The effects of vacuum annealing on the structure and surface chemistry of iron nanoparticles
- First Online:
- Cite this article as:
- Scott, T.B., Dickinson, M., Crane, R.A. et al. J Nanopart Res (2010) 12: 1765. doi:10.1007/s11051-009-9732-9
- 296 Views
In order to increase the longevity of contaminant retention, a method is sought to improve the corrosion resistance of iron nanoparticles (INP) used for remediation of contaminated water and thereby extend their industrial lifetime. A multi-disciplinary approach was used to investigate changes induced by vacuum annealing (<5 × 10−8 mbar) at 500 °C on the bulk and surface chemistry of INP. The particle size did not change significantly as a result of annealing but the surface oxide thickness decreased from an average of 3–4 nm to 2 nm. BET analysis recorded a decrease in INP surface area from 19.0 to 4.8 m2 g−1, consistent with scanning electron microscopy (SEM) and transmission electron microscopy (TEM) observations which indicated the diffusion bonding of previously discrete particles at points of contact. X-ray diffraction (XRD) confirmed that recrystallisation of the metallic cores had occurred, converting a significant fraction of poorly crystalline iron to bcc α-Fe and Fe2B phases. X-ray photoelectron spectroscopy (XPS) indicated a change in the surface oxide stoichiometry from magnetite (Fe3O4) towards wüstite (FeO) and the migration of boron and carbon to the particle surfaces. The improved core crystallinity and the presence of passivating impurity phases at the INP surfaces may act to improve the corrosion resistance and reactive lifespan of the vacuum annealed INP for environmental applications.
KeywordsIronNanoparticlesVacuum annealXPSEnvironmental remediation
An emerging and rapidly developing technology for the remediation of contaminated water is the use of iron nanoparticles (INP). Nanoparticles, by definition, having diameter between 1 and 100 nm are characterised by a high surface area to volume ratio and high surface energies (Zhang et al. 1998). The small size, high reactivity and versatile applicability of nanoscale iron provide a cost-effective tool for the treatment of contaminated industrial- or ground waters. The suitability of zero-valent iron (through its corrosion mechanisms) for the remediation of contaminated water and soil environments has long been known; see Bigg and Judd (2000) for a review of applications through the years. When compared to surfaces of bulk scrap metal or iron filings/granules, nanoscale iron offers a much higher reactivity, and many studies have concluded that INP degrade contaminants more rapidly than the aforementioned forms of zero-valent iron (Choe et al. 2000; Glazier et al. 2003; Li and Zhang 2007; Lien and Zhang 1999, 2001; Ponder et al. 2000; Schrick et al. 2002; Wang and Zhang 1997).
In recent years, INP have been shown to be effective remediators of a range of contaminants, including heavy metals (Alowitz and Scherer 2002; Li and Zhang 2007; Ponder et al. 2000, 2001; Shimotori et al. 2004), radionuclides (Scott 2005; Riba et al. 2008; Dickinson et al. in preparation), chlorinated organics (Cheng et al. 2007; Elliott and Zhang 2001; Lien and Zhang 1999, 2001; Liu et al. 2005; Nurmi et al. 2005; Schrick et al. 2002; Wang and Zhang 1997; Zhang et al. 1998), inorganic anions (Alowitz and Scherer 2002; Cao et al. 2005; Choe et al. 2000; Kanel et al. 2005; Mondal et al. 2004) and other harmful chemicals (Joo et al. 2005; Miehr et al. 2004; Nurmi et al. 2005; Shimotori et al. 2004; Zhang 2003). The mechanism of remediation varies depending on the nature of the contaminant (Miehr et al. 2004); for example, the oxidation of Fe(0) drives the reductive transformation of chlorinated organics, such as the carcinogenic trichloroethylene (TCE), to relatively innocuous hydrocarbons. For waters containing heavy metals and/or radionuclides, decontamination occurs via sorption and/or reduction onto the surface of the iron (Riba et al. 2008). In a study by Li and Zhang (2007), it was demonstrated that for metal ions such as Zn(II) and Cd(II) which have standard potentials (E0) very close to, or more negative than, that of iron (−0.41 V), the removal mechanism is sorption/surface complexation, whilst for metals such as Cu(II), Ag(I) and Hg(II) which have E0 much more positive than iron, removal occurs predominantly via reduction. For those metals with E0 only slightly more positive than iron, e.g. Ni(II) and Pb(II), both sorption and reduction have been shown to occur at the iron surface. The resulting nanoparticulate-contaminant solids can be removed from the water via filtration, settlement or exploitation of the INP’s magnetic properties.
In previous work investigating the uptake mechanisms of uranyl onto the INP (Riba et al. 2008, Dickinson et al. in preparation), it was noted that after a 48-h reaction period the INP displayed partial dissolution in aqueous solutions. A simultaneous, partial re-oxidation of the surface sorbed/reduced U(IV) to U(VI) was also observed. In order to improve the longevity of contaminant retention, a method is sought to improve the INPs’ resistance to corrosion and thereby extend their industrial lifetime. It is hypothesised that the rapid “flash” formation of the INP via the borohydride reduction method leads to a highly disordered crystalline structure of both the bulk metallic Fe and the surface oxide layer. Numerous impurities and defects are present in unaltered metals and are known to significantly impact the reactivity of a material (Bonin et al. 2000). Thermal treatment, or annealing, is commonly used in metallurgy to relieve internal stresses, refine the grain structure and produce equilibrium conditions within a metal. To date, the effect of annealing on the physical and chemical properties of the INP, as well any implication for their remediation abilities has not been thoroughly investigated. In a recent study by Liu et al. (2005), annealing the INP prior to use for the remediation of TCE lowered their efficacy, as the improved crystallinity and resistance to corrosion hindered the activation of H2 for TCE dechlorination. However, it is postulated that annealing should have the opposite effect on the remediation efficacy of INP when the contaminant of interest is a heavy metal or radionuclide.
The aims of this study were to investigate the effect that vacuum annealing has on the bulk and the surface chemistry of nanoscale iron. In order to gain a thorough understanding of the processes involved, a multi-disciplinary approach is used to study the structural and chemical changes induced in the INP by annealing within a vacuum. INP and vacuum-annealed iron nanoparticles (VAINP) were analysed using disc centrifugation to determine the particle size distribution, X-ray diffraction (XRD) to determine the changes in bulk structure; X-ray photoelectron spectroscopy (XPS) and BET to examine the changes in the surface chemistry; and scanning electron microscopy (SEM) and transmission electron microscopy (TEM) to study the form, size and crystallinity. In order to determine the rate of change invoked by vacuum annealing, sequential XPS analysis was conducted whilst annealing the INP on a heat stage within an XPS instrument.
In a concurrent study (Dickinson et al. in preparation), annealing was demonstrated to have no adverse effects on the remediation abilities of the INP for the removal of uranium from solution. However, their resistance to corrosion was greatly improved, with the Fe dissolution rate reduced to a third of that measured for non-annealed INP.
Materials and methods
All chemicals [iron sulphate (FeSO4∙7H2O), sodium hydroxide (NaOH), sodium borohydride (NaBH4)] and solvents (ethanol, acetone) used in this study were of analytical grade, and all solutions were prepared using Milli-Q purified water (resistivity >18.2 MΩ cm).
Nanoparticle synthesis and preparation
INP were synthesised using sodium borohydride to reduce ferrous iron to a metallic state, following an adaptation of the method described by Wang and Zhang (1997). In brief, 7.65 g of FeSO4∙7H2O was dissolved in 50 mL of Milli-Q water (18.2 MΩ cm) and a 4 M NaOH solution was used to adjust the pH to the range 6.2–7. The salts were reduced to metallic nanoparticles by the addition of 3.0 g of NaBH4. The nanoparticle product was isolated through centrifugation and then sequentially washed with water, ethanol and acetone (20 mL of each). The nanoparticles were dried in a desiccator under low vacuum (approx. 10−2 mbar) for 48 h and then stored in the oxygen-free environment of a Saffron Scientific glovebox until required.
In order to enable XPS analysis of the INP to be performed before, during and after annealing the INP, ~0.002 g of material was lightly packed into a 100 μm deep cavity on the surface of a sapphire substrate which was then mounted on the Peltier heat stage of a Thermo Fisher Scientific Escascope, operating at <5 × 10−8 mbar. The instrument is equipped with a dual anode X-ray source (Al Kα 1486.6 eV and Mg Kα 1253.6 eV). The INP were pre-heated at 75 °C overnight to remove surface-sorbed water and then subsequently vacuum annealed for 48 h at 500 °C. During the annealing period, XPS profiles for iron, carbon, oxygen and boron were acquired approximately every hour to determine surface dynamics to 5 nm depth. Al Kα radiation was used at 400 W (15 kV, 23 mA) to generate high resolution scans, using a 30 eV pass energy and 300 ms dwell times. Profiles were collected and analysed using Pisces software (Dayta Systems Ltd, c/o Interface Analysis Centre, http://daytasystems.co.uk) and were corrected to the binding energy of the adventitious carbon (284.8 eV). Backgrounds were subtracted from the spectral range around the peak(s) and curve fitted to χ2 values <2.
A Phillips Xpert Pro diffractometer with a Cu Kα radiation source (λ = 1.5406 Å) was used for XRD analysis (generator voltage of 40 keV; tube current of 30 mA). XRD spectra were acquired over a 2θ range of 0°–90°, with a 0.02° step size and a 12 s dwell time. Analysis was performed on both INP and VAINP to determine bulk changes. Samples were prepared by pipetting the appropriate amount of an acetone/INP suspension onto a glass microscope slide and allowing the acetone to dry prior to analysis.
Morphological analysis of the nanoparticles was performed on a Zeiss NVision 40 Cross Beam system. The dual SEM and FIB capabilities of the system allow for high resolution imaging and precision milling of samples. The SEM column was operated over a range of voltages from 5 kV to 500 V, allowing for various levels of surface sensitivity. For higher resolution at lower accelerating voltage, a working distance of 2–3 mm was used. High resolution images were acquired using both inlens secondary electron and energy selective backscatter (ESB) detectors. The inlens detector allows high sensitivity at extremely short working distances, whereas the ESB detector, which is also situated in the SEM column, features a filter to isolate electrons backscattered from the sample surface from the secondary electrons produced by the beam. The ESB highlights chemical (z) contrast. Samples were prepared by pipetting the appropriate amount of an acetone/INP suspension onto a Si substrate and allowing the acetone to dry prior to analysis.
TEM images and electron diffraction patterns were obtained with a Philips EM 430 TEM operating at 250 keV. A selected area aperture was used to pick out a cluster of nanoparticles, the image recorded and diffraction patterns acquired from the selected area. Nanoparticle samples were mounted on 200 mesh copper grids.
BET-specific surface area was measured using the nitrogen adsorption method with a Micrometrics Gemini VI Surface Area and Pore Size Analyser. Approximately 0.3 g of nanoparticle material was weighed out into a glass tube using a Sartorius four decimal place balance. The particles were flushed with nitrogen for 30 min and then reweighed.
Particle size distribution was measured using a CPS Disc Centrifuge, Model DC24000, CPS Instruments, Inc. In order to prepare the samples for analysis, 0.05 g of nanoparticles was added to 10 mL of pure water and sonicated for 15 min. A 0.1 mL sample of this colloid was drawn into a plastic pipette and injected into the CPS disc centrifuge through the injection point.
Comparison of bulk structures
The observed sizes of the INP and VAINP are consistent with size distribution measurements performed using a CPS Disc Centrifuge (Barnes in preparation), in which 85% of the total INP were in the size range 0–63 nm, 8% were in the range 63–100 nm and only 7% were >100 nm. The distributions for VAINP were comparable at 84, 7 and 9%, respectively.
Using the SEM and ESB electron detector on the NVision system, the density contrast between the metallic core and oxide shell in the INP was readily identified (Fig. 1c, d). Similarly, the outer layer of the VAINP was distinct from the denser metallic cores. However, the boundaries between adjoining individual VAINP were not so easily determined and implied the possible bonding of some particles previously aligned ‘end to end’ by magnetic attractions.
In order to investigate the apparent particle bonding in greater detail, TEM analysis was used. Images recorded of INP (Fig. 1e) showed that the cores of individual particles were metallic surrounded with an apparently amorphous layer of thickness 3–4 nm, ascribed to surface oxide coverage. In the TEM images, dark mottles visible within the metallic cores indicated that individual particles were either polycrystalline or comprised isolated metal crystals in an otherwise amorphous matrix.
TEM images of the VAINP (Fig. 1f) clearly indicated the presence of particles bonded together at points of contact. Metal grains within the particle cores (highlighted by differential contrasting) were visibly larger than those observed in the INP and in some cases were clearly observed to bridge the boundary between previously discrete particles. Consequently, the surface oxide layer (~2 nm thick) around these particles was discontinuous at the point of bonding. Electron diffraction performed in the TEM confirmed that the grains in the cores of the INP and VAINP were metallic α-Fe with bcc structure.
Additional features recorded in the XRD spectra of VAINP were reflections at 43°, 56.5°and 79.5° 2θ, Fig. 2. These peaks are attributed to the presence of Fe2B phases (after Krämer et al. 1994 and Dehlinger et al. 2003); these phases were also observed by Liu et al. (2005) following the annealing of Fe nanoparticles produced using the borohydride synthesis method. The presence of this Fe2B phase is ascribed to the crystallisation of the small amounts of B, originally incorporated into the INP as an impurity during aqueous synthesis, within the Fe lattice. It is known from previous analyses that B can account for up to 7% of the total weight of the INP (Barnes in preparation).
Comparison of surface chemistry
BET analysis indicated that the surface of the INP was markedly greater than the VAINP, recording values of 19.0 and 4.8 m2g−1, respectively (Barnes in preparation). This is in good agreement with the TEM observations which indicate the bonding of individual particles at points of contact, thereby decreasing the total surface area. Assuming the nanoparticles to be spherical, the equivalent particle diameter of the as-synthesised powder as calculated from the BET specific surface area is about 40.1 nm, consistent with the nanoparticles being polycrystalline aggregates. The BET calculated particle diameter of the annealed powder was 158.8 nm. This is significantly larger than indicated by both electron microscopy and disc centrifuge measurements which indicated that particle size remained roughly the same during the annealing process. Assuming that each INP particle is a polycrystalline aggregate, the significant decrease in surface area recorded for VAINP indicates that grain growth and recrystallisation within each particle must significantly decrease the surface roughness and grain boundary area, whilst not significantly impacting the overall particle dimensions. It should also be noted that the BET-based particle diameter calculation for the VAINP is also offset by reduction in surface area caused by particle to particle bonding.
Differences in the surface chemistry of the INP and VAINP were determined using XPS analysis that provided chemical characterisation to a depth of 4–5 nm. Of particular interest were the recorded surface proportions and chemical states of Fe, O, C and B.
Binding Energy (eV)
Allen et al. (1974)
CoFe2O4 and Fe3O4
CoFe2O4 and Fe3O4
CoFe2O4 and Fe3O4
McIntyre and Zetaruk (1977)
NiFe2O4 and Fe3O4
McIntyre and Zetaruk (1977)
NiFe2O4 and Fe3O4
McIntyre and Zetaruk (1977)
A summary of the experimental results regarding the bulk and surface characteristics of the INP and VAINP
Particle size distribution
Crystalline (α-Fe, Fe2B)
Crystallite dimensions (nm)
Oxide thickness (nm)
BET surface area (m2/g)
(Fe(0)/Fe2+ + Fe3+)
From an examination of the O1s spectra (not shown), the pre-treatment at 75° was observed to decrease the amount of chemi- and physi-sorbed waters on the particle surfaces; physi-sorbed water was observed to disappear after a period of 2 h, whilst chemi-sorbed water was observed to decrease over the pre-treatment period but the signal was never completely eradicated. The persistence of the signal from chemi-sorbed water indicates the presence of OH− groups as an integrated component of the surface oxide structure, rather than a surface-sorbed layer. As a result of vacuum annealing, the recorded O1s peaks (ascribed to O in the surface oxide) were observed to shift to from 530.4 eV to a lower binding energy of 530.0 eV, indicating a change in oxide stoichiometry. The typical O1s binding energies for magnetite and wüstite are 530.0 and 529.8 eV, respectively (McIntyre and Zetaruk 1977), and so the data represent a shift consistent with a move to a more Fe2+ based oxide.
In order to compare the physical and chemical properties of the INP before and after vacuum annealing at 500 °C, summarised in Table 2, it is necessary to consider the effects of the heating process in the absence of any significant oxygen. When the INP are supplied with energy in the form of heat it is diffusion related processes that allow the reorganisation of the metallic cores towards an equilibrium state, but without the incorporation of additional oxygen into the structure.
It is generally accepted that the three characteristic stages of annealing include recovery, recrystallisation and grain growth (Callister 2003). During the first of these stages, recovery, an increased rate of atomic diffusion at the elevated temperature allows a proportion of stored internal strain energy to be relieved by dislocation motion. However, an important factor to consider in INP is that a significant proportion of the volume of a nanoparticle consists of grain boundaries, which can be as much as 30% by volume (Flewitt PEJ private communications). This means that during heating the recovery period is very rapid because dislocations have short distances over which to propagate before reaching grain boundaries. TEM and XRD data provide clear indication that the current annealing process resulted in recrystallisation of metal grains within the INP cores to produce what are presumably strain-free equi-axed grains characterised by low-dislocation density. Recrystallisation is driven by the difference in internal energy between strained and unstrained materials. The new grains form as very small nuclei and grow to completely consume the parent material, via short-range diffusion. Grain growth is also apparent in TEM images and occurs to minimise interfacial energy by reducing boundary area. Grains are assumed to grow via short range diffusion and in the case of particles adjoining each other due to magnetic attraction this diffusion process allowed metallic bonding between particles. In order for this bonding to happen, diffusion of both O and Fe must occur to displace the surface oxide of these points of contact.
In the current INP system, the primary impurity species is B which becomes incorporated into the metallic structure during the manufacturing process. XPS data indicate that both B and C (minor impurity species) are moved to the particle surfaces during annealing; this phenomenon is assumed to occur via diffusion to grain boundary surfaces. This observation correlates well with B and C solubility studies in α-Fe, that indicate relatively low solubility of both species, <0.0018 at.% (Lucci and Venturello 1971) and ≤0.018 at.% (Merlin et al. 2004), respectively, at 500 °C. The formation of distinct separate phases is evidenced in the XRD data, which show formation of Fe2B phases, and the XPS data, which clearly show the presence of carbide at the particle surfaces. Due to the similar cationic radii and bonding behaviour of B and C, it is possible that C may substitute for B in the Fe2B phase (Koifman et al. 1969). Although the presence of carbide is clearly indicated by XPS, it would be at volumes too low to be detected by XRD.
Another significant effect of the vacuum heat treatment is on the oxide surface of the particle. Similar to the metal diffusion processes on the bulk, it will result in diffusion-related alteration of the oxide structure. This is clearly evidenced in the XPS results, which show a decrease of the surface oxide thickness and a change in its stoichiometry, an increased proportion of Fe(0) to Fe oxides, as well as an increase in the Fe2+:Fe3+ ratio. Given the melting point of magnetite is 1583–1597 °C (Cornell and Schwertmann 2003), the observed loss of surface oxide is attributed to the diffusion of oxygen into the metallic core with a resultant partial decomposition of the oxide. By comparing the ratios of metallic iron to oxide as annealing progressed, it was evident that diffusion occurred most rapidly in the initial stages of annealing (0–4 h).
The solubility of O in α-Fe has been reported previously by Wriedt (1991) as 0.0024 at.%, and the persistence of oxide on the surface during the annealing period indicates that oxygen saturation of the metal had likely occurred. In a bulk material, it would be expected that the surface oxide would completely disproportionate under the same reaction mechanisms to leave a pristine metal surface. The BET results also indicate a reordering and optimisation of the surface area of the INP. In aqueous conditions, the reactivity of the VAINP is expected to be different to the INP. Corrosion of the particle surfaces is expected to proceed more rapidly for INP due to the greater surface area and greater density of surface defects and dislocations. Any galvanic phenomena that may assist with surface redox reactions would be expected to be enhanced in the VAINP due to the optimised structure of the magnetite overlayer. In using elemental metals for remediation of contaminants in groundwater, a key aspect of the decontamination is that reactions are mediated by the availability of electrons supplied by the metal, either directly at the metal surface or through conductive oxide films such as magnetite Fe3O4. Electron availability may be limited by the presence of non-conductive surface phases such as Fe2O3 and FeOOH which will inhibit electron transfer. Vacuum annealing is observed to promote the electron transfer from the metal to the Fe3+ on the surface and thus increase its reactivity (Moura et al. 2005).
However, reductive removal of contaminants can only proceed once contaminant species are in direct contact with the reactive surfaces. This necessitates that any of the surface reaction must be preceded by adsorption of a contaminant ion to an available surface site. The reduction in surface area induced by vacuum annealing, and presence of B and C phases at the INP surfaces, would proportionally reduce the number of available surface adsorption sites and thus would be expected to reduce the mass of contaminants remediated on the particle surfaces. Whether there exists a play off between the smaller surface area and the enhanced galvanic capability on the reactivity of the VAINP in relation to INP remains to be determined. A concurrent study is being performed to determine the comparative efficacy of the INP and VAINP for the remediation of U-contaminated solutions, with a view to improving their industrial application and reactive lifetime.
Vacuum annealing at 500 °C has been demonstrated to induce structural and chemical changes in INP which are considered to alter the subsequent reactivity with aqueous contaminants. XRD and TEM data indicate reordering and recrystallisation of the bulk metallic cores, an increase in grain size and elimination of defects and impurities. XPS and BET indicate significant alteration of the particle surfaces involving reduction of surface area, thinning and dehydration of the surface oxide and migration of impurities (B and C) to surfaces. The corrosion resistance and longevity imbued by vacuum annealing of the nanoparticles material may prove favourable for applying remediation technologies more effectively in natural and industrial situations.
We thank Dr. Roger Vincent, Department of Physics, University of Bristol, for performing the TEM analysis and Prof. Peter Flewitt for valuable discussion.