Mechanism of powellite crystallite expansion within nano-phase separated amorphous matrices under Au-irradiation

A fundamental approach was taken to understand the implications of increased nuclear waste loading in the search for new materials for long-term radioisotope encapsulation. This study focused on the formation and radiation tolerance of glass ceramics with selectively induced CaMoO4 as a form to trap the problematic fission product molybdenum. Several samples were synthesised with up to 10 mol% MoO3 within a soda lime borosilicate matrix, exhibiting phase separation on the nano scale according to thermal analysis, which detected two glass transition temperatures. It is predicted that these two phases are a result of spinodal decomposition with Si–O–Ca–O–Si and Si–O–Ca–O–B units, with the latter phase acting as a carrier for MoO3. The solubility limit of molybdenum within this matrix was 1 mol%, after which crystallisation of CaMoO4 occurred, with crystallite size (CS) increasing and cell parameters decreasing as a function of [MoO3]. These materials were then subjected to irradiation with 7 MeV Au 3+


Introduction
As new developments are made to nuclear reactor design and performance, greater stress will be placed on materials for waste storage, as these new developments will create different waste streams. [1][2][3] This shift is concurrent with a general desire to decrease the final volume of waste for storage, as more countries turn to nuclear power as a method of decarbonisation. 4 This has led to a renewed interest in the development of alternative glass or glass ceramic compositions for the encapsulation of waste streams that contain a higher concentration of compounds with a limited solubility in borosilicates, such as molybdenum, platinoids, and rare earths (REs).
These compounds have been problematic in the traditional process of vitrification [5][6][7][8][9] as they lead to uncontrolled precipitation of crystalline phases within the glassy matrix that can act as carriers for radioactive caesium, strontium, or minor actinides and lanthanides (Ce 3+ , La 3+ , Sm 3+ , Nd 3+ , Gd 3+ , Y 3+ ), 7,10 that will subsequently alter the psychochemical properties of the wasteform. 11 This type of precipitation can also increase the fracture potential at the interface between these phases, which would similarly affect material properties. [12][13][14] While glasses prove to be useful containment matrices as they can be easily synthesised on a large scale, and show good thermal and radia-weapons decommissioning, post operation clean out, or Generation IV reactors is considered. 10,17,18 In response to this problem single or poly-phase ceramics based on natural analogues, such as apatites, zirconolites or synroc were developed for the encapsulation of specific radioisotopes with desired material properties. 10,[19][20][21][22] Unfortunately, they required high temperatures and pressures during synthesis, making them both time-consuming and costly to manufacture on an industrial scale.
As a result, a new ideology has emerged of inducing controlled crystallisation of predesigned phases within a glass matrix using the existing vitrification infrastructure. These glass ceramics (GCs) or glass composite materials are able to incorporate poorly soluble waste components such as sulfates, chlorides, molybdates and actinides in a durable crystalline phase, while the majority of other shorter-lived radioisotopes can be trapped in a flexible amorphous network. 10,17,23,24 In the context of molybdenum incorporation, selective formation of CaMoO 4 , which is 13 500Â less soluble than alkali molybdates 25 within a borosilicate framework is of interest.
In reprocessed waste streams, the fission product molybdenum is usually found as Mo 6+ in oxidizing or neutral conditions. [26][27][28] It forms the oxyanion (MoO 4 ) 2À according to Mo EXAFS, which are found to be unconnected to each other and to the glassy framework when dissolved into an amorphous structure. 27,29,30 According to Greaves' structural model, this is because they are located in non-bridging oxygen (NBO) channels. 31 Mobile alkali or alkaline earth cations are also found to cluster is these channels, and can coordinate octahedrally to (MoO 4 ) 2À entities by weak long-range ionic forces. 32 A similar configuration is also found for crystalline complexes, 8,30,32 where the process of phase separation stems from the accumulation of (MoO 4 ) 2À oxyanions in close proximity to charge-balancing cations within these depolymerized regions. [33][34][35][36] The preferred formation of crystalline CaMoO 4 over alkali molybdates can be driven by composition, 30,32,34 external heat treatments, 8,34,37 increasing the quench rate during synthesis, 38 or through redox chemistry. 35,39 Compositionally, molybdate speciation is influenced by the preferential charge balancing of BO 4 À and (MoO 4 ) 2À anionic entities by Na + and Ca 2+ ions. 40 Owing to ion charge, mobility, size and sterics both anions prefer Na + as a charge balancer, but BO 4 À will have a stronger affinity to Na + , thus enabling the incorporation of Ca 2+ into CaMoO 4 . 29,41,42 This trend has been observed to increase for increasing amounts of B 2 O 3 or CaO, as it affects the population of BO 4 À species. 32,34,40 Composition is also a significant factor in controlling glassin-glass phase separation within the residual matrix, which will have subsequent effects on molybdate precipitation. In past phase separation studies focusing on molybdate formation, the concentration of CaO in soda lime borosilicate glasses has been limited to B11 mol%, as it is known to cause glass-in-glass phase separation. 34 This limitation has resulted in Na 2 MoO 4 production in most studies, despite the manipulation of other factors affecting speciation. In the calcium borosilicate system, low concentrations of CaO will generally lead to phase separation of calcium borate and calcium silicate-based phases. 43,44 This is also true for alkali borosilicates. 45 For either alkali or alkaline earth borosilicates, the regions of immiscibility decrease with the atomic size of the cation, and increased incorporation can be achieved by rapid quenching, according to temperature profiles.
Base glasses formed solely of network formers can also show phase separation with a dependence on SiO 2 /B 2 O 3 , 46,47 hence the need for cations to induce homogeneity. However, at low alkali or alkaline earth concentrations, regions of immiscibility within borosilicates are evident with temperature-dependent transitions. [43][44][45] Approaching these regions of immiscibility through composition or heat treatments can cause a multicomponent system to separate into distinct phases with different chemical compositions, and therefore different physical properties. 48 It can proceed through spinodal decomposition, or via a nucleation and growth process, as the system moves from metastable to unstable within the immiscibility dome. The first process is a result of small stochastic fluctuations in composition with temperature that occurs without a thermodynamic barrier. 49 It arises within a chemical spinodal area that may be present within a bimodal immiscibility region. Within the spinodal dome, homogeneous phase separation will occur. In contrast, nucleation and growth is a two-step process in which larger fluctuations cause the discrete nucleation of separated regions, which grow as a result of diffusion. This process can occur anywhere in the immiscibility region as the temperature of the system decreases. The distribution and size of secondary phases in this mechanism are governed by the kinetics of these two steps, which are assumed to follow Gaussian profiles with regards to temperature 49,50 and will determine many structural properties of the glass. 51 Phases can be identified through a unique enthalpy signature associated with the glass transition temperature (T g ), which is the process at which a liquid ''freezes'' into a metastable solid. 52 While several factors can impact glass-in-glass phase separation and the speciation of molydates during synthesis, their durability against internal radiation is of considerable importance in determining its viability as a GC wasteform. Materials containing radioisotopes will continuously undergo a-decay of minor actinides and Pu, b-decay of fission products, and transitional g-decay, which is a secondary process following a and b-decay. These collective decay processes can cause significant changes to long-range ordering through atomic displacements, ionisation, and electronic excitation.
Macroscopically, these structural modifications can cause changes to volume, where both swelling and densification have been observed, 6,53,54 and to mechanical properties, such as hardness, fracture toughness and elasticity. [55][56][57] They can also induce phase transformations, such as devitrification of the amorphous network, precipitation, amorphisation of crystalline phases, redistribution of radionuclides to existing molybdenum assemblages, glass-in-glass phase separation, or the clustering of cations. 5,6,13,[58][59][60] The range of effects is of course dependent on composition, and can sometimes result in favourable properties like an increase in fracture toughness, or re-vitrification of unwanted crystalline phases. 5,6 The a-decay process, and specifically the heavy recoil nuclei, are responsible for the for the bulk of these changes. 5,6,61 During a-decay, a fast moving a (or He 2+ ) particle with kinetic energy between 4.5 to 5.5 MeV, and a heavy daughter isotope with kinetic energy between 70-100 keV referred to as the a-recoil are created from a stationary parent isotope. The small a-particle will interact with electrons in a substrate and typically travel 10 to 20 mm, while the a-recoil will interact primarily through nuclear collisions and stop after B30 nm. 5 These nuclear collisions will have a billiard ball-type effect that will initiate a chain reaction, referred to as a displacement cascade. While this process will generate substantial structural reorganization, it is partly mitigated by the a-particle that will initiate some recovery processes through the creation of latent ion tracks that can be considered to mimic high temperature effects. 6,62,63 These competing processes of damage creation and defect annealing have been observed to result in a saturation of changes to mechanical properties, internal energy, and density for a cumulative dose of 4 Â 10 18 a g À1 . 6,64,65 Given current waste loading standards and waste streams, this is expected to occur following B1000 years of storage, 2,6 which corresponds to approximately 1 displacement per atom (dpa). 64,66 By replicating the damage in this timeframe, long-term structural projections can be produced to test the durability of GCs against internal radiation and to understand how radiation damage can be used as a tool to design specific structural components adaptable for transmuted waste created through decay processes, or to increase the solubility of non-radioactive fission products. This paper specifically aimed to answer the question of if irradiation would: (1) lead to powellite (CaMoO 4 ) amorphisation; (2) modify the crystalline structure or induce cationic substitution; or (3) result in modifications to the borosilicate network that would enable precipitation. It further sought to identify the mechanism in which any of the above would occur as a function of composition. It therefore attempts to validate a CaMoO 4 bearing GC as a candidate wasteform with a higher waste loading efficiency on a fundamental level.

Composition and synthesis technique
The GCs in this study were created to test the formation of molybdates when MoO 3 was added in a 1 : 1 ratio to CaO in a borosilicate glass normalised to French nuclear waste glass SON68 (non-active form of R7T7) with respect to SiO 2 , B 2 O 3 and Na 2 O. The excess Ca 2+ ions were required to promote powellite production with the aim of preventing any water soluble Na 2 MoO 4 from forming. See Table 1 for the normalised compositions.
Several of the samples also contained a fixed amount of gadolinium. RE dopants can be considered as an actinide surrogate, and can therefore act as an indicator for potential incorporation sites of radioactive elements. They can further be used to investigate how incorporation would affect powellite crystallinity.
Compositions were first melted for 30 min. They were then cooled, crushed and re-melted for 20 min. A double melt with short melt times was used to try and ensure homogeneity, while minimizing volatilization. Specimens were annealed for 24 h at 520 1C to reduce internal stresses, after which they were cut to B3/4 mm Â 3/4 mm in order to fit the irradiation sample holder (see Fig. 1). These pieces were all hand polished successively using P600, 800, 1200, 2400 and 4000 SiC grit paper, followed by 3 mm and 1 mm diamond polishing to achieve a uniform thickness of approximately 500 mm, and a surface appropriate for analysis.

Irradiation experiment
Irradiation with medium energy ions can be used to simulate some of the damage created by a-decay observed on long timescales. While this method does not simulate the a-particle, which can induce healing, 11,62 the nuclear stopping power of ions such as Pt, Au, and Pb is similar to that of the a-recoil (5 keV per nm), 67 which as previously mentioned is responsible for creating displacement cascades that result in the greatest degree of structural changes.
The radiation experiment conducted in this study was performed with 7 MeV Au 3+ ions at a flux of 2 Â 10 10 ions per cm 2 s on the ARAMIS beamline at CSNSM in France. A total dose of 3 Â 10 14 ions per cm 2 was achieved, which corresponds to approximately 1 dpa with a ion penetration depth of B1.5 mm, as estimated from TRIM simulations 66 (see ESI † for results of simulation). This level of damage should test the stability of any crystalline phases, as well as the flexibility of the residual amorphous phase. Fig. 1 shows the sample prior to and following irradiation.

Characterisation techniques
The primary points of interest in this investigation were radiation affects on crystallinity and the Mo environment. This required use of several analytical techniques as GCs contain multiple phases, which can create complexity when studying modifications following irradiation. Crystal phase and crystallite size identification were performed using X-ray diffraction (XRD) with CuK a1 (l = 0.15406 nm) and CuK a2 (l = 0.15444 nm) wavelengths on a Bruker D8 ADVANCE. The diffractometer was equipped with Göbel mirrors for a parallel primary beam, and a Vantec position sensitive detector. For each sample, data was acquired for a 2y = 10-901 spectral range with a 0.021 step size, and a 10 s per step dwell time. Samples were analysed as monoliths instead of powders to isolate irradiation effects at the surface, and rotated to find the maximum diffracting orientation before final acquisition in order to compare samples containing randomly orientated crystals with some accuracy. Scherrer crystallite size (CS) valuations were determined from structural analysis based on whole pattern Rietveld refinements with the software Topas v4.1. 68 A single parameter approach was utilized based on the quality of data available and the large amorphous content in most samples. 69,70 In this fitting approach, CS is presumed to incorporate contributions from both size and strain, as correlation issues prevented the independent deconvolution and quantification of these two physical properties in GCs.
In order to characterise and assess structural changes to both the crystalline and non-diffracting amorphous phases following irradiation, Raman spectroscopy was utilized. Spectra were collected with a 300 mm confocal Horiba Jobin Yvon LabRam300 spectrometer coupled to a Peltier cooled front illuminated CCD detector over the 150-1600 cm À1 spectral range with a 2 mm spot size and a holographic grating of 1800 grooves mm À1 . The 532 nm excitation line used in this experiment was produced using a diode-pumped solid-state laser (Laser Quantum) with an incident power of 100 mW focused with an Olympus 50Â objective. This set-up enabled a spectral resolution of B1.4 cm À1 per pixel (1024 Â 256 pixels in size), and an estimated penetration depth of B22 mm based on depth profile analysis. This means results also showed a contribution from the pristine layer under the irradiation zone. Despite this fact, these parameters were used based on a number of factors including acquisition time, spectral resolution, minimization of damage, and a desire to describe the amorphous phase. As a result, the recorded trends are indicative of those induced by irradiation, but the magnitude of these trends will be more pronounced if measurements were solely collected from the irradiation zone. Raman spectra of GCs were analysed with PeakFit software using pseudo-Voigt profiles to characterise bands, while spectra of amorphous phases were analysed qualitatively due to uncertainties with comparing overlapping broad bands.
Multi-phase GCs present varying morphologies and phases for quantification. This was analysed using SEM backscattered (BSE) imaging and energy dispersive X-ray spectroscopy (EDS) on a Quanta-650F held at low vacuum (0.06-0.08 mbar) and performed with a 5-7.5 keV beam. This configuration resulted in a maximum penetration depth of B0.05 mm. For EDS analysis, measurements were collected with a 8 mm cone in order to reduce skirting effects when analysing small particles. Images were collected using FEI Maps software, while acquisition and analysis for EDS was performed using Bruker ESPRIT software from mapped areas or 10-15 points. Quantification of particle size and density at the surface was determined by image analysis using ImageJ. 71 While these analytical techniques were useful for comparative analysis of glasses and GCs prior to and following irradiation, addition characterisation of synthesised materials prior to irradiation was required to help explain subsequent modifications. The two bulk characterisation techniques employed for this purpose were 11 B MAS NMR and simultaneous differential scanning calorimetry -thermogravimetric analysis (SDT). In both cases, analysis following irradiation was limited owing to the small irradiation volume and the requirement to powder monoliths, thus creating additional defects and limiting further investigation. 11 B MAS NMR (m z = 2.6887; g = 8.5847 Â 10 7 rad T À1 s À1 ) experiments were conducted on a Varian Infinity Plus spectrometer with a 11.74 T superconducting magnet, tuned to the Larmor frequency of B160.37 MHz using 2.5 mm probes. For each measurement, approximately 9-13 mg of powdered solids were packed in zirconia rotors, and spun at 20 kHz. 901 pulses with a 0.7 s pulse delay, and 5 ms receiver and acquisition delays were used. Chemical shifts were then measured relative to the primary liquid standard BF 3 Et 2 O (0 ppm), and the secondary solid standard of NaBH 4 (À42 ppm). Thermal SDT analysis was performed on a TA Instruments Q600 with Pt/Rh thermocouples arranged for a dual beam balance under nitrogen gas to ensure no weight changes resulted from oxidation at high temperature. The equipment configuration resulted in a standard balance sensitivity of AE0.1 mg, and a AE2% calorimetric accuracy. Two measurements were taken per sample using 10-18 mg of powdered material in alumina pans. A 10 1C min À1 heating rate to a maximum temperature of 1500 1C was applied to each experiment, after which the system was air-cooled. All observable thermal transitions were analysed using TA Instruments Universal Analysis 2000 software.

Bulk material characteristics
The addition of MoO 3 can lead not only to crystallisation of alkali or alkaline earth molybdates, but it can also cause changes to the residual glass matrix. These changes will have subsequent effects on thermal properties, which can be used as metrics for phase separation and crystallisation. The events of interest measured by SDT are T g and the melting temperature T M of crystallites, both of which are summarised in Table 2 (spectra provided in ESI †).
Interestingly two T g s were detected for glasses and GCs. The first transition, T g1 around 600 1C is closer to the value of T g found for similar soda lime borosilicate systems with [SiO 2 ]/ [B 2 O 3 ] 4 3. 72,73 This transition also exhibits the previously observed inverse relationship to increasing [MoO 3 ]. 35 The outlier to this rule was CN10, which most closely emulated the temperature and enthalpy of transition seen in CNO suggesting that with excess MoO 3 the conversion of phase separated Mo-O-Ca-O-B/Si to CaMoO 4 increases, thereby reducing the amount of molybdenum in the amorphous phase. It may also be related to the gadolinium dopant present in the other compositions, though previous studies have indicated that this dopant would reduce primary phase separation instead of increasing it. 8  While the composition dependent shifts of these transitions imply that MoO 3 altered the glass structure, the presence of these two glass transitions implies a degree of heterogeneity. Furthermore, the relative enthalpies of transition suggest that MoO 3 can increase the domain size of separated phases. It is speculated that the relationship observed between T g2 and [MoO 3 ] is correlated to molybdenum phase separation and eventual crystallisation primarily within this higher refractory phase. As more [MoO 3 ] is added to the system, (MoO 4 ) 2À oxyanions accumulate leading to additional phase separation and the tendency for CaMoO 4 crystallisation subsequently increases, which would increase T g of the encapsulating phase. Correspondingly, the number of Ca 2+ ions in the residual glassy matrix of the phase associated with T g1 decreases as they are pulled towards phase 2, thereby allowing T g1 of phase 1 to increase.
Multiple phases cannot be clearly resolved in the micrographs of CNG1 in Fig. 2. This suggests that phase separation proceeded through spinodal decomposition, which would result in two phases that cannot clearly diffuse light due to the nanodomain size, as has been previously found experimentally. 51,76 Further investigations using TEM and utilization of heat treatments above T g1 could be used to confirm this theory.
While a complex relationship was observed for glass transitions with composition, a more straightforward observation was made with respect to the liquidus or melting temperature (T M ) of CaMoO 4 or a CaMo-rich precursor environment (see Fig. 3). Given that CNG1, which appears to be a glass also produced a T M , it is assumed that there is an amorphous precursor environment to crystallisation in which the structure of clustered (MoO 4 ) 2À oxyanions in relation to cations is very similar to that of crystalline powellite, as has been previously observed. 32 It can therefore also produces a similar ''melting'' transition. The presence of this transition could also indicate that there are very small crystallites (o50 nm) in CNG1 that are beyond XRD and SEM detection (o1 vol%).
Generally, T M and the enthalpy of this thermal event (H Tm ) increase with the quantity of crystal matter, which increases with [MoO 3 ] as Fig. 3 indicates. Prior to crystallisation on a detectable level, T M is found at a lower temperature (CNG1). As [MoO 3 ] increases forming GCs, a corresponding increase in T M and H Tm can be correlated to a predicted increase in the crystal content. This reflects the fact that motion in an amorphous precursor environment is easier than in an analogous phase with denser atomic packing and higher ordering resulting in increased CaMoO 4 formation.
The results in Table 2 suggest a level of phase separation within the bulk residual glass matrix, which can be correlated to the extent of Si-O-B mixing within the short-range order of the glass. Initial phase separation is independent of MoO 3 , but inclusion of MoO 3 causes additional phase separation of within the borosilicate phase in addition to the formation of a Ca-Mo-rich environment associated with T M . 11 B MAS NMR has long since been considered one of the primary tools to elucidate this type of mixing and phase separation, and thus Table 2 Glass transition temperature(s) and melting temperature determined by SDT analysis   was employed in this study, the results of which can be seen in Fig. 4. There are two main contributions detected. The first is from (BO 4 ) À groups centred around À1 ppm, with downfield shifts expected for an increasing number of boron nearest neighbours as opposed to silicon, and an upfield shift expected for more silicon nearest neighbours. 33,34,77 The second detected group of overlapping peaks is from BO 3 units centred around 10 ppm. The quadrupolar BO 3 peak is composed of two contributions from ring and non-ring configurations, 34,78 as the labels on Fig. 4 indicate. While there are no drastic changes to the mass normalised 11 B MAS NMR spectra, some minor modifications occurred.
Firstly, the spectra for CN10 and CNO are similar, indicating that MoO 3 has minimal effects on the coordination of boron when molybdenum is primarily found in a crystalline phase. The main modifications observed were a minor increase in non-ring over ring BO 3 (3Si,1B), but none of these changes were occurring on a significant scale. Secondly, a visible spectral broadening for samples CNG1, CNG1.75, CNG2.5, and CNG7 is evident. This is caused by the gadolinium, which with 7 unpaired electrons (spin 7/2) can create a local magnetic field that will induce spectral broadening. The slow electron flipping of gadolinium ions can also cause fluctuations in the relaxation of boron with spin 3/2, which would similarly cause spectral broadening.
The inclusion of a dopant also appeared to increase the relative [BO 3 ]/[BO 4 À ] fraction in all of these samples, with a preference for non-ring BO 3 units as [ MoO 3 ] increases. This could perhaps be tied to structural units in the nano-phase associated with T g1 , as the relative enthalpies of transition follow a similar trend.

Crystallisation
In soda lime borosilicate glasses normalised to SON68, the selective formation of water-durable CaMoO 4 was achieved by introducing up to 10 mol% MoO 3 in a 1 : 1 ratio to CaO for a fixed melting process according to XRD (see ESI † for raw XRD patterns). Spherical particles containing CaMoO 4 crystallites were found to be homogeneously distributed within the amorphous network indicative of a nucleation and growth process (see Fig. 5). The particle sizes (PS) observed were proportional to MoO 3 , with two groups identifiable for low and high [MoO 3 ], as Table 3 indicates. This was also true of the Scherrer CS determined from Rietveld refinements of XRD patterns (see Table 4). Increasing [MoO 3 ] was also observed to be inversely related to the a and c cell parameters of powellite.
An exception to the proportional dependence of CS with [MoO 3 ] occurred in CN10, which mirrors deviations observed in bulk characterisation. This may result from the absence of a Gd 2 O 3 dopant in this composition, which is predicted to improve crystallisation kinetics in the other GCs, or could be a result of sampling variations. For compositions with [MoO 3 ] o 1.75 mol%, samples did not show the formation of distinctive particles according to BSE imagining and showed no diffraction peaks (see Fig. 2 and ESI †), indicating that they were completely amorphous. This retention limit of molybdenum is similar to that observed in previous compositional studies of SON68-type borosilicates. 8,29,30 Following Au-irradiation, CNO and CNG1 remained amorphous with no CaMoO 4 precipitation detected by XRD. Similarly, GCs contained a single powellite crystal phase, with no identifiable Na + or Na-Gd substitution of Ca 2+ within the CaMoO 4 structure. The powellite phase remained spherical in particle morphology and evenly distributed within the amorphous matrix, but a marginal increase in the range of PS was determined, as Table 3 summarises. This coincided with a reduction in the number of particles per unit area, which suggests that PS growth is diffusion based.
This shift in PS was mirrored by growth in both the powellite CS and cell parameters, as can be seen in Table 4. While irradiation-induced growth of both CS and cell parameters occurred for all samples (see Fig. 6), the expansion of the unit cell cannot alone account for the changes in CS, therefore an additional process must be taking place to enable CS growth. This is especially true for high Mo-bearing samples with [MoO 3 ] Z 7 mol%, where DCS 4 50 nm. These results imply either that there was some diffusion-assisted growth of crystallites, which changes to the range and density of PS support, or that some bulk-to-surface precipitation of powellite took place in these compositions. Alternatively, it could indicate a greater sensitivity of these compositions towards accumulated defects and a propensity for void formation within the crystal structure. This latter effect can be a direct result of displaced atoms following nuclear collisions, and would have a similar effect on PS as particles are made out of crystallites.
The magnitude of change for powellite cell parameters was also larger as [MoO 3 ] increased (see Fig. 6). This result corroborates the theory that the crystal phase becomes more susceptible to alteration as [CaMoO 4 ] increases. In contrast, changes to the powellite cell parameters often fell within the estimated standard deviation for compositions with [MoO 3 ] r 2.5 mol%. Nevertheless, expansion occurred in both unit cell directions and growth of CS was always observed in these compositions. Growth of the unit cell also occurred isotropically, which mirrors a temperature based response.
It is interesting that the CS growth induced by Au-irradiation was smallest for CNG2.5. This is an anomaly that has been seen in other studies on compositions with 1.5-4.5 mol% MoO 3 where an initial GC containing 2.5 mol% MoO 3 showed broader diffraction peaks indicative of increased Mo solubility in comparison to other compositions with CaMoO 4 formation. 8,70 This is mirrored by SDT, XRD and SEM results in this study, which show an increased retention of MoO 3 . A similar such property in CNG2.5 could subsequently have inhibited radiation-induced CS growth as well.

Composition and Raman modifications
Changes to CS are concurrent to shifts in the Raman bands associated with internal (MoO 4 ) 2À lattice vibrational modes for powellite with C 4h point symmetry. The main bands of interest are n 1 (A g ) 878 cm À1 , n 3 (B g ) 848 cm À1 , n 3 (E g ) 795 cm À1 , n 4 (E g ) 405 cm À1 , n 4 (B g ) 393 cm À1 and n 2 (A g + B g ) 330 cm À1 . [79][80][81] These vibrations are associated to: symmetric elongation of the   Fig. 7), but are found at lower wavenumbers in comparison to the listed vibrations for single crystals. As MoO 3 increased from 1.75-10 mol%, these vibrational modes experience a B0.8 cm À1 shift to higher wavenumbers, thus approaching theoretical values as the powellite cell parameters decreased towards an ideal crystal structure. Following Au-irradiation these vibrational modes are all still present, indicative of rigidity of the molybdenum tetrahedron, and thus resilience of the crystal phase. However, a shift of B1-2 cm À1 to higher wavenumbers was observed for all (MoO 4 ) 2À lattice vibrational modes, an example of which can be seen in Fig. 8. This is associated to an increase in the Mo-O bond length, 82 which has similarly been observed for CaMoO 4 under increasing temperature or pressure. 83,84 The effect of radiation on vibrational modes via a shift to higher wavenumbers was also dependent on composition, with a more significant change found at low [MoO 3 ], as Fig. 9 indicates. This could be attributed to existing defects within the crystal structures following synthesis that resulted in the (MoO 4 ) 2À lattice vibrational modes starting off at lower wavenumbers.
Simultaneous to a change in peak position was an observed peak broadening for all GC compositions, except CNG2.5 where the peak full-width decreased (see Fig. 9). Peak broadening   878 cm À1 and n 3 (B g ) 848 cm À1 ) in GCs illustrating a spectral shift to higher wavenumbers.
can be associated with distortion within the molybdenum tetrahedron or increased disorder in the stacking of (MoO 4 ) 2À units, whereas a decrease in broadening is associated with greater order within the crystalline environment. 85 Given that this composition showed some anomalies with an increased fraction of molybdenum anion incorporation, it also appeared to modify the radiation response, as CS changes confer. The increased incorporation of (MoO 4 ) 2À units is supported by the band at B910 cm À1 being largest for this composition (see Fig. 7). This is predicted to occur owing to the initial metastable equilibrium state of this composition, which is hypothesized to be associated with a higher fictive temperature (T f ) from internal quenching processes. Therefore, radiation serves initially as a relaxing agent increasing the order within the crystalline phase in this composition, after which structural defects form. While evident modification was observed for the vibrational modes of powellite, very little change was detected from the Raman spectra of fully amorphous samples (CNO and CNG1), as Fig. 7 indicates. In these samples there are several bands classically assigned to silicates or borates. The most significant of which is the R band at B450-520 cm À1 , which at lower wavenumbers is assigned to mixed Si-O-Si and Si-O-B bending and rocking, 77,86,87 and at higher wavenumbers to B-O-B rocking. 88 The second band of interest is at B633 cm À1 and can be attributed to Si-O-B vibrations in danburite-like B 2 O 7 -Si 2 O groups. 64,79 This band experiences minor damping relative to the R band following Au-irradiation indicative of modified ring structures. The last collection of overlapping characteristic bands can be found between B850-1250 cm À1 , and are representative of Si-O stretching vibrational modes for Q n units that represent SiO 4 tetrahedra with n bridging oxygen. 89 The maintenance of the first and last grouping of broad bands in their existing proportions indicates that the borosilicate network experiences little modification on average.
The Raman spectra for the glass containing 1 mol% MoO 3 also showed three broad bands at B330 cm À1 , B870 cm À1 and B910 cm À1 in addition to those listed above. The first two broad bands indicate that while crystallisation has not been observed in this composition, (MoO 4 ) 2À tetrahedron are similarly clustered and coordinated to Ca 2+ ions. 32 The broadness of these bands reflects that they nonetheless remained in an amorphous state. Following irradiation, these bands became even broader, indicative of increased disorder in the molybdenum environment, which can be associated to increased molybdenum dissolution. The band at B910 cm À1 representing Mo-O bond elongation in tetrahedral (MoO 4 ) 2À chains is a metric of this. 30 As this band moves to higher wavenumbers it indicates the dissolution of isolated (MoO 4 ) 2À units in the amorphous network or significantly distorted Ca 2+ -(MoO 4 ) 2À assemblies. In contrast, as it moves to lower wavenumbers crystallisation commences. 82,90 Following irradiation this band underwent the first transformation, with a shift of B2 cm À1 to higher wavenumbers. These results collectively imply that the molybdenum local environment was sensitive to radiation damage, while the borosilicate matrix showed resilience to modification.
In order to understand how the observed change to crystallinity and bonding occurred, EDS analysis was employed. Results in Table 5 indicate that as MoO 3 increased, Ca and Mo ions were leached out of the amorphous phase and clustered in the crystalline phase prior to irradiation. Furthermore, the fairly constant [Ca]/[Mo] ratio of B3.5 in the amorphous phase of all GCs implies that the influence of network modifying ions is independent of [MoO 3 ] following an initial concentration of 1 mol% MoO 3 .
Prior to irradiation, some oddities in concentration were noted representative of equipment limitations. Si was found in Fig. 9 Changes to the position and broadening of the Raman vibrational band n 1 (A g ) as a function of composition and external radiation.  Table 5). This implies cationic depletion at the surface, which has also been observed to occur with alkali ions, 63,91 and irradiation with both b-particles 69,70 and Xe 23+ ions, 92,93 which simulate the electronic components of b and a-decay. This type of surface depletion is due to an electric field gradient created from bombarding ions, which appears translational to irradiation with Au 3+ ions.
While an increase in [Si]/[Ca] and a decrease in [Ca]/[Mo] occurred in both phases, the fractional differences are larger in the amorphous phase. This could be a result of minor changes in the Ca-rich amorphous phase around crystalline particles, while the crystallites themselves remained compositionally unchanged. It could also suggests a secondary effect of cross phase diffusion of Ca ions from the amorphous to the ''crystalline'' phase, though this is difficult to ascertain owing to the significant mechanism of surface to bulk diffusion.
Based on a comparison between the changes seen for CNG7 and CN10, which initially have particles of a similar size, it can be concluded that the mobility of Ca ions increases for Gd-doped samples when exposed to irradiation. Moreover, the normalised [Gd] increases from B0.10 to B0.15 atm% in CNG1.75, CNG2.5 and CNG7 following irradiation. This increase could indicate a charge balancing substitution of Ca 2+ ions for Gd 3+ within the borosilicate network, which would subsequently enable the free migration of Ca 2+ ions, though this effect would be small given the relative dopant concentration.

Multi-scale phase separation
The compositions used in this study facilitated an excess of calcium for CaMoO 4 formation, and had a high enough concentration of B 2 O 3 to ensure that Na + ions were unavailable for Na 2 MoO 4 formation. In excess of 1 mol% MoO 3 , phase separation and consecutive crystallisation occurred during cooling and a single powellite (CaMoO 4 ) phase was observed within an amorphous network. A uniform distribution of spherical and unconnected powellite particles are indicative of a nucleation and growth process. 94 The first stage of this process involves liquid-liquid phase separation of the borosilicate network that created depolymerized regions of the glass where (MoO 4 ) 2À entities accumulate. The initiation of phase separation can be correlated to the field strength of modifying ions, such as Na + , Ca 2+ , and Mo 6+ , with the degree of phase separation proportional to the modifier field strength. 29,35 The second stage of this process involves Ca-O-Mo bond formation leading to crystallisation inside these regions during cooling. 32 The resultant CaMoO 4 crystallites were observed to precipitate on the micron scale following synthesis and took the form of scheelite-type powellite with a tetragonal structure.
The CS values and cell parameters for these synthesised crystals align with studies for similar soda lime borosilicate GC compositions, 8,30,34 as well as calcium borosilicate GCs with a high MoO 3 content. 69 A variance was found to exists between the lattice parameters for crystals within an amorphous matrix, and those of a monocrystal (a = 5.222 Å and c = 11.425 Å). 95 In general, synthesised GCs had a larger unit cell compared to powellite single crystals in both this and previous studies, which suggests that the very nature of an amorphous structure independent of composition causes this effect. It is predicted that contraction of the borosilicate framework during the glass transition process subsequently creates tensile stress on CaMoO 4 crystals. This transition inhibits a full relaxation of powellite crystals to room temperature equilibrium, thus creating the discrepancies in lattice parameters.
In this study, a compositional effect was observed for both of these metrics with CS expansion and lattice parameter contraction occurring for increasing [MoO 3 ]. PS was similarly proportional to increasing [MoO 3 ]. These composition-dependent shifts can be correlated to a change in the temperature of phase transitions that consequently controls the molybdenum environment. In previous studies of sodium borosilicates, the phase separation temperature (T PS ) was observed to increase by B18 1C per mol of MoO 3 . 35 Comparatively, Magnin et al. observed a B50 1C increase in T PS and a B40 1C increase in the crystallisation temperature (T C ) of CaMoO 4 for every 0.5mol% increase in [MoO 3 ]. 34 Increases in T PS and T C would produce larger crystallites that form into larger particles for a fixed synthesis temperature. These crystallites would also have a longer period of relaxation before a glass transition occurred and the matrix goes from liquid to solid. This longer relaxation would subsequently result in smaller lattice parameters. Given that these trends were recorded for increasing [MoO 3 ] in this study, it can be concluded that these principles of increasing T PS and T C apply to these GCs. Similar trends were also observed for T M and H Tm with increasing amounts of MoO 3 , indicating that T PS and T C are related to the melting of crystallites. When T PS and T C increase, larger crystallites with smaller cell parameters are formed, which represent embedded powellite crystals approaching single crystal ideality. A larger amount of these crystals are also formed, which is supported by the larger (B3Â in size) PS found for higher [MoO 3 ]. As a growing number of crystals that are approaching ideality in structure form, a higher temperature would be required to induce thermal motion and eventual melting. The less ordered crystal structures formed when T C is closer to T g would in comparison require less thermal energy to induce amorphisation. For this reason T M is predicted to increase with T PS and T C , and therefore [MoO 3 ]. Theoretically T C and T M are presented as overlapping Gaussian profiles on the Gibbs free energy versus temperature plot. 50 Accordingly, the proportional relationships observed here are not unexpected.
While this trend with respect to [MoO 3 ] held for PS and powellite lattice parameters, a deviation occurred with respect to CS and T M . In this study, CS was initially B20 nm larger for CNG7 than for CN10 despite a 3 mol% difference in [MoO 3 ], and the rate of change following irradiation was likewise larger. Similarly T M was B20 1C cooler for CN10 than for CNG7. These differences can be attributed to the gadolinium dopant, which is predicted to improve crystallisation kinetics during synthesis, in addition to growth kinetics following external radiation. This is in contrast to the theory that REs such as neodymium prevent the initial phase separation of molybdenum-rich regions by disrupting coordination in depolymerized regions of the glass. 30 However, melt crystallisation tendency has been observed to vary with both the type of alkaline earth and RE. 37,40 This observation thus outlines the effects of specie dependent trace incorporation of REs, which may or may not translate to actinides in nuclear waste glasses or GCs.
Crystallisation of CaMoO 4 was a clear process taking place in GCs, however results interestingly suggest that the residual matrix was also phase separated. This is based on the presence of two T g s, which imply the formalization of two distinct structural states. 52 Yet, glasses and the residual matrix of GCs were not characterised by distinct micro-phase separation and remained amorphous, according to SEM and XRD respectively. Furthermore, random Raman spot analysis at the surface gave almost identical spectra (see ESI †), as opposed to those with different proportions of silica and borate-type features indicating that any type of phase separation in these glasses occurred uniformly within the spot size.
It is therefore hypothesized that phase separation in CNO occurred through spinodal decomposition. This process results in strongly interconnected phases with a similar chemical composition that are separated by a diffuse interface, and are thus difficult to discern using SEM microscopy and EDS analysis. 96 This type of heterogeneity on the nano-scale can result in multiple glass transitions, as has been previously predicted. 78 Kobayashi et al. theorised that multiple transitions would occur through the emergence of very small variations in the intermediate-range order (IRO) resulting in a bimodal nano-structure periodically distributed throughout the amorphous network. 97 Moreover, these types of transitions have shown a dependency on composition and the relative cationic content, which determines structural mobility. 98,99 In silicates, the addition of alkali and alkaline earth oxides causes a decrease in T g , through the formation of additional NBOs that increase thermal motion of the glass network with temperature. Whereas, low alkali or alkaline earth concentrations can increase T g in borates through the creation of BO 4 À entities with a proportional dependence on the cationic size as this affects spatial hindrance. 99 Therefore, the population of bridging oxygen and the [BO 3 ]/[BO 4 À ] fraction, which was observed to vary with composition in this study are important factors in determining the IRO, and therefore the transition temperature. 100 Based on this existing knowledge a model is put forth to explain the type of nano-phase separation occurring in these compositions. It is predicted that the two interconnected phases in CNO are attributed to a variations in Si-O-Ca 2+ -O À -Si and Si-O-Ca 2+ -O À -B units. As the CaO content increases with MoO 3 , the separation of borate groups from the silicate network increases, thus resulting in de-mixing of Si-O-Ca 2+ -O À -B units, as has been observed to occur in similar compositions. 51,76 This shift causes the relative changes to the enthalpies of T g1 and T g2 observed in Table 2.
The simultaneous inclusion of MoO 3 distributed between these two phases causes a secondary process of nucleation and growth evident on the sub-micron scale. It is predicted that the first step in this process requires the separation of a MoO 3 -CaO-B 2 O 3 -NaO 2 rich phase associated with T M or the liquidus temperature. Therefore, it is theorised that MoO 3 relies on the de-mixing of Si-O-Ca 2+ -O À -B units. It can further be deduced that decreasing the initial [SiO 2 ], while increasing [CaO] could promote mixing of these two IRO units and therefore reduce secondary Mo phase separation.

Crystallite modification
Radiation was observed to modify the crystal phase in a process of significant CS and lattice parameter expansion. It is hypothesized that this is occurring due to two reasons. The first is a temperature-based component, and the second is an accumulation of defects induced by atomic displacements.
Although Au-irradiation was primarily employed to replicate ballistic collisions, it does not omit electronic interactions, which EDS results indicate are occurring from the observed cationic depletion at the surface. As 7 MeV Au 3+ ions travel through a material, energy from the ion can be transferred to the host lattice electrons via electron-electron and electronphonon coupling. The thermal spike model translates these interactions into a small cylinder of energy characterised by a temperature of B1000 K. 101 It was later theorised that electronic coupling at the site of a nuclear interaction could also create a melt halo around every collision. 60 Therefore, external radiation can theoretically induce temperature-like effects. In this study, changes to the powellite lattice parameters following irradiation are similar to those typically seen for heat treatments between 100-200 1C. 102,103 Furthermore, the isotropic nature in which expansion occurred supports a temperaturelike effect.
Thermal expansion is predicted to contribute to the radiation-induced alteration of crystallites, yet changes to powellite lattice parameters cannot alone account for changes to CS in compositions with [MoO 3 ] Z 7 mol% (where DCS 4 50 nm). Theoretically, additional crystallite growth could have been initiated by a high-temperature diffusion of ions, but this kinetic process has an activation barrier that is affected by the glass structure and requires time at temperature. 104 It is therefore predicted that the accumulation of damage is the driving force to additional structural changes.
As previously mentioned, internal radiation can cause structural modifications that result in changes to mechanical properties and to volume. 6,[53][54][55][56][57] In a crystal structure, these modifications can present themselves as swelling through the displacement of atoms, which can create voids, mobile interstitials that can cluster, and other structural defects such as dislocations and interfacial cracking. 10,105,106 The accumulation of this type of damage from elastic collisions has experimentally been observed to increase T f through a process of ballistic disorder and fast quenching. 65,107 Unit cell expansion and swelling has been previously observed to occur in powellite single crystals undergoing Ar-irradiation, though alteration occurred anisotropically. 85 In this case, expansion occurred through the formation of defect clusters that turned into dislocations as the dose increased to 1 dpa. In this study, increasing disorder was supported by shifts in Raman vibrational modes of (MoO 4 ) 2 , which suggests defects in the stacking of Ca polyhedra along the c axis and the Mo tetrahedra along the a axis. It is therefore assumed that this type of transformation was occurring in irradiated GCs, as opposed to simply a thermal process resulting from electronic coupling, but that it is not the only process given the isotropic nature of modifications observed here.
The mechanism of CS alteration in this study thus combines elements of thermal expansion and defect accumulation leading to crystallite swelling. It can further be summarised that increasing [MoO 3 ], and therefore the fraction of CaMoO 4 within the borosilicate matrix, accelerated damage accumulation within the crystalline structure, and therefore CS expansion.
It can further be deduced from comparison of studies examining the irradiation of GCs with other ions that expansion of powellite CS and lattice parameters has only been found for damage emulating nuclear interactions. b-Irradiation was observed to result in a unit cell contraction with a non-linear CS trend with respect to dose. 69,70 Here the unit cell alteration was hypothesized to occur from thermal relaxation of the crystalline lattice following the energy inputted from irradiation, with changes to CS resulting from the accumulation of point defects. In contrast, irradiation with Xe 23+ ions, 92,93 which replicated the electronic component of a-decay, caused a reduction of CS concurrent to a dose dependent anisotropic unit cell expansion along the c-axis, which led to partial amorphisation. 92,93 Combined, these results suggest that the components of internal radiation will produce opposing forces on structural modifications within powellite crystallites, and that electronic interactions may reduce displacement induced swelling and unit cell expansion.

Ion impact on phase transformations
Following Au-irradiation results confirm a single crystalline phase of powellite (CaMoO 4 ). Had incorporation of Gd 3+ ions into CaMoO 4 taken place, or if Na 2 or Na 0.5 Gd 0.5 substitution of Ca 2+ occurred this would have been reflected in XRD and Raman results. Refinement of the calcium site occupancy in CaMoO 4 would have revealed substitution, particularly for Gd 3+ , where the covalent sizes are substantially different, as has been previously determined. 37,90 Moreover, substitution of Gd 3+ into the powellite structure would have converted the point group symmetry of the octahedrally coordinated Ca 2+ end-member from S 4 to C 2 , which would have altered spectroscopic emissions, in addition to XRD. 108 In this study, no such substitution was detected following refinement of XRD patterns, nor was such inclusion evident in Raman peak analysis.
In addition to a lack of detected Ca 2+ substitution, SEM imaging reported the growth of particles, which coincided with CS expansion for all GCs. This occurred alongside isotropic lattice parameter expansion, most evident in compositions with [MoO 3 ] 4 2.5 mol%. The collective results answers the first two questions posed in this paper in that radiation damage did not induce significant amorphisation, nor did it induce cationic substitution of CaMoO 4 . If amorphisation were occurring, the opposite trend in PS and CS would have been produced, in addition to XRD whole pattern damping, and growth of the Raman band at B910 cm À1 assigned to (MoO 4 ) 2À dissolved in the amorphous network. These changes have been seen for irradiation-induced amorphisation with higher energy ions, 93 though no amorphisation of CaMoO 4 has been previously reported for irradiation with medium-energy ions like 7 MeV Au 3+ . 85,109 The third question to be answered was whether irradiation damage would induce precipitation. Increases in the range of PS, together with a decrease in the particle density at the surface suggest a rearrangement of crystallites as opposed to additional precipitation. This result is supported by cationic depletion at the surface, whereas CaMoO 4 formation would have required Ca 2+ ions to be in excess within the molybdenum environment.
In terms of the amorphous phase very little changes were observed in terms of short to medium-range order. In glasses, the residual matrix showed a minor reformation of ring structures away from danburite-like units and the molybdenum environment showed increasing disorder, which is predicted to discourage crystallisation. Unfortunately, no information could be gained on how microheterogeneities of the matrix changed following irradiation and if this could have played a role in the observed modifications to the powellite phase.

Conclusions
This paper focused on the fundamental processes occurring in glasses and GCs containing MoO 3 as a method to investigate matrices that can increase the incorporation of the fission product molybdenum. Several glasses and GCs were synthesised with up to 10 mol% MoO 3 within a soda lime borosilicate matrix. Two types of phase separation were observed in these systems. The residual glassy matrix is predicted to be phase separated on the nano-scale through a mechanism of spinodal decomposition resulting in small variations in the amount of Si-O-Ca 2+ -O À -Si relative to Si-O-Ca 2+ -O À -B units, with increases in [CaO] causing the de-mixing of these two types of IRO units. Above 1 mol% MoO 3 , a secondary process of nucleation and growth of CaMoO 4 crystals occurred, which pulled Ca and Mo ions out of the residual matrix and toward crystal centres, according to EDS analysis. This process did not significantly alter the presence of heterogeneities in the glass matrix according to thermal analysis, nor the coordination of boron according to 11 B MAS NMR.
These materials were then subjected to a dose of 3 Â 10 14 ions per cm 2 using 7 MeV Au 3+ ions to replicate the damage caused by internal a-decay and thus determine radiation tolerance. This created 1 dpa of damage for a depth of B1.5 mm, according to TRIM calculations. Using this experiment we could test the theory of whether irradiation would: (1) cause amorphisation of CaMoO 4 , (2) change the crystal structure or induce cationic substitution of CaMoO 4 , or (3) induce precipitation of alternative phases. Results imply that amorphisation or precipitation did not occur based on changes to the size and density of crystalline particles. Furthermore, Na 2 /NaGd-MoO 4 assemblages were likewise found not to form through direct crystallisation or substitution, as both would have introduced new XRD and Raman peaks, which were not detected. Au-irradiation was however found to cause expansion of both the unit cell and CS of CaMoO 4 . The mechanism in which this occurs is based on two processes. The first is a thermal expansion resulting from heat generated by electronic interactions. The second is swelling from the creation of defect clusters, such as voids resulting from the accumulation of atomic displacements following ballistic collisions. The collective results suggest that while modification to the crystalline phase of interest may be occurring, they do not alter the nature of this phase, therefore proving that these GCs are tolerant to radiation damage.

Conflicts of interest
The authors declare that are no conflicts of interest.