Characterization of immiscibility in calcium borosilicates used for the immobilization of Mo 6+ under Au-irradiation

The aim of this paper was to assess factors affecting primary and secondary phase separation in simplified calcium borosilicate glasses studied for nuclear waste applica-tions. Several glasses with varying [MoO 3 ] and [B 2 O 3 ] were synthesized and exposed to Au-irradiation to examine compositional effects on glass structure and domain size of separated phases induced by accumulated radiation damage resulting from α-decay over a ~1000 year timeframe. The produced glasses fell within the immiscibility dome of CaO−SiO 2 −B 2 O 3 and showed a unique microstructure of embedded immiscibility with three identifiable amorphous phases according to electron microscopy, Raman spectroscopy, and diffraction. These glasses were then bombarded with 7 MeV Au 3+ ions to a dose of 3 × 10 14 ions/cm 2 creating an estimated ~1 dpa of damage. Several changes to the morphology, spatial distribution, and size of secondary phases were observed, indicative of significant structural reorganization and changes to the chemical composition of each phase. A general mechanism of coalescence to form larger particles was observed for [MoO 3 ] < 2.5 mol%, whereas segregation to form smaller more evenly distributed particles was seen for [B 2 O 3 ] ≤ 15 mol% and [MoO 3 ] ≥ 2.5 mol%. These microscopic changes were concurrent to surface-bulk diffusion of Ca and/or Mo ions, where the direction of diffusion was dependent on [B 2 O 3 ] with a barrier identified at ~20 mol%, as well as cross-phase diffusion of said ions. These modifications occurred in part through the formation of distorted ring structures within the borosilicate network, which enabled the increased dissolution of isolated (MoO 4 ) 2− units. Au-irradiation was therefore able to increase the solubility of molybdenum and alter the structure and composition of secondary phases with the extent of modification varying with [MoO 3 ] and [B 2 O 3 ]/[SiO 2 ], though glasses notably remained heterogeneous. The collective results suggest that radiation and composition can both be used as design tools to modulate the domain size and distribution of separated phases in heterogeneous glasses.


| INTRODUCTION
The storage of nuclear waste is a complex problem, as materials are required to encapsulate a wide variety of radionuclides, show resistance to chemical leaching and radiogenic heat, in addition to maintaining structural integrity following internal decay processes. [1][2][3][4] In many countries, these wasteforms will be buried in a near surface or deep geological repository (~500 m) for millennia. 5 This places a heavy burden on the wasteform to maintain physicochemical properties over long timescales, as nuclear waste will be radioactive for longer than several human lifetimes. 6 It is common practice to immobilize high-level waste (HLW) in a borosilicate or aluminoborosilicate matrix, 6 as these amorphous structures exhibit many of these capabilities. [1][2][3] They can also be manufactured under reasonable temperatures and pressures, necessary for large scale industrial applications. However, waste loading is limited to ~18.5-20 wt% in French borosilicates and US aluminoborosilicates, 7,8 in order to prevent the uncontrolled crystallization of molybdates, rare-earths, silicates, and chromites, [9][10][11] which can act as carriers for radioactive cesium, strontium, or minor actinides (MAs) and lanthanides (Ce 3+ , La 3+ , Sm 3+ , Nd 3+ , Gd 3+ , Y 3+ ). 1,12,13 High concentrations of molybdenum are expected from weapons-grade waste, postoperative clean out, increased fuel utilization, or Generation IV reactors. 1,14,15 It is also a concern if increased waste loading is to be considered, as a means to minimize the final volume of waste for long-term storage. In excess, the fission product molybdenum can lead to the formation of water-soluble alkali molybdates (Na 2 MoO 4 , Cs 2 MoO 4 ), known as yellow phase. 12,16 Given that these compounds exhibit a high water solubility and can act as carriers for radioactive elements, 1,15 the formation of this complex can severely alter the chemical durability of the wasteform during long-term storage. 14,17 In its stead, selective formation of CaMoO 4 , which is 13,500× less soluble than alkali molybdates 18 has become of industrial interest.
In reprocessed waste streams, molybdenum will primarily be hexavalent taking the form of (MoO 4 ) 2− tetrahedral units in oxidizing or neutral conditions. [19][20][21] According to Greaves' random network model for modified silicates 22 and experimental Mo EXAFS, 9,20,23 these tetrahedral units are unconnected to each other and to the surrounding borosilicate framework, and are instead located in non-bridging oxygen (NBO) channels lined with charge-balancing cations. In this configuration, these cations are octahedrally coordinated and bound to (MoO 4 ) 2− entities by weak long-range ionic forces in either the glassy or crystalline phase, [23][24][25] hence why molybdenum has a limited solubility in glasses. 19 The ability to incorporate molybdenum is dependent on the initial glass structure. An amorphous network is classically defined as a random aperiodic structure that lacks the long-range order assigned to crystals making it an ideal host for a variety of nuclei of varying size and charge. There is, however, some general short-and intermediate-range order within glasses that is related to the coordination and covalent bonding of network-forming atoms such as Si, B, Al, or P located within an oxygen sublattice. 22,26 These atoms typically take the form of XO 4 tetrahedral units, where X = Si, B, Al, P. These tetrahedra are connected to each other via X−O−X bridging bonds forming large chains and/or rings. In the case of silicates, the degree of polymerization is annotated by Q n units, where n represents the number of bridging oxygen on SiO 4 tetrahedra. 27 When n is less than 4, a NBO is present. In some cases, these NBO sites will have a residual negative charge that requires charge compensation. This is also the case for (AlO 4 ) − and (BO 4 ) − tetrahedral units, where X 3+ , and thus these units require both an extra oxygen from an added alkali or alkaline earth oxide, as well as charge compensation to maintain electrical neutrality. These types of network-forming units will also be found in their trigonal forms, where XO 4 /XO 3 will have a significant effect on several glass properties. 28,29 This outlines the complexity of borate glasses over silicate glass, where the former offers a much larger range of configurations between other network formers and modifier atoms. 22 This extends to borate-silicate mixed glasses known as borosilicates.
This introduces the second major component of glasses, which are network modifying units constituted of cations, such as alkalis, alkaline earths, or rare earths. These cations can be further divided into network modifiers, which break up the network through the creation of NBOs, and therefore favor phase separation; and charge compensating network formers, which counteract the effects of immiscibility by stabilizing network formers. 30,31 The coordination and connectivity of network-forming oxide units will determine the role of cations. Similarly, the presence of cations can alter the ratio of XO 4 /XO 3 for X 3+ , and therefore other properties of the glass such as viscosity and ion transport. [32][33][34][35] If the ionic bonds between network formers and cations break following an external treatment, the cations will be free to migrate and cluster in depolymerized regions of the glass. This clustering of cations can lead to phase separation [36][37][38][39][40] and the precipitation of complexes such as molybdates. 19 The specific speciation of a CaMoO 4 type molybdate complex in a borosilicate glass can be induced by composition, 12,23,41 controlled cooling during synthesis, 16 external heat treatments, 42,43 rapid quenching, 16 or through redox chemistry. 44,45 In this paper, glass composition will form the primary point of interest with a focus on factors affecting the competitive charge balancing of network forming (BO 4 ) − and network modifying (MoO 4 ) 2− anionic entities by a limited cation content, as this can significantly effect phase separation tendencies. 46 In soda lime borosilicates, both of these anionic units prefer Na + as charge balancers, but (BO 4 ) − network units will have a stronger affinity to this ion, thus enabling the incorporation of Ca 2+ into molybdate complexes. 9,19 Previous studies have found that an increase in [B 2 O 3 ] or [CaO] will promote the formation of CaMoO 4 over Na 2 MoO 4 crystallization, as it affects the population of (BO 4 ) − species. 12,46 However, the concentration of CaO in soda lime borosilicate glasses has typically been limited to <11 mol%, as it is known to cause liquid-liquid phase separation in borosilicate glass. 12,15,47 This limitation has resulted in Na 2 MoO 4 production in most studies, despite the manipulation of the other listed factors affecting speciation. 12,21 For this reason, Na 2 O was excluded from this study even if that resulted in multiphased glasses. This was performed as a means to isolate the key components of CaMoO 4 formation, and to better understand the principles of phase separation as a function [B 2 O 3 ] and [MoO 3 ], in addition to factors affecting the solubility limit of molybdenum within matrices containing varying proportions of network formers (borate and silicate units). It further sought to examine how glass chemistry would affect structural evolutions induced by radiation damage as a metric for long-term performance, and at the same time to assess if internal radiation could be used to create a desired structural configuration.
Materials used for the storage of nuclear waste will continuously undergo internal damage from α-decay of minor actinides and Pu, β-decay of fission products, and transitional γ-decay. These processes will trigger atomic displacements, ionization, and electronic excitations, which can result in changes to the short-and medium-range ordering within the glass structure. While the former is induced by changes to the coordination of atoms resulting from a modification in charge compensation, the latter will result from changes to bond angles between network formers and the distribution of ring sizes. 48,49 These modifications can subsequently lead to changes in volume (swelling or densification), and mechanical properties (hardness, elasticity, and fracture toughness). It can also induce phase transformations, such as devitrification, precipitation, bubble formation, glass-in-glass phase segregation, or the clustering of cations. 6,48,[50][51][52][53] These effects will be predominantly of interest under the scope of multiphased structures as the [SiO 2 ]/[B 2 O 3 ] ratio, and the concentration of cations has also been observed to alter bond angles and the domain size of separated phases. 33,[54][55][56] Therefore, radiation may accelerate or remediate the effects of composition on glass order and subsequently phase-specific properties.
During encapsulation, the α-decay process and specifically the heavy recoil nuclei are responsible for the bulk of these modifications according to both experimental work and to molecular dynamics (MD) simulations. 6,48 During α-decay, an α (or He 2+ ) particle is ejected from a parent isotope. This small α-particle typically has a kinetic energy between 4.5 and 5.5 MeV and will travel 10-20 μm before stopping due to interactions with electrons in a glassy matrix. 6 The decay process will also produce a daughter isotope (parent isotope-He 2+ particle), which is referred to as the heavy α-recoil. This isotope will have a relatively smaller kinetic energy around 70 to 100 keV, and will interact primarily through nuclear (or elastic) collisions. These nuclear interactions will cause a chain reaction that creates a dense displacement cascade, resulting in around ~1500 atomic displacements after travelling ~30 nm. 6 In order to specifically replicate the damage produced from these overlapping displacement cascades on an accelerated timescale, external radiation using medium-energy ions can be used. 48,57,58 Previous research has shown that borosilicate glasses subjected to internal or external irradiation have primarily remained amorphous, 6,48 with a saturation in changes to mechanical properties, internal energy, and density observed for a cumulative dose of 4 × 10 18 α/g, 48,59,60 where an equilibrium between damage creation and thermal annealing from overlapping ion tracks occurs. 48,59 Given current waste loading standards and waste streams, this saturation in structural modifications is expected to occur following ~1000 years, 6,48 which corresponds to approximately 1 displacement per atom (dpa). 58,59 Using this information, experiments can be designed to simulate a maximum damage state, which can be used as a tool to reverse engineer a desired starting encapsulating medium.
This paper principally sought to identify if irradiation simulating the damage from α-decay would: (a) affect the solubility of molybdenum and potentially cause the formation of CaMoO 4 ; (b) promote phase separation in the borosilicate network; or (c) alter existing separated phases, thereby affecting immiscibility tie lines similar to temperature-based effects. In this context, solubility refers to the inclusion of isolated (MoO 4 ) 2− tetrahedron within the borosilicate network, versus those that are clustered together in depolymerized areas of the glass that can serve as precursor environments for crystallization. 19,61,62 Additionally, the terminology for phase transitions refers to both those between amorphous and crystalline phases, and between multiple amorphous phases that are compositionally and structurally different based on network connectivity.

| Composition and synthesis technique
Multiple compositional series were created with varying [B 2 O 3 ] and [MoO 3 ] to study the solubility and phase separation tendencies of molybdenum in calcium borosilicate glasses, and to explore how separated phases were modified when exposed to external radiation that displaced each atom. The first series (labeled CB) has an increasing [B 2 O 3 ] for a fixed [SiO 2 ]/[CaO] ratio and MoO 3 content. This series was used to determine how the ratio of network formers altered the glass structure and phase-specific properties. These samples also contained a fixed amount of rare-earth dopant, which can be considered an actinide surrogate, and can therefore indicate potential incorporation sites of radioactive elements.
The second series (labeled CM) focused on increasing the molybdenum content in simplified calcium borosilicates normalized to French nuclear waste glass SON68 (non-active form of industrial R7T7). It was used to assess the mechanism of MoO 3 incorporation and to test solubility limits, in addition to understanding how encapsulation would be affected by atomic displacements. See Table 1 for the batch compositions within these series. Owing to the heterogeneous nature of the samples it was difficult to confirm the experimental composition, but energy-dispersive X-ray spectroscopy (EDS) and 11 B MAS NMR were used for this purpose in so far as it was possible 63,64 (see Supporting Information).
Note that while changes to [MoO 3 ] and [B 2 O 3 ] are referred to throughout the text, molybdenum will take the form of (MoO 4 ) 2− complexes unconnected to the glass network, as previously mentioned. Similarly, boron will be found as BO 3 or (BO 4 ) − structural units that are connected to other borate or silicate units. While different fractions of tri-fold and tetra-fold co-ordinated boron will alter the connectivity and properties of the glass, they are often lumped together for simpicity in these otherwise complex multiphase glasses. This study thus remarks on the general effect boron oxide has on glass structure.
Glass batches of ~30 g were prepared by milling powders of SiO 2 , H 3 BO 3 , CaCO 3 , MoO 3 , and Gd 2 O 3 , followed by melting in a platinum-rhodium (90/10) crucible at 1500°C. Mixed powders were melted for 3 hours, before crushing and re-melting glasses for 2 hours. A double melt and variance in melt time was used to try and ensure homogeneity of elemental distribution, but all calcium borosilicate compositions remained multiphased despite various techniques to increase the quench rate and induce mixing at high temperature. 63 Due to the high viscosity of these samples, they needed to be quenched in a water bath and tapped out with a hammer. Fragments from this process were annealed for 24 hours at 520°C to reduce minor internal stresses. They were then cut to fit the irradiation sample holder (~4 mm × 4 mm) and hand polished successively using P600, 800, 1200, 2400, and 4000 SiC grit paper, followed by 3 and 1 μm diamond polishing to achieve a uniform thickness of approximately 500 μm.

| Irradiation experiment
Energetic ions can be used to replicate the effects of various irradiation processes, such as neutron damage, β-decay, γ-decay, or α-decay. 6,65,66 The α-decay process is unique in that it will have both electronic and nuclear components, which can be replicated using certain ions and energies. The nuclear stopping power resulting from slow heavy ion irradiation, such as Pt, Au, and Pb ions, is similar to that of the α-recoil (5 keV/nm). 57 Therefore, external radiation experiments of this design can reasonably replicate the damage produced by displacement cascades arising from ballistic events.
This radiation experiment was conducted at CSNSM using 7 MeV Au 3+ ions at a flux of 2 × 10 10 ions/cm 2 ·s based on a target current of 0.15 mA on the ARAMIS beamline in France. A dose of 3 × 10 14 ions/cm 2 was achieved creating approximately 1 dpa. The ion penetration depth was correspondingly estimated to be ~1.6 μm, based on TRIM calculations 58 (see Supplementary Information). This degree of damage can be used to assess the structural modifications within the glassy phase(s) when every atom has been displaced, and whether this will affect the solubility of molybdenum.

| Characterization techniques
The initial microstructure of synthesized samples often contained two to three phases, either a crystalline and glass phase, or three compositionally different glass phases. As a result, multiple analytical techniques were required to assess the modifications to structure induced by irradiation.
Owing to the small irradiation volume, characterization was limited to surface analysis using accessible techniques that would describe the short-and medium-range order of each phase. The synthesized multiphase glasses and glass ceramics (GCs) presented varying morphologies and phases for quantification, which were analyzed using Scanning Electron Microscopy (SEM) backscattered (BSE) imaging and energy-dispersive X-ray spectroscopy (EDS). These were performed on a Quanta-650F held at low vacuum (0.06-0.08 mbar) with a 5-7.5 keV beam, resulting in a penetration depth of ~1 μm. EDS measurements were collected with an additional 8 mm cone, to reduce skirting effects, thus providing information on the interface between phases, and the relative composition of each identifiable phase. Images were collected using FEI Maps software, whereas acquisition and analysis for EDS was performed using Bruker ESPRIT software from 5 to 15 mapped areas for each phase.

Series
While electron microscopy was useful for investigating the microstructure of these complex glasses, Raman spectroscopy was used to determine bonding configurations. This was particularly useful in identifying the characteristic medium-range order of each amorphous phase. A Horiba Jobin Yvon LabRam300 spectrometer with a 300 μm confocal hole, coupled to a Peltier cooled front-illuminated CCD detector with a holographic grating of 1800 grooves mm −1 was employed. It utilized a diode-pumped solid-state laser (Laser Quantum) with an incident power of 100 mW focused with an Olympus 50× objective to produce a 532 nm excitation line. Spectra were collected over a 150-1600 cm −1 range with a 2 μm spot size, which resulted in a spectral resolution of ~1.4 cm −1 per pixel (1024 × 256 pixels in size). Multiple acquisitions (3)(4) were made for each phase within each sample.
While most of the compositions in this study were glasses, X-ray diffraction (XRD) was primarily used to confirm this was true in the observed heterogeneous structures, and to identify any molybdate formation. A Bruker D8 ADVANCE spectrometer using CuK α1 (λ = 0.15406 nm) equipped with Göbel mirrors for a parallel primary beam, and a Vantec position sensitive detector was enlisted for this purpose. Acquisitions were made over the 2θ = 10-90° spectral range with a 0.02° step size, and a 10 s per step dwell time, resulting in a ~10 hours run per sample.

| Microstructure
The microstructure of synthesized calcium borosilicate glasses in the CB and CM series showed embedded immiscibility with three identifiable phases prior to irradiation, as a result of compositions falling under the immiscibility dome in the SiO 2 -B 2 O 3 -CaO system. 47 The residual matrix is Si-rich and referred to as phase A. Within this phase, there are several large droplets rich in Ca/Mo referred to as phase B. These regions of phase B are found to form in varying sizes, and are randomly distributed within phase A. Within phase B there are smaller droplets of a glassy phase that is compositionally a mixture of phases A and B. 63 These droplets are referred to as phase C and are found in an array of geometries. The most common of which are referred to as spherical, distended (a sphere with rounded protrusions), and spatters (resembles the shapes created by paint splatters). A schematic of these heterogeneous structures and their terminology can be seen in Figure 1. Owing to the complex nature of these heterogeneous structures, changes following irradiation focused on making comparisons to the morphology, size, and distribution of phase C within phase B.
While this general structure of embedded immiscibility was observed in most of the samples, variations in size, distribution, and morphology of phase B and phase C deposits could be observed in the CB and CM series. In the CB F I G U R E 1 Schematic of embedded immiscibility within synthesized calcium borosilicates. 49 Phase A is a Si-rich residual phase within which heterogeneously distributed phase B deposits rich in Ca/Mo and of varying size are located. Within these deposits are smaller phase C droplets, which are found to form in a variety of geometries as labeled above. [Color figure can be viewed at wileyonlinelibrary.com] series, phase B deposits are initially found to form in small (2-18 μm) and medium (75-150 μm) sizes. As [B 2 O 3 ] increases, the number of phase B deposits increases and larger deposits (>200 μm) could also be observed. This result suggests that phase B is a carrier for boron.
Within these deposits, the distribution and number of phase C droplets also altered as [B 2 O 3 ] increased. At low [B 2 O 3 ], phase C droplets were primarily found in a spattered morphology with a ~3-18 μm in diameter range. As the regions of phase B grew with increasing [B 2 O 3 ], so too did the size of phase C droplets forming those up to 50 μm in diameter in an increasingly spherical morphology. This change in the size range of phase C droplets was concurrent to a shift in the distribution of phase C within phase B. As [B 2 O 3 ] increased, phase C droplets were found to form closer to the A-B interface and these droplets were larger in size compared to those in the center of phase B.
Following Au-irradiation, the morphology of the spattered phase C droplets started to coalesce into a spherical geometry for [B 2 O 3 ] ≤ 15 mol% (see Figure 2). This was most pronounced in the center of phase B where the average diameter of phase C droplets also decreased by ~1/3 of the original size. Comparatively, the sample with [B 2 O 3 ] > 15 mol% showed a dispersion of spherical droplets toward that of a spattered morphology. This was particularly evident near the A-B interface as Figure 2D illustrates.
Synthesized glasses in the CM series also exhibited a change in morphology of phase B deposits from ellipsoidal to spherical as [MoO 3 ] increased. These deposits also experienced a marginal increase in size with increasing [MoO 3 ]. This observation suggests either that phase B acted as a carrier for Mo, in addition to boron and Ca ions, or that MoO 3 induced further phase separation of network formers resulting in larger phase separated deposits. In terms of phase C, there was a small increase of ~10 μm in the size distribution of particles with increasing [MoO 3 ], and larger droplets were similarly found near the A-B interface. For compositions with [MoO 3 ] > 2.5 mol%, phase B became a powellite phase with crystallites forming clusters embedded in a residual amorphous phase.
Following Au-irradiation, the sample without MoO 3 (CaBSi) showed several changes to the size and distribution of phase C deposits within phase B (see Figure 3). First, there was the formation of a belt of small deposits (< 4 μm) close to the A-B interface. Second, the size of the remaining phase C droplets increased to ~5-40 μm in diameter. Third, these newly formed larger droplets appeared to be undergoing a process of dispersion, following an initial coalescence of smaller droplets. This suggests that modification to separated phases may be dose-dependent.
Once MoO 3 was introduced into the system, the effects of irradiation varied depending on [MoO 3 ]. For low [MoO 3 ], the formation of larger phase C deposits up to a diameter of ~50 μm similar to that seen in CaBSi was observed (see Figure 4C). As [MoO 3 ] increased to CM2.5, phase C droplets were observed to experience a significant reduction in size, with most droplets found to be <5 μm, as Figure 5 indicates. These droplets also appeared to become more evenly distributed within phase B. These collective observations indicate a possible threshold at 1 < [MoO 3 ] < 2.5 mol% where the mechanism of alteration shifts from coalescence to dispersion of phase C droplets. When [MoO 3 ] increased to 7 mol%, the system shifted from being one of embedded immiscibility with three amorphous phases to that of a GC with heterogeneously distributed CaMoO 4 crystal clusters within a residual amorphous network. This was determined by XRD, which confirmed the formation of CaMoO 4 and by BSE imaging, which showed only two phases. Following irradiation very few alterations can be seen with regards to the morphology or distribution of crystal clusters (see Figure 4), indicating stability and microstructural radiation resistance of this GC.

| Compositional changes
Given that several changes to the microstructure of these glasses were observed, EDS analysis was used as a supplementary tool to discover how some of these modifications took place, and if a correlation between morphology and composition could be drawn.  Figure  6 and Supplementary Information for additional maps).  To gain insight on the location of molybdenum and the mechanism of phase separation prior to irradiation these two ratios were combined to yield [Si]/[Mo] trends (see Figure  7). These trends suggest that cooling of glass melts led to the formation of phase A enriched in SiO 2  While Table 2 provides insight into the bulk properties of each phase, properties across the interface are also of interest in order to explore the mechanism of alteration following Au-irradiation. Figure 8 illustrates  Following irradiation, the magnitude of ratios and the slope of trends changes. The results in Figure 8 indicate that Mo ions move from the bulk of phases A and C to phase B. This reduces the discrete boundary between phases and the accumulation of either Ca or Mo ions along the interface. This may be related to the more significant morphological changes observed near to the A-B interface. Moreover, while [Ca] is primarily observed to decrease at the surface, relative changes to the slope of trends also imply a similar cross-phase diffusion of Ca ions, though the diffusion of Mo ions out of phase C is more significant. The significant compositional changes in and around phase C are of importance given the evident morphological alterations observed following irradiation The general mechanism of diffusion in the CM series was similar to the CB series as discussed above, with [MoO 3 ] also inducing directional changes to the diffusion of atoms. Prior to F I G U R E 6 EDS maps of CB23 irradiated with Au ions showing three immiscible phases with the interface between phases enriched in Mo and Ca. Several areas from each phase within these maps were used for the calculations in Table 2 irradiation, introducing MoO 3 to a calcium borosilicate caused Ca ions to move from phase B to phase A. As [MoO 3 ] increased to 2.5 mol%, a reversion of trends was observed with Ca ions moving from phases A and C toward phase B. This coincided with a cross sample increase of [Mo], with the most significant changes seen in phase B, and to a somewhat lesser extent in phase C. This suggests that 1 mol% MoO 3 can be incorporated in multiple phases within calcium borosilicates aided by the diffusion of Ca ions into the SiO 2 -rich phase, but as [MoO 3 ] increases it preferentially accumulates in phase (B), which encourages additional phase separation marked by Ca diffusion. When [MoO 3 ] was increased to 7 mol% a GC was formed. This resulted in Ca and Mo ions being pulled out of phase A and towards crystal clusters, for which phase B acts as a precursor.
Following Au-irradiation, CaBSi showed signs of Ca diffusion from phase B toward phases C and A, as Table 2 indicates. Once MoO 3 was introduced into the system, the primary mechanism of alteration following irradiation became a surface-to-bulk diffusion of both Ca and Mo ions, with a predominant change observed in phase A. As [MoO 3 ] increased to 2.5 mol%, Ca diffusion followed a pattern similar to that which was seen in CaBSi, with Mo diffusion following a parallel pathway. This occurred alongside a surface-to-bulk diffusion of Ca and Mo ions that was similarly observed in both CM1 and CB23.
These observations suggest two primary points of interest. The first is that as [MoO 3 ] increases to 2.5 mol% and both Ca and Mo ions accumulate in phase B, irradiation causes the cross-phase diffusion of atoms to behave in a similar manner to a glass without molybdenum. The second is that [B 2 O 3 ] appears to be the controlling factor in surface-bulk directional diffusion independent of [MoO 3 ] with the following relationship:

| Raman analyses
In order to gain additional insight on the role of network formers and network modifying ions Raman spectroscopy was utilized. The glasses in this study showed several characteristic bands attributed to the Si-O-B structure, and to the molybdenum local environment with a dependence on composition and phase (see . Across all multiphase glasses, phase A had several bands that are typical of silicate and borate-based glasses. The most easily identifiable is the R band (~ 450 cm −1 ) attributed to mixed Si-O-Si and Si-O-B bending and rocking, 59,67-69 as well as B-O-B rocking (~500 cm −1 ). 70 Additionally, there were two defect bands, D 1 and D 2 , which are assigned to four-membered and three-membered SiO 4 rings, respectively. Silicate-type structures also produced overlapping bands between 850 and 1250 cm −1 that represent Si-O stretching vibrational modes for Q n units. 30 Furthermore, there was a supplementary weak band at ~800 cm −1 assigned to O-Si-O symmetric bond stretching associated with Si motion against its surrounding oxygen cage. 29,68,71 This latter band often overlapped with a sharp peak at ~807 cm −1 attributed to the symmetric vibrations of six-membered BO 3 -triangles in boroxyl rings. 68,[71][72][73] In addition to broad bands characteristic of the borosilicate network, there were several additional sharp Raman bands attributed to the internal vibrations of (MoO 4 ) 2− in a powellite-type structure with C 4h point symmetry. They are the ν 1 (A g ) 878 cm −1 , ν 3 (B g ) 848 cm −1 , ν 3 (E g ) 795 cm −1 , ν 4 (E g ) 405 cm −1 , ν 4 (B g ) 393 cm −1 , and ν 2 (A g +B g ) 330 cm −1 vibrational modes, which represent symmetric elongation of the molybdenum tetrahedron, asymmetrical translation of double degenerate modes, and symmetric and asymmetrical bending, respectively. 74,75 There was an additional band between ~910-920 cm −1 associated with the symmetric stretching vibrational mode of (MoO 4 ) 2− units dissolved in amorphous systems. 21  and thus produce similar vibrations with these bands narrowing as the system becomes crystalline.
Following Au-irradiation, several structural modifications are observed for samples in the CM series. Primarily, the R band is observed to shift to higher wave numbers for all multi-amorphous phased samples. This shift indicates a reduction in the intertetrahedral angles between network formers, implying the formation of smaller or distorted ring structures. This occurs alongside a marginal growth of the D 1 and D 2 defect bands for samples with no or low [MoO 3 ], which further supports the formation of smaller Si-type ring structures. In CaBSi, there was also growth of Q 3 relative to Q 2 in phases A and B, indicating an increase in the general degree of silica polymerization across all phases following irradiation exposure. An opposing affect has been observed to occur in irradiated sodium borosilicate glasses, 48,59 which suggests that Ca 2+ can reduce the population of NBOs, whereas Na + promotes their formation following irradiation.
For samples with [MoO 3 ] ≥ 2.5 mol%, spectral changes were predominately associated with (MoO 4 ) 2− vibrational modes, and the boroxyl ring band. In CM2.5 most (MoO 4 ) 2− vibrational modes within phase A (see Figure 9) and phase B (see Supplementary Information for additional spectra) could not be seen following irradiation. In their place, growth of the band associated with isolated (MoO 4 ) 2− units in the amorphous network grew. This occurred alongside the growth of the D 1 defect and damping of the O-Si-O stretching band, whereas little change to the boroxyl ring band was observed. In contrast, only minor damping of the (MoO 4 ) 2− bands were observed in CM7 where CaMoO 4 crystallites were detected. In this composition the most significant change was a damping of the boroxyl ring band, suggesting radiation-induced ring cleavage. The collective results suggest that Au-irradiation can significantly alter the molybdenum environment and increase dissolution of isolated (MoO 4 ) 2− units through Si-O-B ring distortion, but only when CaMoO 4 crystallization has not occurred.
In compositions with a fixed [MoO 3 ], increasing [B 2 O 3 ] caused an intensity increase in the D 1 defect, concurrent to damping of the D 2 defect prior to irradiation (see Figure 10). This coincided with a sharp growth in the band associated with boroxyl rings, and with (MoO 4 ) 2− vibrational modes. This result suggests that increasing [B 2 O 3 ] increased the order of (MoO 4 ) 2− units toward a crystalline phase, while promoting phase separation within the borosilicate network through the formation of single-atom type rings. This effect was more pronounced in phase B, which further supports the statement that B 2 O 3 F I G U R E 1 0 Raman spectra of the Sirich phase A (bottom half) and the Ca/Morich phase B (top half) of CB7 and CB23 prior to (blue solid line) and following Auirradiation (red dashed line) [Color figure can be viewed at wileyonlinelibrary.com] preferentially separates into phase B, hence why characteristic bands of silicate-based features become less significant in this phase.
Following irradiation, the (MoO 4 ) 2− vibrational bands are observed to broaden within all phases, and within phase A to dampen significantly, as Figure 10 illustrates. This band broadening and damping occurred alongside growth of the band at ~910 cm −1 attributed to (MoO 4 ) 2− units dissolved in the amorphous network, suggesting increased solubility of molybdenum units, and the prevention of molybdate crystallization from ordered environments. While this shift occurred in both phases A and B, the amplitude of these bands suggest that molybdenum is predominantly found in phase B. Changes to the R band, D 1 and D 2 defects, and O-Si-O stretching were similar to those observed in the CM series following irradiation with an exception seen for CB23. While the boroxyl band was persistent for glasses with respect to [MoO 3 ] in the CM series, a minor damping of this band was observed for CB23 following irradiation. This occurred alongside damping of the D 1 defect. These results indicate that Au-irradiation is more easily able to cleave smaller single-atom type rings and induce the formation of larger Si-O-B ones for compositions with [SiO 2 ]/[B 2 O 3 ] < 2.8.
Although some phase-specific trends are shown in Figure  10, a large variation in structural features was actually observed by Raman spectroscopy following Au-irradiation. The spectra for CB15 in Figure 11 illustrate this point. This variance in bands can be used to gain insight into possible mechanisms of phase separation or reintegration, as these spectra were collected across phases.
Whilst general modification to the R band, D 2 defect, and boroxyl ring bands remained similar across the spectra, the proportion of (MoO 4 ) 2− vibrational modes to these bands changed. This result suggests clustering and increased order of (MoO 4 ) 2− oxyanions within some localized areas, while increased dissolution of isolated units in other areas. Therefore, it can be deduced that Au-irradiation created several mixed regions within the amorphous network that showed characteristic bands of phases A and B, or of phases B and C. The predicted mixing between phases represented by these new spectra is supported by a change to the microstructure of samples, and cross-phase ion diffusion suggested by EDS analysis in Table 2.

| DISCUSSION
It is evident from the complex microstructures observed in this study that these borosilicate glasses underwent liquidliquid phase separation during synthesis. This resulted in a unique heterogeneous glass for investigation. While a multiphase glass is unconventional for industrial applications, the chosen compositional range was used to: (a) identify factors influencing molybdenum solubility and the process of CaMoO 4 formation, and (b) study the compositional impact on immiscibility properties at a fundamental level when the materials were subjected to Au-irradiation, which replicated the ballistic collisions from α-decay on a geological timescale.

| Evolution of separated phases
Glasses are by nature non-equilibrium materials where material properties are dependent on composition and environmental factors such as temperature, pressure, or high-energy events like radiation damage. 76 These external factors can have substantial impacts on the structural components that define short-range order within the glassy matrix, especially in the case of irradiation where interactions create ballistic motion of atoms, significant heat, as well as an input of charge.
In this study, structural modifications following irradiation were most evident in phase C, where several changes to the morphology, size, and spatial distribution of droplets were observed. The mechanism of alteration appeared to be dependent on composition. At low [MoO 3 ], phase C droplets were observed to coalesce into larger particles that became spherical, similar to the changes seen in the calcium borosilicate without MoO 3 . This type of mechanism is similar to temperature-based relaxation and particle growth during cooling above T g , and can be translated to growth of the domain size in phase-separated amorphous materials. 44,56,[77][78][79] Increasing [MoO 3 ] to 2.5 mol% had an opposing effect on microstructural modifications, where separation and segregation of phase C droplets was the main mechanism of alteration. This was evident from a reduction in the size of phase C droplets that also appeared to be more evenly distributed within phase B. Similar modification to distribution and particle segregation was also observable for compositions with [B 2 O 3 ] ≤ 15 mol%. This type of mechanism is similar to the disorder created by radiation damage in organized systems like crystals. [80][81][82] A schematic summarizing these collective transformations as a function of composition can be seen in Figure 12.
The observed changes to morphology can be attributed to two factors. The first being the initial state of synthesized samples. Thermodynamically, immiscibility in glasses is controlled by the free energy of mixing, and occurs when the entropy of phase separation is lower than that of a homogeneous state. 78 This happens when the system falls below a characteristic temperature, referred to as the phase separation temperature T PS . Composition can affect this threshold, with both network formers (SiO 2 , B 2 O 2 , and Al 2 O 3 ) and network modifiers (alkali and alkaline earths) having an impact. 83 In a binary SiO 2 -B 2 O 2 system, low [B 2 O 3 ] will induce phase separation unless a high temperature of synthesis and rapid quenching techniques are used. 71 Similarly, the inclusion of an alkali or alkaline earth at low concentrations will induce the segregation of cation/borate-rich domains in a residual Si-rich matrix, with the regions of immiscibly increasing as the atomic number decreases. 55,84,85 In this set of experiments varying the melt temperature, time at temperature or quench rate did not induce homogeneity, indicating that external parameters cannot always remediate the chemical immiscibility induced by composition, or at least not within feasible laboratory ranges.
Furthermore, the inclusion of MoO 3 into these types of phase-separated glasses will further promote the separation of these two domains with molybdenum favoring the borate-rich phase in SiO 2 -B 2 O 2 -Na 2 O glasses. 44,61 This has been quantified to a ~50°C increase in T PS per 0.5 mol% MoO 3 . 12 It can be assumed that MoO 3 inclusion causes an analogous increase in T PS for the SiO 2 -B 2 O 2 -CaO system. 63 This inference is based on the observation that increasing [MoO 3 ] increased the size of phase B domains, as well as [Mo] in this phase relative to phase A. Additionally, increasing the incorporation of MoO 3 caused the formation of boroxyl rings, increased the ordering of (MoO 4 ) 2− oxyanions toward a crystalline precursor, and caused a reduction in the coordination of boron. 64,86 All of these observations point to higher T PS that would increase ΔT = T PS -T g and therefore the relaxation time for a fixed cooling rate. Similar observations were F I G U R E 1 2 Compositional-dependent mechanisms of modification for immiscible phase C following Au-irradiation. The direction of arrows represent an increasing value of concentration or particle size. Coalescence is akin to temperature-based particle growth, while segregation is related to radiation-induced disorder. [Color figure can be viewed at wileyonlinelibrary.com] seen for increasing [B 2 O 3 ], which indicates that this too may increase T PS .
In this system of embedded immiscibility, secondary phase separation also occurred. It is predicted that the de-mixing of the separated phase into phases B and C is driven by the residual silica content in this liquid phase during cooling. 85 While the mechanism of secondary phase formation is still under investigation it is clear that it is a sequential process. Primary phase separation of a melt leads to the formation of two phases, referred to as phases X and Y in this example. Phase X represents the SiO 2 residual phase, and phase Y represents an alkaline earth borosilicate. Secondary phase separation causes phase Y to be further separated into two phases: Y 1 and Y 2 . In this step, phase separation is driven by the coalescence of a SiO 2 -rich phase Y 2 within a cation-rich borate phase Y 1 . 87 Compositionally Y 2 will be different from X, though both are relatively rich in SiO 2 .
While an increase in T PS would affect primary phase separation, it is predicted that this would have subsequent effects on secondary phase formation if not by increasing the temperature of this transition, then by changing the viscosity of the various phases, which would alter the relaxation period needed for secondary phase formation and droplet conglomeration. This is supported by the larger and more spherical phase C deposits observed for higher [MoO 3 ] or [B 2 O 3 ], which implies that droplet rearrangement and relaxation are occurring faster during cooling in this composition range. As previously mentioned, increasing [MoO 3 ] or [B 2 O 3 ] is also associated with an increased T PS , which suggests a corresponding change to secondary phase formation. Furthermore, previous investigations have yielded that an increase in T PS can decrease the fictive temperature (T f ) of the residual matrix, 88 which is the point at which a liquid at metastable equilibrium transitions into a solid, and relaxation can no longer take place. This further indicates that increases in [MoO 3 ] or [B 2 O 3 ] may increase the period available for primary and secondary phase formation and coalescence.
The second factor in determining why certain changes to morphology occurred is to understand how external radiation is interacting with matter, and how this mechanism of alteration may change depending on glass structure. There are two components to consider here. The first is the structural reorganization taking place from nuclear collisions, which can alter material properties. 48,59 The second involves a thermal response from localized heating along ion tracks created by fast-moving ions used during external irradiation. This effect is often described by the thermal spike model, which translates interactions of electrons in the target material with the irradiating ion to a small cylinder of energy with a characteristic temperature of ~1000 K. 89 In real systems undergoing α-decay, these two processes can combine and result in an increase in T f through a process of ballistic disorder (inclusive of melting) and fast quenching. 60,90 This can subsequently remediate some phase separation, especially in systems with a higher T PS .

| Mechanism of radiation damage
Based on existing knowledge and results from this experiment, it hypothesized that the mechanism of alteration occurs in a three-step process: (a) the displacement of atoms through nuclear collision; (b) localized high-temperature heating akin to melting at the collision site and along the ion track; (c) rapid quenching of these structural changes with the degree of alteration dependent on the composition of each phase.
The coalescence or segregation of particles is predicted to occur based on phase-specific properties. Phases A, B, and C are compositionally different and will therefore have a different packing factor, if this term can be translated to an amorphous system. It is theorized that the varying densities within these phases will direct the ion interaction pathway, and therefore the accumulation of damage within each phase. When a high-energy ion impacts a surface it will hit some atom referred to as the primary knock-on atom (PKA). The depth that this ion will travel before it collides with another atom will depend on its energy (and therefore speed), mass, and charge. The PKA will then behave like a α-recoil and set off a billiard ball-type interaction, while the penetrating ion will continue traveling through the target material and initiate a series of additional displacement cascades interacting with target atoms at a higher frequency until it stops. As the ion travels along this pathway its electrons will also be interacting with the target material's electrons, creating a melt halo around every collision. 91 As the ion and knock-on atoms move through matter they will have a lower energy and will take the path of least resistance at a higher probability creating the most damage, in comparison to the higher energy PKA. 92 In heterogeneous systems, this means that the accumulated damage volume may vary from phase to phase characterized by varying compositions. 58,93 It also means that the interface between phases where a density variation exists will be more susceptible to ion interactions. This has actually been thought to improve the radiation tolerance of multiphase materials. 94 For systems without or with low [MoO 3 ], the allowable relaxation time correlated with ΔT = T PS -T g is shorter meaning that phases B and C will be compositionally more similar than if a longer relaxation time was available for additional SiO 2 to move out of phase B. Therefore, it is theorized that accumulated damage was more uniform within these phases, thus enabling a thermal relaxation type coalescence of phase C particles similar to that occurring during cooling. 16 For [MoO 3 ] ≥ 2.5 mol%, which is associated with an increase in T PS and thus a faster formation and relaxation of chemically different phases, a non-uniform response is expected. It is predicted that densification of phase C (Si-based droplets) and swelling of the phase B (cation-and borate-rich matrix) takes place, as has been shown to occur in silicate and ternary glasses. 48,49,95 Given that the CB series had a fixed 2.5 mol% MoO 3 concentration we would expect these glasses to behave in a similar manner, but it was observed that increasing [B 2 O 3 ] to 23 mol% hindered this process of phase C densification and phase B expansion. Here it is theorized that the three-dimensional connectivity of the borosilicate network provided improved elasticity to deformation and thus affected the thermal expansion coefficient. Based on previously published 11 B MAS NMR results, synthesized CB23 showed a significantly higher fraction of BO 3 relative to (BO 4 ) − , 64 with BO 3 groups primarily found as "ring" structures, as opposed to the higher fraction of "non-ring" seen for samples with [B 2 O 3 ] ≤ 15 mol% (see Supplementary Information). A higher concentration of these ring structures can be associated with restoring forces against deformation. 96 Testing of mechanical properties could be used as a first step to verify this theory. While performing 11 B MAS NMR was difficult on irradiated samples given the very small irradiation volume, other high-resolution techniques may be employed in future research for additional analysis and quantification of B 2 O 3 , which could not be accurately measured by EDS. This may provide greater insight and support the theories presented in this paper.

| Direction of diffusion
In this study, EDS analysis was used to indicate how phases A, B, and C differed compositionally prior to and following irradiation. This yielded information on the migration of ions and the possible integration or segregation of phases occurring during irradiation. Two general trends could be identified. One was a surface-bulk diffusion of Ca, and occasionally Mo ions. The other was a cross-phase diffusion of these ions with trends varying for the CB and CM series.
These were also the main mechanisms of modification observed following irradiation with β-particles 63 or Xe 23+ ions, 64 though variations in the direction of diffusion were observed with the type of irradiation. This is partly owing to β-irradiation remediating Si-O-B phase separation, whereas Xe-irradiation propagated de-mixing of the phases. These alterations were concurrent to different changes to the domain size of separated phases, with Au-irradiation producing the most significant modification. While some structural changes are attributed to composition and type of irradiation, the primary mechanism of cationic depletion at the surface is associate with an electric field gradient created by ions bombarding the surface. 97,98 This side-effect to using external irradiation made crossphase diffusion more difficult to analyze, but most interestingly the direction of this surface-bulk diffusion showed a dependency on composition.  64 This initial structure plays a role in the concentration of network modifying cations and the degree of glass network connectivity.
The second factor to consider is that irradiation has previously been observed to cause a reduction in the coordination of boron that correspondingly increases the NBO population. 48,59,[100][101][102] It is subsequently thought that in compositions with an initially higher [(BO 4 ) − ] and with a higher fraction of non-ring BO 3  This has similarly been observed to occur in sodium borosilicates under irradiation. 97 The diffusion of ions can therefore be primarily correlated to the strucutre of the borosilicate network, which is itself affected by synthesis conditions, composition, and external treatments, as opposed to the secondary factor of the ion properties.
Reducing the [SiO 2 ]/[B 2 O 3 ] also enabled the cleavage of smaller single atom-type rings and induced the formation of larger Si-O-B ones. This outlines the importance of B 2 O 3 in controlling network connectivity and by extension the role of cations, which have a significant impact on the radiation response and mobility of ions. It may also affect the ability for certain phases to absorb incoming ionization, and therefore elicit precipitation versus ion depletion.
Independent of the type of diffusion, this process would first require the formation of ion migration channels. This can be achieved through accumulated ring cleavage between network formers, which creates NBOs. This type of bond cleavage can also modify the role of cations from network forming to network modifying. Together this can increase the diffusivity of ions, as has been previously observed to occur in irradiated glasses. 31,97 While irradiation was generally observed to increase disorder within the molybdenum environment, thus enabling the increased integration of (MoO 4 ) 2− , the results in Figure  11 suggest that in pockets the opposite was true and these moieties clustered in depolymerized regions of the glass. The varied results suggest that radiation-induced diffusion of ions could contribute to both phase separation and re-integration with a dependence on the initial properties of each phase. These properties also control the direction of ion diffusion between phases A/C↔B.

| CONCLUSIONS
Several simplified calcium borosilicate glasses with varying [MoO 3 ] and [B 2 O 3 ] were synthesized and irradiated with 7 MeV Au 3+ ions used to simulate nuclear collisions resulting from the α-decay process. A dose of 3 × 10 14 ions/cm 2 was achieved creating ~1 dpa of damage to a depth of 1.6 μm. Samples were then analyzed using BSE imaging, EDS analysis, and Raman spectroscopy to identify factors affecting the chemistry of phase separation within heterogeneous glasses, and the solubility limit of problematic molybdenum groups. Radiation was observed to create several changes to the morphology, spatial distribution, and size of separated phases within these three-phased (A, B, and C) heterogeneous glasses. At low [MoO 3 ], phase C droplets were observed to coalesce into larger particles that became spherical emulating a temperature-based particle growth, whereas increasing [MoO 3 ] to 2.5 mol% was observed to encourage droplet separation and segregation forming more evenly distributed particles. This latter type of segregation and distribution was also observable for compositions with [B 2 O 3 ] ≤ 15 mol% and is predicted to result from phase-specific properties created related to T PS , and therefore the degree of heterogeneity. Increasing [B 2 O 3 ] appeared to partially hinder this process owing to the initial defect structure, which subsequently impacted the direction of ion diffusion. The process of surface-bulk and cross phase diffusion were enabled through increased disorder within the borosilicate network that increased the dissolution of (MoO 4 ) 2− units in most areas based on damping of (MoO 4 ) 2− Raman vibrational modes within a powellite-type structure, combined with the growth of a band ~910 cm −1 associated with molybdate moieties dissolved into an amorphous framework. Conclusions, therefore, report several effects of [MoO 3 ] and [B 2 O 3 ] in directing structural modifications and altering the size of immiscible regions in multiphase calcium borosilicate glasses following external Au-irradiation simulating α-decay on geological timescales.