ARTICLE pubs.acs.org/JPCC
Nucleation and Growth of Fe Nanoparticles in SiO2: A TEM, XPS, and Fe L-Edge XANES Investigation Jer^ome Leveneur,†,‡ Geoffrey I. N. Waterhouse,‡,§ John Kennedy,*,†,§ James B. Metson,‡,§ and David R. G. Mitchell|| †
National Isotope Centre, GNS Science, 30 Gracefield Road, Lower Hutt, PO Box 31312, New Zealand School of Chemical Sciences, The University of Auckland, Private Bag 92019, Auckland, New Zealand § The MacDiarmid Institute for Advanced Materials and Nanotechnology, New Zealand Australian Centre for Microscopy and Microanalysis, University of Sydney, NSW 2006, Australia
)
‡
bS Supporting Information ABSTRACT: Magnetic iron nanoparticles embedded in insulating oxides matrices are prized targets for “on chip” magnetic sensors, nano fluxgates and nano hard magnets. In this study, the nucleation and growth of iron nanoparticles in the near surface region of 400 nm silica thin films (on silicon substrates) during ion implantation and post- implantation electron beam annealing was systematically investigated by transmission electron microscopy (TEM), X-ray photoelectron spectroscopy (XPS), and Fe L-edge X-ray absorption near edge spectroscopy (XANES). Results show the presence of Fe oxides after low-fluence low-energy ion implantation in SiO2, suggesting that initially Fe substitutes for Si in the silica matrix. Larger Fe fluences lead to the formation of sub-2 nm metallic Fe nuclei. Postimplantation annealing transformed the dispersed cationic Fe species into metallic Fe nanoclusters (diameter 1 10 nm) that are stabilized by a thin passivating surface oxide film. The versatility of ion implantation and electron beam annealing for the synthesis iron nanoparticles in silica matrices is demonstrated.
’ INTRODUCTION Metallic nanoparticles exhibit many remarkable and unique properties due to size-related quantum effects.1 For example, magnetic nanoparticles have been shown to exhibit enhanced magnetization and an absence of magnetization hysteresis. Such characteristics have stimulated interest for applications in magnetic sensing or magnetoresistance devices or in medicine for cell filtering or drug targeting.2 4 Magnetic nanoparticles may also be exploited in high-density magnetic data storage if the superparamagnetic limit, that is, the limit in nanoparticle size below which the remnant magnetization vanishes, can be overcome.5 Many chemical and physical approaches are effective for producing metallic nanoclusters such as: sol gel,6 magnetron sputtering,7 sonochemical synthesis,8 or ion beam synthesis.9,10 Recently, we reported that low-fluence low-energy iron implantation in silica (SiO2), followed by electron beam annealing, can be used to fabricate surface-embedded magnetic iron nanoparticles.11 These nanostructured materials showed very interesting magnetic and magnetoresistive properties.11,12 Structural changes occurring during the growth of nanoparticles were studied with atomic force microscopy (AFM), transmission electron microscopy (TEM), and Rutherford backscattering spectrometry (RBS). During growth, the iron nanoclusters were observed to evolve from a Fe-rich SiO2 amorphous phase with possible crystalline Fe nuclei to a Fe metal(core)/amorphous Fe oxide-rich(shell) structure to fully crystalline spherical metallic r 2011 American Chemical Society
Fe nanoparticles mostly at the silica surface. The mechanisms likely to be involved in iron nanopartucle evolution are shown schematically in Figure 1. The nanoparticle growth involved precipitation of Fe nuclei from a Fe:SiO2 solid solution coalescence to form a core/shell structure, phase changes leading to single crystal nanoparticles, and ripening. The ripening is likely to occur through Ostwald ripening or thermal coalescence.10,7 Two other mechanisms related to the annealing step were also observed: the electron beam etches the SiO2 layer resulting in iron nanoparticles protruding at the surface; a diffusion-sink phenomenon was observed with the fast diffusion and segregation at the SiO2/Si interface. In addition, the magnetic properties were observed to change in accordance with these structural changes. Therefore, the tailoring of the materials properties, and importantly the magnetic properties, could be achieved by controlling the structural properties. To optimize the structure and magnetic properties of ion beam synthesized Fe nanoclusters, an improved understanding of the growth mechanism of Fe nanoclusters during ion beam synthesis is required. Whereas the mechanisms of nanoparticle growth are now well-known for most liquid- and gas-phase preparation methods, the chemistry occurring in the solid state for methods such as ion Received: July 6, 2011 Revised: September 15, 2011 Published: September 21, 2011 20978
dx.doi.org/10.1021/jp206357c | J. Phys. Chem. C 2011, 115, 20978–20985
The Journal of Physical Chemistry C
ARTICLE
Figure 1. Schematic diagram of the mechanisms involved in the synthesis of the iron nanoparticles. Not to scale.
implantation is less well understood and demands further investigation. For instance, whereas the physical mechanism driving the growth of nanoclusters for long, high fluence implantation13 or during annealing has been described,14,15 the first steps of the growth mechanism during the implantation, including the nucleation, still remain elusive. Previous studies indicated that ion implantation in silica could result in the formation of oxides, silicides, or aggregates of the implanted species, depending on its electronegativity and fluence. Semiempirical studies indicated that electropositive implants (M) react chemically with oxygen atoms in the substrate structure to form M O bonds while simultaneously creating Si Si bonds at concentrations comparable to the implants. Conversely, electronegative implants (A) replace oxygen atoms in the substrate structure to form Si A bonds, liberating molecular O2 and peroxy radicals. Such implants may have weak chemical interactions and occur primarily as a neutral state as homomolecules or colloids.16 Another model emphasizes the importance of chemical interactions during ion implantation.17 Therefore, in the Fe:SiO2 system, the formation of Fe O bonds is expected to prevail over the Fe Si bonds. Studies of iron implantation into SiO2 suggest the formation of dispersed cationic iron during the implantation. In particular, medium energy Fe-implantation of SiO2 with fluencies above 1016 at cm 2 was shown to result in the formation of metallic Fe, isolated Fe2+, and both Fe2+ and Fe3+ possibly in Fe3O4 nanoparticles.13,18 Similar results were obtained with a ZnO substrate.19 It has been shown that Fe was observed in four different states depending on the fluence, temperature of implantation, and annealing parameters. Dispersed Fe2+ and Fe3+ are seen at low Fe concentrations and low processing temperatures, Fe2+ in FeZnO4 is seen at very high processing temperatures, whereas Fe metal is observed for higher fluence. In both cases, the neutrality of the system is assumed to be conserved by the formation of negative ions. In the case of the ZnO substrate, isolated Fe cations occupied tetrahedral sites substituting for Zn2+. More recently, a similar study was performed to study the Co/ ZnO system, where implanted Co2+ substituted tetrahedrally
coordinated Zn2+ and promoted nanocluster formation.20 Theoretical studies also predicted the formation of nuclei of a few angstroms in diameter after 1 s of implantation of 1.5 1012 at cm 2 s 1 of germanium in silicon oxide at 120 keV. (Germanium has a solubility limit of 1.01 108 at cm 3 in SiO2.)17 This study aims to provide a better understanding of the chemical interactions that occur between the ion beam implanted iron and the silica matrix that lead to the formation of metallic Fe or Fe oxides. We also examine the formation of Fe nanoclusters at the surface of SiO2 after low fluence ion implantation and electron beam annealing (EBA). Using TEM, X-ray photoelectron spectroscopy (XPS), and X-ray absorption near-edge spectroscopy (XANES), we demonstrate that ion implantation initially results in the formation of dilute cationic Fe2+ species and then, at higher dissolved iron concentrations, in the formation of small metallic nuclei, which will seed the growth of nanoclusters during prolonged implantation or annealing. A value for the critical concentration threshold above which Fe starts precipitating was established using two different photon energies for XPS measurements and knowing the precise probed depth of these two analyses. Surprisingly, the Fe metal nanoparticles formed exhibited remarkable oxidation resistance, likely because of a small passivating oxide shell that formed at the surface of the metallic nanoparticles.
’ METHODS Material Synthesis. The samples were fabricated using the same conditions and methods as those previously described.11,12 Low-energy ion implantation is used to incorporate iron at different concentrations up to 25 nm deep in a 400 nm silica matrix. The samples were then annealed using high-temperature (1000 °C) EBA for several different durations. Reference materials Fe, α-Fe2O3, and Fe3O4 were obtained from Aldrich. Samples were prepared for Synchrotron XPS and Fe L-edge and O K-edge XANES measurements by light Ar+ sputtering to remove surface contaminants. A 2 keV Ar+ gun 20979
dx.doi.org/10.1021/jp206357c |J. Phys. Chem. C 2011, 115, 20978–20985
The Journal of Physical Chemistry C
ARTICLE
Figure 2. Cross-section TEM image of the as-implanted surface prepared at a fluence of 1016 at cm 2.
under high vacuum (base ∼6 10 9 mbar, working ∼5 10 6 mbar) was used for that purpose. Oxide samples were subsequently heated under an oxygen flow to recover from the argoninduced reduction of the surface. The same Ar+ sputtering beam was used on some samples to investigate the electronic properties at different depths by removing the superficial layers. Analysis. RBS was performed systematically to verify the Fe concentration profile. The spectra were analyzed using the RUMP program.21 The position, crystalline order, and composition of the nanoclusters were examined using a JEOL 2010F TEM operating at 200 kV. X-ray photoelectron spectroscopy (XPS) data were collected using two different experimental setups. The first was a conventional lab instrument, a Kratos Axis UltraDLD equipped with a hemispherical electron energy analyzer. Spectra were excited using monochromatic Al Kα X-rays (1486.69 eV) with the X-ray source operating at 150 W. The hemispherical electron analyzer was set with a resolution of 0.1 eV with a pass energy of 40 eV and a dwell time of 0.136 s. This instrument illuminates a large area on the surface and then using hybrid magnetic and electrostatic lenses collects photoelectrons from a desired location on the surface. In this case, the analysis area was a 220 220 μm spot. The measurements were carried out in the normal emission geometry. A charge neutralization system was used to alleviate sample charge buildup, resulting in a shift of ∼3 eV to lower binding energy. The binding energy scale was corrected using adventitious hydrocarbon (C 1s = 285.0 eV). Survey scans were collected with 160 eV pass energy, whereas core level scans were collected with a pass energy of 20 eV. The analysis chamber was at pressures in the 10 9 Torr range throughout the data collection. Synchrotron XPS data was collected at a photon energy of 900 eV on the Soft-X-ray beamline at the Australian Synchrotron (AS). The hemispherical electron analyzer was set with a resolution of 0.01 eV with a pass energy of 20 eV and a dwell time of 0.1 s. Si L-edge, O K-edge, and Fe L-edge X-ray absorption near-edge spectra (XANES) were collected on the same beamline using the total electron yield (TEY) mode. The TEY absorption signal was normalized against the current induced on a gold mesh in the beam path as a function of the incident photon energy. The energies were calibrated using the Si 2p = 103.4 eV peak of SiO2, which showed no significant changes across the samples.
’ RESULTS AND DISCUSSION Ion Implantation Induced Nucleation. The initial objective of this study was to identify the depth distribution and speciation
Figure 3. Estimation of the Fe 2p XPS amplitude signal at different X-ray energies (full line) using Dynamic-TRIM simulated profile (theoretical) of the 1016 at cm 2 Fe-implanted sample. The squares shows the experimental Fe depth profile extracted from RBS spectrum using RUMP.
of iron in the silica substrate following ion implantation. Crosssection TEM analysis (Figure 2) revealed the presence of a Ferich layer 5 30 nm below the silica surface. No crystalline Fe or Fe oxide nanoparticles were identified by TEM, suggesting that the implanted iron was well-dispersed in the silica matrix. Ion implantation is a nonequilibrium process. Indeed, the incident ions can induce structural and chemical changes metastable with respect to thermal annealing, an effect that has long been used in the formation of metastable alloys.22 It is also known that ion implantation can lead to a supersaturated solid solution that will quickly form precipitates under annealing.23 In a crystal, the implanted atoms can take interstitial or substitutional positions, leading to different bonding environments in the crystal structure. In an amorphous material, implantation will induce changes in the short-range ordering due to point substitutions and formation of new bonds. RBS confirmed that the implanted Fe stays mostly within the first 25 nm of the silica (Figure 3), in agreement with previous work.11 The concentration of implanted iron reached a maximum ∼15 nm below the silica surface. To examine the speciation of the implanted iron, XPS measurements at two different X-ray photon energies (hυ = 1486.7 and 900 eV) and Fe L-edge XANES measurements were carried out on the samples. XPS is a powerful technique for nondestructive depth profile analyses as the escape depth of emitted photoelectrons, and hence the information depth can be changed by varying the source X-ray energy. Using the NIST Electron-Effective-Absorption-Length database (EAL)24 and XCOM: Photon Cross Sections Database,25 we evaluated the depth of the electrons emitted from interaction of the incident X-ray with the implanted iron using profiles produced by Dynamic-TRIM simulations.26 Note that the resolution of the standard RBS experiments could not be used to resolve variations at the nanometer level. Nevertheless, the total amount of Fe, depth of maximum concentration, and implanted range retrieved from the simulation were in good agreement with the values obtained from the RBS results. Assuming the homogeneous distribution of matter in the sample and low roughness of the surface, the number of electrons emitted from a particular depth z due to an interaction of an incident X-ray beam with every iron atoms can be approximated to N(z,Ep) = NFe(z)σ(Ep)p(z,EK), where NFe(z) is the number 20980
dx.doi.org/10.1021/jp206357c |J. Phys. Chem. C 2011, 115, 20978–20985
The Journal of Physical Chemistry C
ARTICLE
Figure 4. Synchrotron Fe 2p XPS (left) and Fe L-edge XANES (right) spectra for different as-implanted samples with different fluences (shown in at/cm2).
of atoms at a certain depth z in Fe at cm , σ(Ep) in cm at is the cross section for photoelectron emission at a photon energy EP retrieved from NIST XCOM database, and p(z,EK) is the probability of having an emitted electron to come from an interaction at a depth z retrieved from the depth distribution function of electrons emitted with the kinetic energy EK from photoelectric interaction in SiO2 using the NIST EAL database. This was modified from the first principle of quantitative XPS that gives an approximation of the measured photoelectron current from a homogeneous surface for a particular element.27 Figure 3 displays the results of the simulation for photon energies of 900 and 1486 eV for an Fe implantation at 15 keV and a fluence of 1016 Fe at cm 2. Lower energy X-rays will produce information from shallower depths in the sample. Whereas ∼99% of the measured Fe signal will come from the first 5 nm for 900 eV X-ray, ∼75% will come from the same layer using 1486.7 eV X-rays. For comparison, the probing depth for the total electron yield (TEY) XANES measurements can be evaluated similarly at ∼7 nm with 50% of the signal expected to be emitted from the upper 4 nm. Figure 4 shows synchrotron Fe 2p XPS (hυ = 900 eV) and Fe L-edge XANES data for samples implanted with iron at fluences of 5 1015, 1 1016, and 2 1016 at cm 2. The Fe 2p XPS spectra show features at 710.9 and 724.1 eV in a 2:1 peak area ratio, which are assigned to the Fe 2p3/2 and Fe 2p1/2 signals, respectively. The peak areas for the Fe 2p signals increase linearly with fluence. By comparison with the synchrotron XPS for Fe, Fe2O3, and Fe3O4 reference samples (Figure 5), which have Fe 2p3/2 binding energies of 707.1, 710.4, and 710.9 eV, respectively, we confirmed that Fe was present in an oxidized state after ion implantation. The presence of “shake up” satellites on the high binding energy side of the Fe 2p signals for the implanted samples validates this because such features are characteristic for Fe2+ or Fe3+. Whereas the Fe 2p chemical shift for the implanted samples would appear to suggest the presence of Fe3+, it must be remembered that the implanted Fe cations are dispersed in a silica matrix that is expected to change the Madelung potential around the Fe cations significantly compared with a bulk Fe oxide and as such affect the Fe 2p binding energies. Fe L-edge XANES measurements for the implanted samples indeed confirm that the samples contain Fe2+. XANES is a technique far more sensitive to 2
1
Figure 5. Synchrotron Fe 2p XPS (left) and Fe L-edge XANES (right) spectra for three references samples.
coordination and bonding environment than XPS as it probes the unoccupied electronic states of atoms and as such can reveal information about the crystal field (octahedral, square pyramidal, or tetrahedral) that the Fe cations occupy. Fe L-edge XANES spectra for the ion implanted samples are shown in Figure 4 and are dominated by a set of peaks around 708.2 eV (L3 region, 2p3/2 f 3d transitions) and another set of peaks around 721.5 eV (L2 region, 2p1/2 f 3d transitions). Corresponding Fe L-edge XANES spectra for the reference materials are shown in Figure 5. In the case of Fe3O4 and Fe2O3, the L2 and L3 features split further because of crystal field splitting effects (i.e., due to presence of nonequivalent t2g and eg orbitals). In α-Fe2O3, Fe3 + ions sit in octahedral sites. In the case of Fe3O4, half of Fe3+ ions sit in octahedral sites and half sit in tetrahedral sites, whereas all Fe2+ ions sit in octahedral sites.28 We noted the positions of the main Fe L3 peaks as 708 eV for Fe0, 710.1 eV for Fe2+, and 711 eV for Fe3+. The Fe L-edge XANES spectra for the as-implanted samples indicate majoritarly the presence Fe2+. In particular, those of the samples prepared at higher fluences are comparable to Fe L-edge literature data for FeO.29 This tends to indicate that the samples contain predominantly Fe2+ either from dispersed ionic Fe2+ or from Fe(II) in FeO. Nevertheless, the slight increase observed in the first peak (708.2 eV) relative to its shoulder (709.8 eV) could correspond to the presence of more Fe0 with increasing the fluence. This suggests that the formation of a metallic precipitate prevails over the formation of an oxide precipitate. However, the presence of FeO cannot be completely overruled. Minimal crystal field splitting of the Fe L-edge peaks could be seen for the ion implanted samples, thereby eliminating the possibility that these samples contained significant amounts of Fe3+. This work demonstrates the value of using both XPS and XANES in concert for this type of structural analysis. Figure 6 shows XPS data collected at a photon energy hυ = 1486.7 eV on a laboratory XPS source. At low ion fluences, the Fe 2p spectra show only the presence of Fe2+/Fe3+, as was seen in the XPS studies at hυ = 900 eV (Figure 4). At high fluences (2 1016 at cm 2), the presence of weak shoulders on the low binding energy side of Fe 2p oxide peaks can be seen. The binding energies of these features match perfectly those of the Fe metal reference sample (Figure 5). This clearly indicates that metallic iron is formed during ion implantation at high fluences. As shown in Figure 3, the depths of analysis for the different XPS 20981
dx.doi.org/10.1021/jp206357c |J. Phys. Chem. C 2011, 115, 20978–20985
The Journal of Physical Chemistry C
Figure 6. Lab Fe 2p XPS spectra (hν = 1486.7 eV) of as-implanted (top) with different fluences (shown in at cm 2) and 1 h 1000 °C EBA annealed (bottom) samples.
experiments differed slightly, with data collected at hυ = 1486.7 eV probing deeper in the bulk. A simplified profile of the oxidation state of Fe over depth can be deduced from these two energies measurements. From the surface to ∼5 nm depth, Fe is found nearly exclusively as Fe2+ possibly originating from dissolved Fe2+ sitting in tetrahedral sites substituting Si. The next 5 nm likely includes Fe2+ and Fe0 probably in the form of sub2 nm metallic nanoclusters. This is in accordance with the depth where different contrast can be observed by TEM imaging (Figure 2) and the XANES measurements. Indeed, we noted that the XANES probing depth, which is defined mostly by the escape depth of the Auger electron (651 eV), is similar to the lab XPS experiment. Therefore, it is understandable that both will probe the metallic phase deeper in the material compared with the 900 eV synchrotron XPS. Increasing the fluence will tend to bring the formation of Fe metal nanoclusters closer to the surface. The increase in the metallic iron shoulder between 1016 and 2 1016 at cm 2 determined in the lab XPS experiment (1486.7 eV) reflects this increase in Fe concentration nearer to the silica surface. These results are in accordance with previously published X-ray absorption and emission spectroscopy investigation of 100 keV and 1016 at cm 2 Fe implanted SiO2.30 Because the implantation is deeper and because the analyses cannot probe the region of higher Fe concentration, they have only depicted Fe2+. From the XPS analyses at two different photon energies (and hence information depths), it is possible to estimate the depth at which the concentration of Fe atoms exceeds the limit of solubility of iron in the SiO2 matrix: between 3.0 and 10.4 nm for the 1016 at cm 2 implanted sample. The lower value (3.0 nm) is the depth from which 90% of the expected signal for a photon energy of 900 eV is emitted, and the upper value (10.4 nm) is the depth from which 90% of the expected signal for a photon energy of 1486.7 eV is emitted. The depth where the Fe-rich region depicted using TEM is found (Figure 2) fits well within this range and lends further support to the hypothesis that the darker regions in the TEM image contain a metallic Fe phase. If we assume a uniform SiO2 matrix and the concentration profile of Fe atoms to match the simulated profile, then a value for the solubility limit of Fe in the SiO2 matrix can be found in the range 2.3 6.7 at %, (1.8 4.6) 1021 at cm 3, or 0.17 to 0.43 g cm 3. These values are above typical values found in the literature: 0.009 to 0.06 g cm 3 in the 500 600 °C temperature
ARTICLE
range.31,32 The large difference can be attributed to the fact that ion implantation might allow supersaturation of the solid solution. Indeed, the damage induced by the incident beam may also dissociate the Fe metal nuclei. Also, differences with the literature values are expected due to temperature differences, different hydration of the SiO2 matrix, and the use of ion implantation to dope the Fe as compared with diffusion or codeposition. Alternatively, one could consider the solubility limit presented in this Article as the high-temperature solubility limit. The local temperature during the solvation process could be as high as the “fictive temperature” Tf = 2727 °C, estimated in previous studies to evaluate the Gibbs energy of oxidation reaction during ion implantation in a silica matrix.33 This temperature was related to the “thermal spike” model of particle irradiation damages.34 Indeed, theoretical studies of ion/matter interactions often consider that every implanting ion brings the material temperature to very high values along the ion path for a very short time.35,36 In the present case, once the ion is stopped the temperature drops dramatically, which quenches the material, potentially allowing apparent supersaturation. Furthermore, the Fe solubility limit (1.4 1021 at cm 3), obtained from taking the theoretical value of Fe at the depth corresponding to the region of the Fe-rich layer furthest from the surface, is about half the value presented for the upper region (2.6 1021 at cm 3). The energy losses at this depth are known to be lower, causing less damage, and as a result the local temperature will be much smaller, thus allowing more nucleation. This indicates that in the first stage, ion implantation of Fe into silica tends to break SiO2 bonds to form “iron oxide-like” species, which remain soluble in the silica matrix. Si atoms, previously placed in the center of a tetrahedral surrounded by four O atoms, are substituted by Fe atoms. The most likely oxidation state for these is Fe2+. During the process, Si Si bonds are most probably created in concentrations similar to those of the implanted species, as suggested in previous studies.17 Increasing the concentration of Fe increases the probability of interaction between new implanted Fe ions with the dissolved Fe2+, which results in the formation of Fe Fe bonds and thus metallic Fe nuclei. Therefore, in the Fe-rich amorphous phase, larger fluences induce the formation of metallic nuclei. Increasing the temperature of the implantation is likely to increase further the solubility limit of Fe in SiO2 in silica. A binary diagram can be drawn on which the transition from Fe:SiO2 solid solution to the formation of a precipitate is shown (Figure 1). Such results have recently been evidenced using Monte Carlo simulation and high-resolution TEM studies.37 These simulations do not take into account the chemistry of the implanted species. Therefore, more accurate models and simulations could consider displacement energies and radii of capture of Fe atoms to form a cluster. These are collision cascade ballistic parameters, the values of which are related to the chemical potential for the formation of Fe Fe bonds from an oxide. In the extreme, very large fluence above 5 1016 at cm 2 was shown to produce large metallic nanoclusters on implantation.10 Effects of Annealing. TEM images revealed the structural changes occurring during the EBA. We observed that short electron beam annealing times at 1000 °C induce the formation of Fe-rich droplets containing a Fe metal core and possibly a Ferich amorphous oxide shell (Figure 7). Further annealing promotes the growth of the crystalline core until crystallization is complete. Whereas the Fe-rich region observed in as-implanted samples is located between 6 and 25 nm beneath the SiO2 surface 20982
dx.doi.org/10.1021/jp206357c |J. Phys. Chem. C 2011, 115, 20978–20985
The Journal of Physical Chemistry C
ARTICLE
Figure 7. Core shell structured nanoparticle. TEM image of sample prepared at a fluence of 15 keV 1016 at cm 2 and annealed under EBA at 1000 °C for 60 s. Figure 9. Evidence of the presence of a small oxide layer of surface nanoparticles. Fe L-edge XANES spectra for 1016 at cm 2 implanted and EBA annealed at 1000 °C for 1 h before and after 20 s Ar+ sputtering.
Figure 8. Fe 2p XPS spectra collected using a lab source (hν = 1486.7 eV) for ion-implanted samples (1016 at cm 2) subjected to electron beam annealing for different lengths of time.
(Figure 2), the nanoparticles observed after EBA are located just below the surface or on the surface.13 We have previously shown that the protruding Fe nanoclusters were the result of EBAinduced desorption of the SiO2 layer.11 XPS measurements were used to document the chemical state changes that occurred in the top 10 nm of samples during EBA. Lab XPS spectra (Figure 8) taken at 1486.7 eV indicate that the concentration of Fe in the near surface region of the samples first increases over the first 30s of annealing then decreases with prolonged annealing treatments. We interpret these results by suggesting that four phenomena are occurring: • EBA is selectively etching the SiO2 surface through desorption of SiO, resulting in an increase in Fe concentration near the surface; clear evidence for etching was provided by the RBS measurements.11 Etching of the outermost SiO2 layers moves the XPS experiment progressively toward the maximum Fe concentration of the ion implantation (initially 12 nm from the air/silica interface). • The concentration of Fe dissolved in silica by implantation (an average of 3 1021 at cm 3) greatly exceeds the solubility limit of 2 1015 at cm 3 at 1000 °C even close to the surface. Hence the formation of Fe clusters will reduce the concentration of dissolved Fe in the silica matrix back to a level below the solubility limit.11 Initially formed Fe metal nanoclusters with Fe oxide-rich shells morph into fully crystalline Fe metal nanoclusters over time, potentially
driven by the reduction of the oxide shell by excess Si atoms in the silica matrix. As explained above, Si Si bonds are formed during the implantation. In addition, the mechanisms driving the EBA-enhanced desorption of the silica layer also tend to provide an excess of Si atoms, through the desorption of oxygen atoms. Furthermore, EBA also provides a reducing environment, which can favor the formation of Fe metal. Indeed, whereas previous studies have shown that annealing under an oxygen-rich atmosphere13 leads to the formation of oxide nanoparticles, vacuum annealing11 or annealing under a hydrogen environment38 induces the formation of Fe-metal nanoparticles. Note that vacuum annealing leads to the formation of fewer nanoparticles with poorer crystalline state, highlighting the importance of the electron beam. A representative time temperature diagram is displayed in Figure 1 to illustrate the dynamics of the process. • Implanted Fe ions will diffuse inside the SiO2 matrix and toward the surface. Some Fe atoms might leave the sample under high-vacuum and high-temperature conditions, thereby reducing the concentration of surface Fe metal. • Fe nanoclusters at the gas/solid interface might get oxidized in air and develop a thin oxide shell. Absence of Deep Oxidation in Air. Evidence of a thin oxide shell on Fe nanoparticles grown on the silica surface after longer annealing is confirmed by the Fe L-edge XANES spectra of Figure 9, taken before and after mild Ar+ etching. Before etching the Fe L-edge spectrum contains contributions from both Fe metal and FeO or Fe3O4 (cf. Figure 5). By comparison, our TEM results indicated that these nanoparticles were mostly bcc Fe crystalline.11 Magnetic measurement also indicated that the Fe nanoparticles were mostly metallic iron.11,12 Brief Ar+ sputtering on the sample completely removed the oxide layer, giving rise to Fe L-edge spectrum consistent with metallic Fe (cf. Figure 5). This confirms the core shell structure of these nanoparticles, even though the nanoparticles were shown to be fully crystalline with TEM. The lattice fringes from the α-Fe core may have prevented the identification of a thin amorphous oxide shell in the TEM experiment. Nevertheless, the oxide layer on the α-Fe core is thin and likely only corresponds to several atomic oxide layers. Interestingly, contrary to bulk Fe, the oxide layer does not seem to grow thicker over time. The oxide shell thus appears to act simply as a passivating layer, preventing further oxidation. It is 20983
dx.doi.org/10.1021/jp206357c |J. Phys. Chem. C 2011, 115, 20978–20985
The Journal of Physical Chemistry C also possible that the stress induced by the small dimensions and spherical shape on the Fe crystal prevents the diffusion of O inside the nanoparticles in a process similar to self-limited oxidation of silicon wires.39 Such absence of oxidation has already been observed with other Fe nanoparticles systems and seems to be dependent on whether the synthesis was performed using a dry or wet fabrication method.40 42 Recent studies on cobalt nanoparticles oxidation and reduction also showed a very different behavior from bulk cobalt.43 Influence from finite-size effects and effects of the carbon surface used as a supporting layer were proposed to explain the observed differences.
’ CONCLUSIONS The ion implantation and EBA of Fe in silica have been systematically examined using a variety of complementary microscopic and X-ray spectroscopic techniques. The results provide new understanding of the mechanisms involved in the formation of Fe nanoclusters in a solid phase solution. It has been demonstrated that low fluences Fe implantation into silicon oxide first leads to dispersed Fe2+ species, possibly dissolved Fe2+ ions, and less likely in FeO nuclei. Metallic Fe nuclei are quickly formed from ballistic displacement-induced diffusion and thermal diffusion. Such a nucleation model could be used to develop more accurate simulations that will consider not only mechanical interaction but also chemical interactions. During annealing, Fe atoms aggregate around these nuclei forming Fe oxide-rich amorphous shells. Further annealing induces the reduction of the shells in favor of the growth of the α-Fe cores. The growth of Fe nanoclusters then continues via Ostwald ripening and coalescence. The use of EBA as an annealing method allows the production of surface and near-surface Fe nanoclusters without requiring large ion fluence during the implantation stage. Because some of the Fe nanoclusters are at the surface of the silica film, the nanoclusters will be in contact with the air and become partially oxidized at their surface. The thin oxide layer prevents further oxidation of the Fe nanoclusters, thus stabilizing the Fe nanoclusters. ’ ASSOCIATED CONTENT
bS
Supporting Information. XANES O K-edge and XPS O1s spectra for the reference samples and their analysis. Additional Si 2p spectra used for energy calibration are shown. Details from probing depth calculations of the different methods are given. This material is available free of charge via the Internet at http://pubs.acs.org.
’ ACKNOWLEDGMENT This work was supported by the Foundation for Research Science and Technology of New Zealand (C05X0802), AINSE (AINGRA08036), and the New Zealand Synchrotron Group. The assistance of Dr. Colin Doyle at the Research Centre for Surface and Materials Science, The University of Auckland, New Zealand is also acknowledged. This research was undertaken on the soft X-ray beamline at the Australian Synchrotron, Victoria, Australia. ’ REFERENCES (1) Burda, C.; Chen, X.; Narayanan, R.; El-Sayed, M. A. Chem. Rev. 2005, 105, 1025–1102. (2) Akinaga, H. Semicond. Sci. Technol. 2002, 17, 322–326.
ARTICLE
(3) Sieben, S.; Bergemann, C.; L€ubbe, A.; Brockmann, B.; Rescheleit, D. J. Magn. Magn. Mater. 2001, 225, 175–179. (4) Liu, Y; Wang, C.; Wei, Y.; Zhu, L.; Li, D; Jiang, J. S.; Markovic, N. M.; Stamenkovic, V. R.; Sun, S. Nano Lett. 2011, 11, 1614–1617. (5) Skumryev, V.; Stoyanov, S.; Zhang, Y.; Hadjipanayis, G.; Givord, D.; Nogues, J. Nature 2003, 423, 850–853. (6) Lu, Y.; Yin, Y.; Mayers, B. T.; Xia, Y. Nano Lett. 2002, 2, 183–186. (7) Ago, H.; Ayakagi, T.; Ogawa, Y.; Tsuji, M. J. Phys. Chem. C 2011, 115, 13247–13253. (8) Xu, H; Suslick, K. S. ACS Nano 2010, 4, 3209–3214. (9) Rodríguez-Iglesias, V.; Pe~ na-Rodríguez, O.; Silva-Pereyra, H. G.; Rodríguez-Fernandez, L.; Kellermann, G.; Cheang-Wong, J. C.; Crespo-Sosa, A.; Oliver, A. J. Phys. Chem. C 2010, 114, 746–751. (10) Lobotka, P.; Derer, J.; Vavra, I.; de Julian Fernandez, C.; Mattei, G.; Mazzoldi, P. Phys. Rev. B 2007, 75, 024423. (11) Kennedy, J.; Leveneur, J.; Williams, G. V. M.; Mitchell, D. R. G.; Markwitz, A. Nanotechnology 2011, 22, 115602. (12) Leveneur, J.; Kennedy, J.; Williams, G. V. M.; Metson, J.; Markwitz, A. Appl. Phys. Lett. 2011, 98, 053111. (13) Perez, A.; Treilleux, M.; Capra, T.; Griscom, D. L. J. Mater. Res. 1987, 2, 910–917. (14) Stepanov, A. L.; Khaibullin, I. B. Rev. Adv. Mater. Sci. 2005, 9, 109–129. (15) Dhara, S. Crit. Rev. Solid State Mater. Sci 2007, 32, 1–50. (16) Hosono, H. J. Non-Cryst. Solids 1995, 187, 457–472. (17) Cattaruzza, E. Nucl. Instrum. Methods Phys. Res., Sect. B 2000, 169, 141–155. (18) Nomura, K.; Reuther, H. J. Radioanal. Nucl. Chem. 2011, 287, 341. (19) Zhou, S.; Potzger, K.; Talut, G.; Reuther, H.; von Borany, J.; Gr€otzschel, R.; Skorupa, W.; Helm, M.; Fassbender, J.; Volbers, N.; Lorenz, M.; Herrmannsd€orfer, T. J. Appl. Phys. 2008, 103, 023902. (20) Ney, A.; Opel, M.; Kaspar, T. C.; Ney, V.; Ye, S.; Ollefs, K.; Kammermeier, T.; Bauer, S; Nielsen, K.-W.; Goennenwein, S. T. B.; Engelhard, M. H.; Zhou, S.; Potzger, K.; Simon, J.; Mader, W.; Heald, S. M.; Cezar, J. C.; Wilhelm, F.; Rogalev, A.; Gross, R.; Chambers, S. A. New J. Phys. 2010, 12, 013020. (21) Doolittle, L. R. Nucl. Instrum. Methods Phys. Res., Sect. B 1985, 9, 344. (22) Borders, J. A. Annu. Rev. Mater. Sci. 1979, 9, 313–339. (23) Luce, F. P.; Kremer, F.; Reboh, S.; Fabrim, Z. E.; Sanchez, D. F.; Zawislak, F. C.; Fichtner, P. F. P. J. Appl. Phys. 2011, 109, 014320. (24) Powell, C. J; Jablonski, A. NIST Electron Effective-AttenuationLength Database, version 1.2; National Institute of Standards and Technology: Gaithersburg, MD, 2001. (25) Berger, M. J.; Hubbell, J. H. NIST XCOM: Photon Cross Sections Database, version 3.1; National Institute of Standards and Technology: Gaithersburg, MD, 2001. (26) Chakarov, I. R.; Todorov, S. S.; Karpuzov, D. S. Nucl. Instrum. Methods Phys. Res., Sect. B 1992, 69, 193–199. (27) Grant, J. T. Surf. Interface Anal. 1989, 14, 271–283. (28) Crocombette, J. P.; Pollak, M.; Jollet, F.; Thromat, N.; GautierSoyer, M. Phys. Rev. B 1995, 52, 3143–3150. (29) Regan, T. J.; Ohldag, H.; Stamm, C.; Nolting, F.; L€uning, J.; St€ ohr, J.; White, R. L. Phys. Rev. B 2001, 64, 214422. . Z.; Zatsepin, D. A.; Cholakh, S. O.; Schmidt, B.; (30) Kurmaev, E Harada, Y.; Tokushima, T.; Osawa, H.; Shin, S.; Takeuchi, T. Phys. Solid State 2005, 47, 754. (31) Atkinson, A.; Gardner, J. W. Corros. Sci. 1981, 21, 49–58. (32) Ramappa, D. A.; Henley, W. B. J. Electrochem. Soc. 1999, 146, 3773–3777. (33) Hideo, H. Jpn. J. Appl. Phys. 2007, 32, 3892–3894. (34) Primak, W.; Edwards, E.; Keiffer, D.; Szymanski, H. Phys. Rev. 1964, 133, A531–A535. (35) Ossi, P. M.; Pastorelli, R. J. Appl. Phys. 1999, 85, 1387–1394. (36) Hsieh, H.; Diaz de la Rubia, T.; Averback, R. S. Phys. Rev. B 1989, 40, 9986–9988. 20984
dx.doi.org/10.1021/jp206357c |J. Phys. Chem. C 2011, 115, 20978–20985
The Journal of Physical Chemistry C
ARTICLE
(37) Yuan, C. W.; Yi, D. O.; Sharp, I. D.; Shin, S. J.; Liao, C. Y.; Guzman, J.; Ager, J. W., III; Haller, E. E.; Chrzan, D. C. Phys. Rev. B 2009, 80, 134121. (38) Zhang, G. L.; Liu, W. H.; Hy, W. X. Appl. Phys. Lett. 1992, 61, 2527. (39) Fazzini, P. F.; Bonafos, C.; Hubert, A.; Colonna, J.-P.; Ernst, T.; Respaud, M.; Gloux, F. Mater. Res. Soc. Symp. Proc. 2009, 1144, 117–122. (40) Wang, C.; Bae, D. R.; Amoette, J. E.; Engelhard, M. H.; Antony., J.; Qiang, Y. J. Am. Chem. Soc. 2009, 131, 8824–8832. (41) Kim, H.-S.; Ahn, J.-Y.; Hwang, K.-Y.; Kim, I.-K.; Hwang, I. Environ. Sci. Technol. 2010, 44, 1760–1766. (42) Simeonidis, K.; Mourdikoudis, S.; Tsiaoussis, I.; Frangis, N.; Angelakeris, M.; Kalogirou, O. Mod. Phys. Lett. B 2007, 21, 1143–1151. (43) Papaefthimiou, V.; Dintzer, T.; Dupuis, V.; Tamion, A.; Tournus, F.; Hillion, A; Teschner, D.; H€avecker, M.; Knop-Gericke, A.; Schl€ogl, R.; Zafeiratos, S. ACS Nano 2011, 5, 2182–219.
20985
dx.doi.org/10.1021/jp206357c |J. Phys. Chem. C 2011, 115, 20978–20985
