Radiation damage from long-term alpha particle bombardment of silicates – a microfocus XRD and Fe K-edge XANES study

Abstract A detailed understanding of the response of mineral phases to the radiation fields experienced in a geological disposal facility (GDF) is currently poorly constrained. Prolongued ion irradiation has the potential to affect both the physical integrity and oxidation state of materials and therefore may alter a structure's ability to react with radionuclides. Radiohalos (spheres of radiation damage in minerals surrounding radioactive (α-emitting) inclusions) provide useful analogues for studying long term α-particle damage accumulation. In this study, silicate minerals adjacent to Th- and U-rich monazite and zircon were probed for redox changes and long/short range disorder using microfocus X-ray absorption spectroscopy (XAS) and high resolution X-ray diffraction (XRD) at Beamline I18, Diamond Light Source. Fe3+ → Fe2+ reduction has been demonstrated in an amphibole sample containing structural OH– groups – a trend not observed in anhydrous phases such as garnet. Coincident with the findings of Pattrick et al. (2013), the radiolytic breakdown of OH– groups is postulated to liberate Fe3+ reducing electrons. Across all samples, high point defect densities and minor lattice aberrations are apparent adjacent to the radioactive inclusion, demonstrated by micro-XRD.


Introduction
WITH the majority of the global stockpiles of radioactive waste destined for long-term (>100,000 years) isolation in a deep geological disposal facility (GDF), a critical understanding of the mechanisms and consequences of radiation damage across mineral phases is essential for building a safety case. Regardless of lithology, the host rock will be expected to act as the final barrier towards radionuclide migration following eventual waste canister failure (NDA, 2010a,b) and radiation damage will occur where escaping radionuclides accumulate. Additionally, the performance assessment of near-field barrier materials (i.e. bentonite backfill, cements, etc.) and the wasteforms themselves employed in a GDF will be aided by detailed studies of radiation damage effects in minerals (Ewing, 2001). Natural mineral assemblages containing radioactive inclusions provide useful proxies for the response of silicate phases to prolongued α-irradiation. Aureoles of radiation damage (r = ∼30-50 μm) form in minerals that surround α-particle emitting inclusions [e.g. Th-rich monazite (CePO 4 ), U-rich zircon (ZrSiO 4 )], resulting from the high-energy α-particles ( 4 He 2+ ions) penetrating into the neighbouring crystal and causing hundreds of atomic displacements at the end of their projected range. An α-particle will lose most of its energy via ionization of the structure through which it penetrates, eventually resulting in a concentration of Frenkel (interstitial) defect accumulation following sufficient energy loss. Ionization effects (such as electron holes and consequent charge imbalances) are likely to anneal or recombine through time, whilst the more enduring structural defects can remain for millions of years (Nasdala et al., 2001(Nasdala et al., , 2006. Petrographically termed 'radiohalos', silicate minerals commonly exhibit marked discolouration across an irradiated area as a result of the accumulation of these point defects and ionization effects. Prior studies into the extent of structural damage in radiohalos primarily include sheet silicates such as biotite mica and chlorite; these phases often exhibit the most marked discolouration. In phyllosilicates, where the radioactive inclusions are often only a few microns across, the radiohalos are identified optically FIG. 1. (a) Plane-polarized optical microscope image of radiohalos in grunerite (Amp) surrounding monazite inclusions (Mon) within an almandine-grunerite-biotite (Bi) meta-ironstone from Tunaberg, Sweden. Note the dark brown, smeared discolouration surrounding the inclusion resulting from prolongued radiation damage (∼1.8 Ga). (b) Example of a radiohalo in grunerite surrounding an inclusion of monazite, displaying an inner, darker halo surrounded by a more diffuse outer ring. (c) Cross-polarized optical microscope image of inclusions of monazite (also zircon) in a garnet (Gnt)-cordierite-biotite (Bi) gneiss taken from the Huntly complex, NE Scotland. Note the prominent radiohalos (∼470 Ga) surrounding the monazite inclusions within the neighbouring biotite crystal, which are not visible in the garnet. (d ) Plane-polarized optical microscope image of a pale yellow radiohalo in cordierite (Cord) surrounding an inclusion of zircon (Zir) in granulite-facies pelitic gneiss from Madagascar. Across all images, representative areas for the presented data have been marked with dashed boxes. as a series of concentric darkened spheres, the diameter of each sphere (seen as circular ring in thin section) corresponding to the energy of the α-particle along the uranium/thorium decay chain (∼4-8 MeV) (Demayo et al., 1981;Nasdala et al., 2001Nasdala et al., , 2006Pal, 2004). If the Th/U-bearing phases are larger, the damage halo mimics the grain shape and rings become poorly defined. Radiohalos in biotite mica ( Fig. 1) have been demonstrated to contain extensive loss of short-range order by α-particle bombardment  (Nicollet, 1990;Markl, et al., 2000).
Anhedral zircons have produced radiohaloes with a morphology matching the dimensions of the emitter. The discolouration, whilst not displaying any concentric patterns, is a far lighter, pale yellow colour than the dark brown present in the amphibole. A band of cordierite directly adjacent to the zircon is paler and appears not to have suffered the same degree of discolouration, see Fig. 1.

Zircon (anhedral)
∼300 μm × 100 μm and the electron liberating radiolysis of structural OH − groups has been shown to reduce Fe 3+ within the metal-rich mineral layers (Pattrick et al., 2013;Bower et al., 2015). Whilst not all silicate minerals display intense discolouration with damage, subtle optical effects of radiation damage have been identified here in iron-rich garnet, cordierite and amphibole adjacent to α-emitting inclusions. The following study is a synchrotron Fe K-edge X-ray absorption, near-edge spectroscopy (XANES) and microfocus XRD investigation across silicate phases in the vicinity of α-emitters to examine changes in Fe oxidation state, short-range order and structural defect accumulation, respectively.

Samples
Four radiation damaged silicate samples have been analysed by synchrotron microfocus XRD and Fe K-edge XANES. All samples examined envelop inclusions of zircon or monazite which have emitted high energy (5-8 MeV) α-particles for at least 470 Ma and display various optical effects of the resulting radiation damage, see Fig. 1. Samples with a significant Fe content were deliberately selected for optimal Fe K-edge XAS resolution, which is used here as an indicator of changes in short-range order and redox reactions (Table 1).

Electron probe microanalysis (EPMA)
All major-element analyses of the silicates were undertaken by WDS-spectrometry using the Cameca SX-100 electron-microprobe at the University of Manchester. Standards are as detailed in Pattrick et al. (2013). Reported formulae are an average of high-totalling positions (>96%) over n points (n quoted alongside each formula). The suboptimal cross section size of many of the radioactive inclusions yielded very low totals and consequently monazite and zircon chemistry is semi-quantitative.

Synchrotron methods
Analysis of samples on Beamline I18, Diamond Light Source, UK (Mosselmans et al., 2009) was carried out as detailed in Pattrick et al. (2013) with the initial microfocus study of radiohalos in biotite. Both XRD and XAS data were collected as 'transects' across the irradiated region and towards the α-emitter in 6 μm steps (beam spot = 3 μm).
Samples were prepared as ∼20 µm 'thin' sections; radiohalos were identified optically and the wafers then positioned onto glass slides with a transmission hole below the analysis area.
Microfocus X-ray diffraction X-ray diffraction patterns were acquired at 12 keV using a Photonic Science XDI-VHR 125 CCD and peak positions calibrated with either powdered LaB 6 or Si. Acquisition times ranged from 20 s to 1 min. 1D spectra were subsequently reduced and fitted in Igor Pro 6.34A (2014) using the Nika plug in (Ilavsky, 2012). Where possible, XRD peaks were indexed using the modelling software Jems 6.84 (Stadelmann, 2012) or relevant ICDD index cards. Diffraction peaks across all traverses have been fitted with a combination of Voigt and (in the case of more damaged areas) Gaussian line profiles to attain more precise peak positions.

Microfocus Fe K-edge X-ray Absorption Spectroscopy (XAS)
Fe K-edge XANES data were collected in fluorescence mode. Data were reduced and analysed using the Demeter software package (Ravel and Newville, 2005). All plots displayed are background-subtracted, intensity-normalized spectra; edge positions are reported according to the initial first derivative peaks. High-resolution XANES data were collected up to 50 eV beyond the Fe K-edge. No pre-edge trends have been reported; however this is due to data quality and there is scope here for further study.
Relative changes in first shell EXAFS fitting have been displayed where relevant to aid quantification of Fe oxidation state changes. Optimal fits are limited due to the extent of the useful XAS data. Whilst first shell EXAFS fitting at low k ranges (50 eV past the edge, k = 6) will have inadequacies in fitting, relative changes in fit results will be valid. Fit results have been presented in Supplementary Information. (deposited with the Principal Editor of Mineralogical Magazine and available at www.minersoc.org/pages/ e_journals/dep_mat_mm.html)

Grunerite
The amphibole studied has been assigned as grunerite; EPMA both inside and outside the halo regions yielded an average composition (±0.01) of: The XANES spectra across the grunerite radiohalo show a change in the relative white line heights of the double-peaked edge (see Fig. 2). An increase in the ratio of peak i (7127 eV) to peak ii (7131 eV) represents a higher proportion of Fe with a lower average binding energy neighbouring the monazite and suggests a reduction of a proportion of the structural Fe 3+ with increasing proximity to the αemitter. Work by Dyar et al. (2002) and Monkawa et al. (2006) revealed the relative changes in Fe K-edge XANES peak heights in amphibole represent differing Fe 3+ /ƩFe ratios. First-shell EXAFS fitting was possible beyond the absorption edge. Relative changes in octahedral Fe-O interatomic distances also suggest Fe 3+ reduction in the halo region; an overall increase in average bond length is indicative of this process (Fig. 2b). The average Fe-O distance outside the radiohalo is 2.073 Å, whilst within the discoloured, irradiated region this value increases to a maximum of 2.19 Å, a consequence of a higher average signal from 6-coordinated Fe 2+ . XRD peak fitting across the traverse reveals extensive structural damage across the irradiated region. Figure 3 illustrates three examples of XRD peak changes as a product of accumulated αirradiation. With the exception of reflection (202) which is isolated in 2θ space, all peaks in close proximity merge and broaden with decreasing distance to the monazite, demonstrating increasing radiation damage. In the case of reflections (281) and (312), this merging is accompanied by an overall shift of the peak position of the individual reflections to higher 2θ angles (with the exception of peak (191), which has merged to lower angles or disappeared). This shift is also observed in peak (202), but the FWHM (full width at half maximum) and overall intensity of the peak has decreased. At this resolution, most reflections expand to form overlapping peaks and closely neighbouring lattice reflections are individually lost; this is especially noticeable with reflections (201), (060) and (241).
Reflection (202) represents a series of M4 site cations (Fe, Mg) in-plane oblique to the tetrahedral chains. Similarly, reflection (281) links octahedral cations between chains in a similar vector, both trends indicative of a chain-parallel contraction. Large voids occur periodically along the chains in pristine amphibole, a result of the alternating orientation of the tetrahedra. It is possible that increased point defects from α-particle bombardment may cause minor structural collapse into these regions that will act as sinks for displaced atoms; the amphibole structure may be highly susceptible to loss of short-range order as a result.

Almandine
The two garnets studied yielded an average formula of (±0.01 p.f.u.):  2. (a) Normalized Fe K-edge XANES across a radiohalo in a crystal of grunerite with increasing proximity to the αemitter (monazite). The white line is split into two peaks (∼4 eV apart), interpreted as a signal from both Fe 2+ (∼7127 eV) and Fe 3+ (∼7131 eV). Within the halo, the relative intensity of peak i increases over peak ii. Spectra have been spaced for clarity; each spectrum represents a 6 µm shift along the line of traverse towards the monazite/grunerite mix (shown by a dashed line). (b) Absolute changes in intensities for normalized Fe K-edge peaks i and ii across a radiohalo transect. The increasing ratio of peak i (Fe 2+ ) to ii (Fe 3+ ) suggests a reduction of Fe 3+ within the halo area as a product of α-particle damage. (c) First shell distance (Fe-O) EXAFS fit results across the XAS transect, showing an overall increase in bond lengths in the irradiated halo region. More detailed fit data are included in Supplementary Information (deposited at www.minersoc.org/pages/e_journals/dep_mat_mm.html).
FIG. 3. Representative, fitted XRD peaks displayed at intervals along the grunerite → monazite transect. Legend denotes distances from the α-emitter. At ∼35 μm, the edge of the radiohalo is approached, thus reflections at 48 μm are assumed to represent unirradiated crystal. Indexed reflections are shown above undamaged peaks where identification was possible. Note the different 2θ scales.
Fe-rich member of the pyrope(Mg)-almandine(Fe)spesssartine(Mn) series; its structure consists of an orthosilicate framework of tetrahedrally and octahedrally coordinated cations, with the addition of central domains containing dominantly Fe cations surrounded by eight oxygens (Deer et al., 1992). The monazite in almandine B contains ∼3.1 at.% Th, slightly higher than the amphibole sample from Tunaberg. The actinide content of the zircon in almandine A is unknown; however, it is likely to contain less uranium (the primary α-emitter) than the monazites; zircons typically contain ∼0.1 at.% uranium (Cuttitta and Daniels, 1959;Kusiak et al., 2009;Sano et al., 2000). It is important to note the differences in size between the two α-emitters, which may account for radiation damage trends. Radiation damage induced changes are apparent in the XANES spectra of almandine A surrounding the zircon. Two spectra are shown in Fig. 4a, one ∼6 µm from the zircon edge and a second ∼48 µm from the emitter in the unirradiated garnet. Whilst there is no optical halo to act as a reference, Monte Carlo based simulation software, the Stopping and Range of Ions in Matter (SRIM) (Ziegler, 2013) predicts that an 8 MeV α-particle travelling through garnet with a density of 4.19 g/cm 3 and composition shown, has a maximum range of 31 µm. This excludes energy lost upon exiting the zircon. It is assumed therefore that the point at 48 µm from the emitter is unirradiated.
An overall broadening of the modulations beyond the edge is apparent in the irradiated region in almandine A. A minor shift of the white line maximum to higher energy is apparent; however the top of the edge is relatively 'flat' and the edge itself broadens with proximity to the monazite. Such a small shift is not definitive of an Fe valence change and could represent a local, natural variation or damage induced structural disorder.
It is likely that an increase in point defect density accounts for the changes in the spectra above the edge. Farges et al. (1997) demonstrate a broadening of the Ti K-edge with decreased periodicity in Ti oxide samples, attributable to a variation in edge position on an atomic scale as a function of nonuniform absorption sites. As not all atomic path lengths will be similar across irradiated samples, broader XANES (and extended) features will result. X-ray diffraction peak fitting across almandine A shows radiation damage induced trends similar to those observed in the grunerite sample. Almandine A adjacent to zircon displays peak broadening and overall reflection intensity reduction into the 'halo' zone (Fig. 4b). Reflection (233) broadens and shifts to higher 2θ angles within 6 μm of the margin of the monazite, indicative of a contraction of the cyrstal in this plane. Peak (026) splits into two broad, lowintensity peaks located either side of the original reflection position, indicative of both lattice expansion and contraction in this plane (Li et al., 2004). An overall increase in FWHM across most peaks within 24 µm of the zircon is again indicative of minor structural aberrations, whilst reflections (246) and (237) are significantly diminished in intensity (Fig. 4b).
In contrast, almandine B presents far fewer effects of radiation damage in the XRD data. Reflection (046) displays a shift to lower reflection angles within 36 µm of the margin of the monazite (Fig. 5b), indicative of a very slight expansionary strain of the crystal in this plane. No peak broadening is observed. All other reflections in the traverse across almandine B displayed no changes with respect to the alpha emitter, suggesting relatively low structural change as a consequence of irradiation. A similar damping effect upon the XANES signal was observed into the halo region, although not as pronounced as the trend in almandine A. First shell EXAFS fitting was possible across the transect in almandine B, Fig. 5a shows a slight decrease in Fe-O distance into the 'halo' region, in contrast to the trend observed within the grunerite sample.
Whilst α-particle bombardment appears to have a similar effect regardless of emitter (U vs. Th), almandine A (surrounding zircon) has clearly accumulated a higher degree of radiation damage, despite a lower actinide content of the α-source. This trend is difficult to fully explain. However two factors may contribute; the inclusion size and its consequent stopping power, as well as the level of damage sustained by the host lattice; in each case the zircon may allow more α-particles to escape into the garnet.

Cordierite
The cordierite studied also showed similar chemistry across the sample (±0.01): Cordierite is an orthorhombic ( pseudohexagonal) cyclosilicate comprising a framework of sixmembered tetrahedral silica/alumina rings joined laterally by further tetrahedral Al or Si as well as periodic octahedral Mg or Fe. Water molecules can be present within the rings (Rigby and Droop, 2008;Deer et al., 1992). Rigby and Droop (2008) measured the volatile content of this cordierite, presenting mean values of 0.31 moles H 2 O and 0.012 moles CO 2 per formula unit.
No changes were observed in the XANES transect across the cordierite radiohalo. Whilst it is likely that the cordierite has received a lower αflux than the other samples analysed (assumed by the 'pale' discolouration intensity); the Fe content of the sample is far lower in comparison to those previously analysed; therefore redox changes may be less easily observed. The presence of a small amount of water within the structure may also be contributing to the slight colour change, not observed in anhydrous phases.
Data from points both in the pale yellow halo region and in the relatively clear band adjacent to the zircon were collected; however X-ray diffraction results did not show a trend in radiation damage coincident with the pattern of discolouration. Indeed, the highest density of point defects are neighbouring the zircon edge, demonstrated by line broadening (Fig. 6).

Discussion
This investigation has revealed the extent of radiation damage manifested in radiohalos within a range of silicate minerals. The discolouration seen around α-emitters reflects the earliest stages of metamictization, similar to that recorded in phyllosilicates (Pal, 2004;Nasdala et al., 2001Nasdala et al., , 2006Pattrick et al., 2013). Across all phases presented here, the intrinsic mineral structure is preserved within the irradiated regions (as evidenced by retention of the main XRD peak intensities and FIG. 6. XRD traverse across a radiohalo in cordierite surrounding an inclusion of zircon. Note the overall broadening of the fitted peaks within ∼18 µm of the α-source; no relationship between the more intensely discoloured band and the paler region close to the zircon is observed. Note the different 2θ scales. positions); however radiation damage is clearly present, to differing degrees. Varying contributions from point (Frenkel) defect accumulation are pervasive across all radiohalos, regardless of mineral phase. Diffuse scattering (broadening of XRD peaks) is a common characteristic of low-level point defect accumulation and represents imperfections in lattice periodicity (Chailley et al., 1994), with individual domains of variable defect densities existing over sub-micron scale volumes. In contrast to perfectcrystal order with sharp-peak Bragg scattering, diffuse scattering denotes non-uniform distortion along a crystal plane, thereby increasing the range of permitted diffraction angles and broadening XRD peaks. Where peaks are close in 2θ space, individual reflections merge and are lost as a product of broadening. Whilst Bragg diffraction peaks generally lose intensity with radiation damage, merging has led to some diffuse peaks apparently gaining intensity across the traverses shown here. In no instance were the samples entirely amorphous or 'metamict' (Ewing et al., 1988;Tomavsićet al., 2008) within the halo regions; indeed it is likely that amorphization (if any) occurs over small, discrete areas, resulting from sporadic, high concentrations of point defects. Over geological timescales, minerals may form stable clusters of defects within a broadly crystalline lattice, allowing for long-term retention of radiation damage without detriment to the overall structure (Ewing et al., 2000).
Minor shifts in peak position with radiation damage are indicative of volume changes in the lattice. In the XRD patterns shown here, peaks display small movements in 2θ space, indicative of damage-induced lattice expansion and contraction on the order of +/-0.002 Å. In the instance of Frenkel defect formation, a trade-off exists between a minor volume increase (interstitial relocation) and decrease (vacancy formation and relaxation) (Grigull et al., 2001). Both mechanisms appear to be present across the damaged regions identified here, with no apparent patterns across areas with higher defect densities.
Microfocus analysis suggests that the highest level of structural damage occurs directly adjacent to the α-emitter. This is in contrast with conventional models of α-particle energy deposition, whereby an α-particle creates a narrow domain of structural defects only over the final ∼6 µm of its range (as predicted by SRIM) (Ziegler, 2013). This trend may be partially a factor of alpha particle energy loss before leaving the source lattice.
Another possibility is the potential for a 'wandering' α-recoil effect beyond the extent of the host lattice. Upon ejecting an α-particle, the subsequent daughter nucleus will recoil (E = ∼0.1 MeV) in a random direction. Multiple recoil effects along a similar vector have the potential to create a highdefect region as long as 100 nm (Seydoux-Guillaume et al., 2009); this has the potential to increase over geological timescales. Nasdala et al. (2006) suggest the presence of a far smaller 'α-recoil halo' directly adjacent to a decaying emitter may contain extensive radiation defects.
The exact mechanism of discolouration (ionization vs. structural defects) across radiohalos is still a subject of debate (Nasdala et al., 2006). This study finds that discolouration and structural damage do not always directly coincide; it is possible that colour-inducing radiation 'damage' may be more a product of ionization mechanisms, whilst point defects alone do not directly produce a pattern of discolouration. It should be noted that 'darkening' over irradiated areas is only present in hydrated mineral phases (biotite, chlorite, cordierite, amphibole) (Nasdala et al., 2001(Nasdala et al., , 2006Pal, 2004).
Iron redox chemistry has been shown to be influenced by prolongued α-irradiation. A Fe 3+ to Fe 2+ reduction mechanism via the radiolysis of structural OHwithin α-irradiated biotite has been suggested by Pattrick et al. (2013); a similar trend has also been demonstrated for grunerite in this study (also containing structural hydroxyl groups within the lattice). In direct comparison, anhydrous structures such as garnet show no such trends. The presence of water within the structure appears to be a major driver of Fe redox reactions, further experimental work into which is necessary, as OH-bearing phyllosilicates are important in the near-and far-field of a GDF and the consequences of radiolytic degradation are a necessary consideration for hydraulic barriers such as bentonite.

Conclusion
Silicate minerals hosting radioactive inclusions provide useful analogues for the structural and chemical response of such phases to long-term alpha particle bombardment. Whilst radiation damage at these (unknown) doses is broadly accommodated by the structure, absorbing phases within the host rock that isolate and even incorporate actinides will suffer a degree of αparticle damage. There is potential for short-range, highly amorphizing α-recoil effects (Chakoumakos et al., 1987;Meldrum et al., 1998;Weber et al., 1994). Increased defect densities and consequent structural accommodation/phase changes may affect mineral properties and behaviour, which may prove to be highly variable over micrometre (even nanometre) scale regions. Microfocus study of radiohalos is a useful tool to constrain the spatial extent of damage and its relationship to a radiation source.
Some drawbacks arise from the study of radiohalos as radiation damage proxies. The stochastic nature of α-particle damage accumulation from radioactive inclusions of differing morphologies makes direct comparison between (and within) halos difficult. More generally, all studies of naturally accumulated radiation effects must take into account the thermal history of the sample; however this is not a critical issue in the samples analysed here. Analysis of radiohalos yields overall, time-averaged trends resulting from an exceptionally long timescale of defect accumulation and natural crystal relaxation/defect stabilization must be considered.
Despite the advantages of studying long-term radiation damage accumulation in radiohalos, a critical concern for durability and performance studies is the need to accurately constrain doses and dose rates across irradiated materials. The αdose rate across radiohalos is likely to be extremely low, such that recombination of defects or lattice stabilization is accommodated despite prolonged irradiation; however damage accumulation adjacent to an emitter appears far more lasting. There is doubtless a structure effect on lattice susceptibility to radiation damage; whilst direct radiation dose comparisons cannot be made, the amphibole studied here displays the highest degree of structural discontinuity in comparison to the likely more robust, isotropic silicate frameworks of the garnet. Significant further work with controlled α-irradiations is required to determine the effect of lattice structure upon damage manifestation.