using the software package Athena.(Demeter 0.9.24;Ravel and Newville, 2005).Sample composition was then determined using the linear combination fitting (LCF) routine of Athena.The X-ray absorption near edge structure (XANES) spectra were fitted against spectra of known standards using a fitting window of À20 to +53 eV from 7112 eV [34].Inductively-coupled plasma optical emission spectroscopy (ICP-OES) was used to analyse aqueous metal concentrations, the instrument employed being an Agilent 710.

Changes to structure due to aqueous corrosion
The changes to nZVI crystallinity (XRD scans) and size (TEM images) due to aqueous oxidation are recorded in Figs. 1 and 2 respectively, while XANES spectra are shown in Fig. 3. Surface area as determined by BET, and results from XPS fitting of the Fe 2p 3/2 and O 1s peaks are presented in Table 2.The identified iron forms as identified by XRD and XANES analysis are also presented.Due to the limited extent of the oxide layer in nZVI and the similar spinel crystal structure of both maghemite and magnetite, their differentiation by XRD is extremely difficult [24,35].These phases were differentiated using XANES analysis.

As-formed nZVI
Characterisation of the as-formed nZVI revealed typical features associated with other studies using borohydride-reduced zerovalent iron nanoparticles [29].Analysis by XRD (Fig. 1), XAS (Fig. 3) and XPS (Figs.S1 and S2 -Supporting information) indicated that the core of nZVI was poorly crystalline a-Fe (characteristic XRD peak at 44.9 °2h) with a shell of mixed valent iron oxides/ oxyhydroxides.Fitting of the XANES data (Fig. 3) indicated the presence of Fe 0 , along with 2-line ferrihydrite, (tentatively given as Fe 5 O 8 HÁH 2 O) [36], although near 20% of the total fit was undetermined.The presence of 2-line ferrihydrite is discussed in more detail below, although the presence of this iron form does suggest that the surface of the unaged nZVI contained little crystalline iron oxide phases that have previously been reported, such as wüstite [37], magnetite [38], maghemite or FeOOH [30,39].Siskova et al. (2012) stated that the shell of borohydride-reduced particles were Fe 3+ oxide/hydroxide of an unspecified ratio [40], which is in broad agreement with the results in this work.The as-formed shell composed of a number of mixed-valent iron oxide hydroxides, with hydroxide being the more prevalent form (XPS determined stoichiometric ratio of the O 1 s peak of OH À :O 2À of 5).TEM analysis determined that the particles were roughly spherical (Fig. 2, Panel 'unaged'), with an approximate range of 40-100 nm, and an observable dark core surrounded by a lighter oxide layer, although no clear grain structure was identifiable.BET analysis determined that the surface area of nZVI was 11.5 m 2 g À1 .

Up to 1 d aqueous exposure
After 1 d of exposure, the surface area of the corroding nZVI increased over tenfold, to 160.8 m 2 g À1 .This change was attributable to the formation of a large volume of corrosion products, and/or cracking and pitting of the shell due to corrosion propagated strain; both processes have been previously observed [41,42].The corrosion products can be observed as 'lacy' areas on the surface of the nanoparticles (Fig. 2, Panel 2h and 1d), although the core of the particles remains dark, indicative of zero-valent iron, confirmed by the XRD analysis.It is not possible to determine whether these lacy products are from dissolution and precipitation, or from surface reactions [42].However, the broad arc around 25 °2h in the 2 h XRD plot has been assigned as ferrihydrite in other corrosion studies [43,44], which would form at the surface of corroding nZVI.Linear combination fitting of the XANES 2 h data does indicate the presence of 2-line ferrihydrite, although this nZVI sample appears more oxidised than the 1 d or 3 d spectra; this could be due to the presence of greater amounts of heterogeneous amorphous metastable oxidised iron species than at later time points, or possibly oxidation during analysis.This fits well with the observed increase in SSA, as ferrihydrite has a reported volume 6.4 times larger than the equivalent mass of nZVI [42].Ferrihydrite has various reported formulas, including Fe 5 HO 8 ÁH 2 O [36] and FeOH 3 Á3H 2 O [42] and so the XPS determined surface stoichiometry in this study agrees well with the latter.Ferrihydrite will not develop greater crystallinity over time, but transforms to other iron hydroxides with associated dehydration [36].The emergent XRD peaks at 35.6 and 57.6 °2h were assigned to magnetite/maghemite, the first phase detected other than Fe 0 .Both these iron oxides contain octahedral and tetrahedral iron [45,46].At the shell, the lack of Fe 2+ probably excludes the presence of magnetite [36], and therefore it is proposed that magnetite transforms, at least in part, to maghemite.This agrees with a study by Nagayama and Cohen [47], which concluded that the oxide layer on iron consists of an inner layer of magnetite and an outer layer of maghemite.However, as stated, in this work ferrihydrite was also identified.

Up to 4 w aqueous exposure
Linear combination fitting of the XANES analysis of nZVI aged between 3 d and 4 w indicated the formation of magnetite (24.1%) after 1 w, and reduction in the proportion of maghemite to 45.6%.After 2 w exposure, LCF indicated the sample was composed of magnetite (61.3%), maghemite (27.2%) and lepidocrocite (4.3%).After 4 w, XANES LCF analysis showed maghemite and lepidocrocite dominate (33 and 53.6% respectively), with no magnetite identified.Various surface and bulk parameters indicated a slowing of the transformation processes (Fig. 4).An increase in XPS determined Fe 3+ octrahedral :Fe 3+ tetrahedral ratio suggests the formation of more thermodynamically stable Fe 3+ compounds [48].Images obtained from TEM indicated the reduction in the particle size and the amounts of 'lacy' corrosion products (Fig. 2, Panel 1w).A lightening of the nanoparticle core colour, with larger areas of grey oxides/ hydroxides, inferred the expenditure of the Fe 0 fraction of the particle.At 4 w, XRD plot and XAS LCF analysis clearly demonstrated the mixed iron nature of the particles with zero-valent iron (44.9 °2h), magnetite/maghemite (30.2, 35.6 and 57.6 °2h) and lepidocrocite (14.1, 27.1, 36.3 and 38.1 °2h) all identified.This latter phase has been observed as a corrosion product in previous studies [12,18,36].After this time point, the zero-valent peak at 44.9 °2h was either absent or too small in intensity to be detectable.This is in agreement with the XAS LCF analysis.

Post 4 w aqueous exposure
This period of nZVI corrosion could be described as a slow transformation of the iron oxide and oxyhydroxide phases, rather than the more rapid changes identified at shorter timescales.Maghemite (as identified by XANES analysis, Fig. 3 and Table 2) and lepidocrocite were detected in the 8 w XRD scan, but at 16 w, goethite was also identified (21.3 °2h).Goethite could be included in the fit of the XANES signal (at $3%), although the fit was not significantly improved by its inclusion.Goethite is, along with hematite, the most thermodynamically stable iron compound and therefore the end member of many transformation routes [36].Analysis by XPS indicated that (physically and/or chemically) adsorbed surface water had reduced from 13.7% of the O signal at 4 w to 1.3% by 16 w.All fine grained iron materials contain up to several percent of adsorbed water [15], and so this value could be indicative of a final, stable state.At 4 w, the OH À :O 2À ratio of the nanoparticle surface was 0.8, and Fe 3+ octahedral :Fe 3+ tetrahedral ratio was 2.6; by 16 w, these values were 0.9 and 2.3 respectively.The 16 w O 2À :OH À ratio agreed well with reported values for iron oxyhydroxide phases such as lepidocrocite (0.9-1.1).[33,49] However, the continued presence of tetrahedral Fe 3+ indicated that the shell chemistry is not homogeneous; the latter phases of iron oxidation, such as lepidocrocite, goethite and hematite do not contain Fe 3+ in tetrahedral coordination [45].The 8 w, and to a greater extent the 16 w, XRD scans indicated greater crystallisation than the 4 w nanoparticles, with more defined peaks detected for each phase, and a continued reduction  in the size of the predominantly grey particles (42 nm at 8 w, 36 nm at 16 w).Specific surface area had stabilised after 8 w at $57 m 2 g À1 .Reported surface areas of nanoparticulate iron corrosion products include goethite, $25 m 2 g À1 [15], lepidocrocite, $68 m 2 g À1 [50].maghemite, $40 m 2 g À1 [51], and magnetite, $50 m 2 g À1 [7].These figures indicate that the 16 w nanoparticles are a mixed-phase iron corrosion product.Of note, LCF of the XANES signal indicate the presence of 2-line ferrihydrite in all aged nZVI, indicating that transformation processes were continuing up to the termination of the experiment.

Summary of pathway of nZVI aqueous corrosion
With respect to the above, the first step in the transformation of nZVI is the conversion of zero-valent (or ferrous iron) by oxidation with DO or water to form magnetite.Magnetite can be further oxidised to maghemite [36].Lepidocrocite is formed through the oxidation of dissolved ferrous ions or magnetite/maghemite.Finally, lepidocrocite oxidises to the thermodynamically more stable goethite [36,52].This reaction series can be inferred from the recovery and analysis of hydrogen-reduced iron nanomaterial from a field pilot study [53].After near four months aqueous oxidation of nZVI, the resultant material was identified as mixed phase magnetite/maghemite, lepidocrocite, goethite nanoparticles.The emergence of this latter phase revealed that corrosive transformation had not reached an endpoint.

Starting solution geochemistry
For batches treated with unaged (i.e.as-formed) nZVI, with the addition of the contaminant and nZVI, a rapid shift to chemically reducing, anaerobic, raised pH ($8) conditions was observed, with Eh minima recorded in all systems within 120 min (Fig. S3 -Supporting information).All 2 h, 1 d and 3 d solutions were experiencing negative redox, raised pH and low DO (<2.0 mg L À1 ) conditions at the commencement of the experiment, indicating that nZVI corrosion was well advanced.This is in contrast to both the 1 w, 2 w and 4 w solutions, where starting Eh, pH and DO conditions were similar to the control.Starting iron concentrations are shown in Table 3.

Changes to solution metal concentration
The maximum uptake and final retention of metals by nZVI aged for differing times, as well as rate constants, are shown in Table 4 and Fig. S4 -Supporting information.A comparison of reaction kinetics is shown in Fig. 5. From these, it is evident that irrespective of nanoparticle age, a metal uptake response was observed in all solutions.A linear regression of measured data points indicates a pseudo-first-order reaction [20].The first step in any surface mediated reaction is the adsorption of the reactants onto the adsorbent surface [54], and the subsequent slowing of the  rate has been ascribed to a physical occlusion of the zero-valent iron surface sites, rather than a chemical mechanism [20].From Fig. 5, this phenomenon can be observed during which the contaminant is removed from solution significantly faster at the commencement of the experiment than at later times.Due to the swift reaction rates, only the most aged nZVI (1 w, 2 w and 4 w) can be plotted for the Cu solution.It can be observed that a different response is elicited by aged nZVI depending on the contaminant treated, although a general decrease in rate constant is commensurate with an increase in the oxidation/age of the nZVI.Furthermore, it can also be seen that for Cu and Zn, final solution concentrations (after 7 d) were generally higher than the lowest concentration recorded, i.e. without external influence, the previously removed metal was released from the nZVI surface.This contaminant remobilisation mechanism is discussed below.
In the Cr(VI) batch systems, only the 4 w nZVI did not exhibit significant removal; no remobilisation was recorded in any of the solutions.In the Cu batch systems,<3 d nZVI completely removed the contaminant from solution, but all nZVI batches experienced contaminant remobilisation after an initial removal spike.Irrespective of age, no nZVI removed the totality of Zn from the solution.these batches, the contaminant was released from the less corroded nZVI (<1 w), but not from nZVI that had been aged for greater periods.
From the above, it can be stated that the removal efficiency of nZVI is significantly affected by its corrosion state, although the reduction in efficiency is specific to the contaminant uptake mechanism (discussed below).However, a general trend was observed, with decreasing reactivity as nanoparticle age increased.This is attributable to a reduction in electron transfer, due either to a reduction in Fe 0 volume/Fe 2+ dissolution, or an increase in the surface oxide layer hindering electron transfer.
Chromate removal by nZVI occurs via reduction and precipitation [55], although aqueous Fe rapidly reacts with Cr 6+ to promote precipitation of insoluble Cr(OH) 3 , Fe 3+ -Cr 3+ hydroxide and/or Fe 3+ -Cr 3+ oxyhydroxide [56][57][58][59].In this experiment, it is thought that this second reaction dominated, with calculated rate constants generally decreasing with increasing age, tallying with starting Fe concentration (Table 3).However, the unaged nZVI solution, with no initial aqueous Fe, had the lowest reaction rate of any tested.In this solution, removal would predominantly be via the former  mechanism.This process is accompanied by Cr 3+ incorporation into the iron oxyhydroxide shell to form mixed Fe and Cr oxyhydroxides [60,61].This layer acts as an inhibitory layer on the iron surface [58,62], reducing the availability of surface adsorption sites and blocking Fe dissolution, and perhaps also inhibiting the formation of the voluminous corrosion products as observed in the structural experiments (and as such, the unaged nZVI would have the smallest SSA of all particles tested).This inhibition effect was observed in the dampened Cr(VI) solution geochemical response (as compared to the Cu and Zn solutions) on addition of unaged nZVI (Fig. S3 -Supporting information), although XRD scans of all post experimental nanomaterial indicated corrosion reactions had occurred (i.e.no Fe 0 remaining).The 2 h nanoparticles, with the largest surface area, removed chromate at the fastest rate.This, as stated, would be due to the 'primed' aqueous Fe in the solution, but may also be ascribed to the increase in SSA, substantiated by a previous study that found the second recharge of recycled nZVI (with the greatest surface area) reduced nitrate at the fastest rate of the six recharges tested [63].
As stated, for Cu, the initial uptake reaction was so rapid that a rate constant could not be calculated for the nZVI aged <3 d.However, like Cr, rate constants decreased with increasing nZVI age.The Cu uptake mechanism by nZVI has been reported to be surface reduction [30], and therefore a source of electrons must be present, as evidenced in all nZVI batch systems, irrespective of age.Unlike Cr, Cu does not become incorporated into the iron structure, and is susceptible to remobilisation, as evidenced in all nZVI batch systems, irrespective of age.Geochemical data (Figs.S3 and S4 -Supporting information) indicated that Cu remobilisation began after DO entered the system.The availability of oxygen determines the oxidation of Cu in aqueous solutions, with Cu generally assumed to be immune to corrosion by water itself [64].Therefore, oxygen ingress did not initiate desorption of Cu from the nZVI surface, which suggests that continuing electron transfer is essential for the Cu to remain in a reduced state when DO is present.
The removal mechanism of Zn by nZVI is thought to be by adsorption to the iron hydroxide shell, followed by zinc hydroxide complexation [65], and that this reaction includes electrostatic interactions and specific surface bonding, with no net electron transfer [30].The reduced contaminant uptake as compared to Cu suggests that this uptake mechanism requires greater, perhaps continued electron flow for Zn to associate with the nZVI surface, and that when electron flow ceases, the cation would be repelled as the Fe surface would be positively charged at the pH values encountered after $360 min [36,37].Remobilisation did not occur with the 2 w and 4 w nZVI.It is thought that >2 w, FeOOH is the dominant iron compound at the surface, which has been shown to have a high adsorption affinity for aqueous Zn [65].This suggests that perhaps two uptake mechanisms are occurring, one propagated by electron flow, and a second (more stable) Zn-FeOOH adsorption reaction.

Conclusions
i) Zero-valent iron nanoparticles retain a zero-valent core for over 4 w when aged in pure water; ii) The corrosion of nZVI is characterised by three periods of activity.Initially, (<1 d) a swift reaction occurs with the formation of large volumes of amorphous corrosion products and the emergence of magnetite.Following this, corrosion slows (<4 w), with the conversion of magnetite to maghemite and the formation of the metastable iron product lepidocrocite.After this period, nZVI corrosion was observed to be very slow, with increasingly small changes and the emergence of goethite, which is likely to be the final state of nZVI in pure solutions; iii) The aqueous ageing of nZVI has a significant negative impact on the rate constants of metal removal by nZVI, although the reduction in efficiency is specific to the contaminant uptake mechanism.This was attributable to a reduction in electron transfer, due either to a reduction in Fe 0 volume/Fe 2+ dissolution, or a thickening of the surface oxide/hydroxide layer hindering electron transfer.However, even after 4 w aqueous exposure, nZVI removed 36% Cr 6+ from a 20 ppm solution, and 55% and 31% of Cu and Zn respectively, from a 50 ppm solution; iv) Despite impressive initial uptake, both Cu and Zn subsequently remobilised from the surface of nZVI, ascribed foremost to the disappearance of the zero-valent core, and then subsequently by the ingress of DO.Zinc was observed to require greater electron flow to remain associated with the Fe surface than Cu.

Context and implications
The current work attempted to assess the longevity and corrosive transformation products of nZVI in the environment when they are utilised for the remediation of contamination.It appears that the 16 w experimental period was not of sufficient duration to assess the problem fully, but certain conclusions may be drawn: the ultimate iron corrosion product we ascribe as goethite, and the transformational process to reach this phase is significant, in the order of months, rather than days.The above suggests that the risk to receptors by nZVI may be a longer-term problem than previously thought.If deployed on site, would require periodic sampling of downstream boreholes to assess the migration of the reactive particles.However, reactivity is a loose term; a stricter definition of reactivity is needed to assess whether only transformation of nZVI to magnetite/maghemite would be considered reactive, or does this term include the further transformation to oxyhydroxides.This, obviously, will determine the regulatory defined length of any sampling regime.
An interesting finding in the current work is the continued reactivity of nZVI, even after 4 w of aqueous ageing.Furthermore, this experiment indicated that, with only a small fraction of the zerovalent core remaining, reductive reactions could still occur.This suggests that deployed nZVI could be suitable for the treatment of a migrating contaminant plume.Due to the significant desorption exhibited in both the Zn and Cu solutions, it is unlikely that nZVI would be employed for the long-term treatment of these contaminants in groundwaters readily supplied with atmospheric oxygen.Chromate, conversely, experienced no desorption, despite a fall in rate constant as the age of the nZVI increased.Therefore, nZVI could prove effective for the treatment of Cr 6+ plumes at most any subsurface depth for days or weeks after initial dosing, and as reactive rates exhibit a delay before activation, no reductive loss would be experienced if the particles were deployed as a slurry.In fact, a small amount of ageing proved advantageous for the treatment of this contaminant.
It is accepted that the above work has concentrated on geochemically simple waters, and does not take into account the influence of a more complex groundwater on nZVI corrosion.This suggests that a field-scale pilot-study should be undertaken prior to deployment to assess the influence of site geochemistry on the corrosion/reactivity of nZVI.

Fig. 5 .
Fig. 5. Comparison of first-order kinetics for the uptake of Cr(VI) (top left), Cu (top right) and Zn (bottom left) from solution when treated with aged nZVI.Key: X -Unaged, Open Square -2 h, Open Triangle -1 d, Open Circle -3 d, Closed Square -1 w, Closed Triangle -2 w, Closed Circle -4 w and Cross -control.

Table 1
Summary of published data on the effect of corrosion on the reactivity of nZVI.

Table 2
Identified iron forms, SSA values and XPS fitting of the Fe 2p 3/2 and O 1s spectra.Identified through fitting of the Fe 2p 3/2 XPS spectra (Figs.S1 and S2 -Supporting information).Due to limited experimental time, XANES analysis was not performed on nZVI aged for 8 w.This information is from XRD only.

Table 4
Maximum uptake and retention of metals by nZVI aged for differing times.

Table 3
Starting iron concentrations (mg L À1 ) for solutions of nZVI aged for differing times.