Thermal and Crystallisation Kinetics of Yttrium-Doped Phosphate-Based Glasses

Yttrium doped glasses have been utilised for biomedical applications such as radiotherapy especially for liver cancer treatment. In this paper, the crystallisation behaviour of phosphate-based glasses doped with yttrium (in the system 45P 2 O 5 -(30-x) Na 2 O-25CaO-xY 2 O 3 – where x= 0, 1, 3 and 5) have been investigated via Differential Scanning Calorimetry (DSC) using non-isothermal technique at different heating rates (5°C, 10°C, 15°C and 20°C/min). The glass compositions were characterised via EDX, XRD, Density and Molar volume analysis. The Moynihan and Kissinger methods were used for the determination of glass transition activation energy (E g ) which decreased from 192 kJ/mol to 118 kJ/mol (Moynihan) and 183 kJ/mol to 113 kJ/mol (Kissinger) with increasing yttrium oxide content. Incorporation of 0 to 5 mol% Y 2 O 3 revealed an approximate decrease of 71 kJ/mol (Ozawa and Augis) for onset crystallisation (E x ) and 26 kJ/mol (Kissinger) for crystallisation peak activation energies (E c ). Avrami index (n) value analysed via Matusita-Sakka equation suggested a one-dimensional crystal growth for the glasses investigated. SEM analysis explored the crystalline morphologies and revealed one-dimensional needle-like crystals. Overall, it was found that these glass formulations remained amorphous with up to 5 mol% Y 2 O 3 addition with further increases in Y 2 O 3 content resulting in significant crystallisation.


Introduction:
Phosphate-based glasses have been widely investigated for biomedical applications and can provide highly useful properties such as low glass transition and melting temperatures, high thermal expansion coefficients, biocompatibility, high refractive indices and one of their most unique offerings is their controlled dissolution properties. (1). The structural, chemical and physical durability of phosphate glasses has been shown to be improved by the inclusion of several oxide additions such as ZnO (2,3), Fe2O3 (4,5), Y2O3 (6,7), Al2O3 (8,9), TiO2 (10,11) and Bi2O3 (12,13).
Amongst these, yttrium oxide containing glasses are an area of interest due to their characteristic structural and physical properties such as high refractive index, high thermal expansion coefficient, density and excellent infrared transmission (14). Moreover, it has also demonstrated highly useful application in radiotherapy (15)(16)(17) due to the relatively short half-life and beta emission properties of the 90 Y isotope (18)(19)(20).
Y2O3 plays a vital role in improving the properties of glasses. Within glass systems, yttrium oxide can act as a network modifier or as a network former depending on glass composition and the amount of Y2O3 within the glass (14,21,22). Singh et al. investigated the 55 SiO2-30 B2O3-x Li2O-(15-x) Y2O3 glass system and reported that Y2O3 changed its role from glass former to network modifier as its concentration exceeded 5 mol% (14). Furthermore, incorporation of Y2O3 into the zinc borosilicate glass system was reported to increase glass stability by 20°C with increasing yttria content from 0 to 3 mol% (23). Mahdy et al. investigated yttrium aluminosilicate glasses and stated that the glass stability increased due to replacing Al2O3 with Y2O3 resulting in greater strength of the cross-links between the Y 3+ cation and oxygen atoms, leading to an increase in the structural rigidity of the glasses (22). This increase in stability can be related to the crystallisation kinetics and activation energy of the glass systems (24) which also correlates with the crystallisation behaviour of the glasses.
The higher the activation energy, the less prone the glass will be to crystallise (25).
Wang et al. studied yttrium zinc borosilicate glasses and stated that addition of Y2O3 enhanced the glass thermal stability and addition of more than 8 mol% of Y2O3 revealed that the glass began to crystallise due to tendency of supersaturation in the glassy state (30). Vishal et al.
studied the crystallisation behaviour of a yttrium and lanthanum calcium borosilicate glass to explore the suitability of this glass as a sealant and they found that this glass was a good candidate as compared to yttrium calcium borosilicate glass due to higher thermal expansion coefficient, low fragility index value and viscosity profile (26). Studies on crystallisation kinetics of lithium (31), copper (32), calcium (33) and aluminium (9) phosphate glasses have been reported. Limited studies have been reported on the crystallisation behaviour of yttrium phosphate glasses. Petra et al. studied on crystallisation kinetics of erbium doped yttrium phosphate glasses based on particle size of the powder and bulk sample (34).
In order to explore crystallisation kinetics and thermal stability of glasses, differential thermal analysis (DTA) and differential scanning calorimetry (DSC) are popular methods for studying the kinetics of non-isothermal data (35,36). These methods are relatively straight-forward to carry out, requiring only small sample volumes and possessing flexibility in the ability to vary the heating rates. Between these two techniques, DSC is widely used due to its simplicity and sensitivity (37,38). Isothermal and non-isothermal methods are the two basic methods in DSC analysis. For the isothermal method, the sample is heated up to a temperature above glass transition (Tg) and the heat evolution during the crystallisation process is recorded as a function of time where the temperature is kept constant. For non-isothermal analysis, the sample is heated at a fixed rate and the heat evolved is recorded as a function of temperature and time (39) and is more widely used due to its simplicity and rapid method (40,41).
In the present investigation, the glass transition temperature and crystallisation kinetics were studied as a function of the Y2O3 content in phosphate glasses using non-isothermal DSC analysis. The Moynihan (42), Kissinger (43), Ozawa (44), Augis-Bennette (45) and Matusita-Sakka (46) equations were used for the determination of crystallisation kinetics and activation energies for glass transition, onset crystallisation and crystallisation temperature.

Glass synthesis:
Four different glass compositions were prepared using sodium dihydrogen phosphate (NaH2PO4), calcium hydrogen phosphate (CaHPO4), yttrium oxide (Y2O3) and phosphorous pentoxide (P2O5) (Sigma Aldrich, UK) as starting materials. The precursors were mixed together and transferred to a 10%Rh/Pt crucible (Birmingham Metal Company, UK) which was placed in a furnace preheated to 350°C for half an hour and 500°C for an hour depending on the glass compositions for the removal of H2O. The mixtures were then heated at 1150°C for the yttrium free glass and 1300°C for the yttrium containing glasses, for 2 hours as shown in Table 1. The resultant melts were poured onto a steel plate and left to cool to room temperature. The glass was then crushed using a planetary ball mill (Retsch PM100) and sieved to obtain a glass particle size range of between 25-45 µm.

Powder X-ray diffraction analysis:
The as-quenched glass was used for XRD analysis to investigate the amorphous nature of the glass compositions produced. The data was collected on a Bruker AXS-D8 Advance powder diffractometer in flat plate geometry using Ni-filtered Cu-Kα radiation (λ=0.15418 nm), operated at 40kV and 35mA. The angular range 2 for each scan was from 10° to 70° with a step size of 0.1° and a step time of 5s. The phases were identified using the EVA software via Crystallographic Search-Match icon (DIFFRACplus Suite,Bruker-AXS) and the International Centre for Diffraction Data (2005).

Energy dispersive X-ray analysis:
Portions of the glass samples from each composition were polished using SiC paper and diamond cloths with industrial methylated spirit (IMS) as eluent. The samples were then dried and cleaned by air dry spray before being carbon coated. Energy dispersive x-ray (EDX) analysis was performed on an EDAX model DX 4 using ZAF quantitative analysis. The accelerating voltage was maintained at 15 kV and the system resolution was 60 eV with a live time of 120 seconds. Jadeite (for Na), gallium phosphide (for P), wollastonite (for Ca) and yttrium (for Y) were used as a standard for analysis. The working distance was 10 mm and the EDX spectrometer was connected to the Philips XL30 SEM. An average of five separate areas were analysed for the final composition of the materials.

Helium pycnometry:
Density measurements were conducted using a Micromeritics AccuPyc 1330 helium pycnometer (Norcross, GA, USA). 'Archimedes' principle of fluid displacement was employed to determine the volume of the solid objects. Calibration of the equipment was conducted using a standard calibration ball (3.18551 cm 3 ) with errors of ±0·05%. Approximately 3.5 g of the samples were used for measurement and the process was repeated three times. For each formulation the molar volume was calculated using the equation 1: where xi is the molar fraction of the ith component, ρ is the density, and Mi is the molecular weight of the ith component.

Thermal analysis:
Approximately 10 mg samples of the glass powders (size range of 25 and 45 μm) were tested to determine the thermal properties such as glass transition (Tg), onset of crystallisation (Tx), crystallisation (Tc) and melting (Tm) temperature using a simultaneous thermal analysis (TA Instruments SDT Q600, USA). Glass particles were placed into a platinum pan and heated from room temperature to 1100°C at different heating rates (i.e. 5, 10, 15 and 20°C/min) under argon gas flow (50 ml/min). A blank run was carried out to determine the baseline which was then subtracted from the plot obtained using TA universal analysis 2000 software. The thermal stability of the glasses was measured in terms of the processing window by taking the temperature difference between the glass transition temperature (Tg) and the onset of crystallisation temperature (Tx) as shown in the following: Processing window = Tx -Tg (2)

SEM analysis:
For imaging analysis, a JEOL 7100F Field-Emission Gun Scanning Electron Microscope (FEG-SEM) and scanning electron microscope (SEM, Phillips FEI XL30, USA) utilising a 15 kV beam set at a working distance of 10 mm, were used to investigate the glass formulations produced and the samples which crystallised. Prior to acquisition, the samples were mounted in an epoxy resin block, which were ground using varying grits (P240-P1200), followed by polishing on 6 and 1 μm diamond polishing pads respectively. The polished blocks were then eroded in 4 % (mass concentration) HF for 20 s each. The etched samples were subsequently cleaned with deionised water and air dried. A conductive carbon coat of ca. 10 nm was applied before SEM analysis.

EDX analysis:
EDX analysis was performed to confirm the compositions of the different glass formulations prepared in comparison with expected formulations (see Table 2). All the compositions were found to be close to the expected compositions (within a 2 % error margin). Figure 1 represents the 8 mol% and 10 mol% Y2O3 containing glass sample where 8 mol% glass became crystalline and the glass did not melt for 10 mol% glass sample.

XRD analysis:
The X-ray diffraction spectra of the different glass compositions are presented in Figure 2.
Absence of any sharp crystalline peaks suggested that the glasses as prepared were amorphous. A single broad halo shaped peak was observed between 20° to 40° diffraction angle for Y0, Y1, Y3 and Y5 formulations. Tetragonal yttrium phosphate (YPO4) crystal peaks were observed for Y8 glass formulation.

Density and Molar volume analysis:
The density and molar volume for the multicomponent glasses are presented in Figure 3. The density of glass particles increased from 2.56 to 2.74 g/cm 3 with the incorporation of 0 to 5 mol% Y2O3 into the glass system. In contrast, the change observed in molar volume was from 37.65 cm 3 mol -1 to 38.1 cm 3 mol -1 for the same glasses across the compositional variation. Figure 4 shows the DSC scans obtained for 0 to 5 mol% yttrium containing phosphate glass system at different heating rates (i.e. 5°C/min -1 , 10°C/min -1 , 15°C/min -1 and 20°C/min -1 ). The DSC traces exhibited clear endothermic and exothermic peaks at glass transition (Tg) and crystallisation (Tc) temperature, respectively. The onset crystallisation temperature (Tx) has been defined as temperature corresponding to the intersection of the two linear portions of the transition between glass transition and crystallisation temperature. The values of Tg, TX, Tc and processing window (i.e. Tx-Tg) at different heating rates are summarised in Table 3. As can be seen in Table 3, the Tg at different heating rates was similar for the yttrium free phosphate glass (355°-361°C) whereas the Tg at different heating rates increased from 368°-376°C, 386°-399°C and 415°-429°C for the glasses investigated with increasing Y2O3 content.

Thermal analysis:
The variation of shifts in Tx for the varying glass formulations at different heating rates are presented in Figure 5. The Tc at different heating rates increased from 471°C to 537°C for 5°C, 480°C to 552°C for 10°C, 485°C to 564°C for 15°C and 492°C to 571°C for 20°C, with increasing (0 to 5 mol%) Y2O3 content, respectively as shown in Figure 4. The values of Tx and Tc are listed in Table 3.
The processing window (Tx-Tg) was also evaluated for the different compositions investigated which is an indication of their thermal stability (47). From Figure 6, it was seen that the processing window values increased as the Y2O3 concentration increased to 3 mol% and further addition of Y2O3 to 5 mol% showed a slight reduction in the thermal stability. It was also observed that the values of processing window were seen to increase with increase of heating rates for each formulation.
To validate the results, the glass formulations were heat-treated at two different temperature regimes; a) between 24-50°C above Tg to avoid crystallisation and explore any physical changes in their properties and b) between Tx and Tc to deliberately crystallise the samples and observe the crystal morphology formed.
"From Figures 7 (a-d), it was observed that needle shaped crystals had nucleated near to the glass surface. However, fewer needle shaped crystals were observed on the glass surface for the 5 mol % Y2O3 samples as seen in Fig. 7 (d)." A heat treatment process (between Tx and Tc) at a heating rate of 10°C/min was applied to the different glass formulations investigated to explore the growth and morphology of the crystals observed. However, it was also observed that the needle like crystalline morphologies formed decreased in width with the addition of yttrium oxide.

Activation energies:
The activation energies for the Tg, Tx, Tc and overall crystallisation (E) region of the glasses were determined by utilising the equations outlined in supplementary section S1.

Activation energy of glass transition temperature:
The glass transition activation energies (Eg) obtained using equation 3 in supplementary section S1 are represented in Figure 9 and highlight the plots of lnα versus 1000/Tg. The Eg values obtained from the slope of the straight lines are also highlighted in Figure 9. As can be seen, the Eg obtained using the Moynihan equation decreased from 191.82 to 118.12 KJ/mol with increasing Y2O3 from 0 to 5 mol%.
The Eg can also be determined utilising the Kissinger equation (see equation 4 in S1). A plot of ln (α/Tg 2 ) versus (1000/Tg) is depicted in Figure 9. The activation energies here were also

Onset crystallisation temperature activation energy:
The values of the onset crystallisation activation energy (Ex) were obtained using the Ozawa equation (see equation 5) based on the slope of lnα vs 1000/Tx as shown in Figure 10. The values of Ex decreased from 189.74 to 118.27 KJ/mol with the incorporation of Y2O3 into the glass system.
The onset of crystallisation activation energies using Augis and Bennett method was also determined (based on equation 6 in S1). The Ex of the glasses estimated from plots of ln (α/Tx) versus 1000/Tx are shown in Figure 10. The values of Ex were seen to decrease from 184.26 to 114.88KJ/mol with the incorporation of Y2O3 in the glass system.

Peak crystallisation activation energy:
The activation energy of crystallisation peak temperature (Ec) was also explored based on the Kissinger equation (see equation 7 in S1). The plot of ln (α/Tc 2 ) versus 1000/Tc for the samples explored are shown in Figure 11. The value of Ec was calculated from the slope of each curve.
It was clear from Figure 10 that the activation energy of crystallisation (Ec) decreased from 120.34 to 94.27 KJ/mol with increasing Y2O3 content.

Overall crystallisation activation energy:
The overall non-isothermal crystallisation activation energy (E) was calculated using the Matusita-Sakka equation, which represents the volume fraction of crystals (X) precipitating in a glass heated at a constant rate (α). From equation 8, the nucleation parameter (n) was determined by plotting ln[-ln(1-X)] versus lnα for three selected temperatures (within the crystallisation peaks of the DSC traces obtained at each of the heating rates explored) as shown in Figure 12. The slope of the straight lines provided the value of nucleation and growth parameters (n) and the corresponding values are listed in Table 4. The nucleation parameters or Avrami exponent (n) were almost similar for the glass formulations tested. The dimensionality of crystal growth (m) depends on the Avrami exponent (n) and usually the dimensionality of crystal growth m is equal to n-1 (23). As such, a one-dimensional crystal growth was suggested based on the Avrami exponent values obtained for the glasses investigated (see Table 4).
With the order parameter n, m and plots of ln [-ln(1-X)] versus 1000/T at different heating rates, the overall crystallisation activation energy (E) of the glass formulations was estimated from Equation 9. These plots are presented in Figure 13. Overall the crystallisation activation energy decreased from 253.46 to 205.52 KJ/mol with the incorporation of Y2O3 in the glass system (see Table 4).

Discussion:
Yttrium containing phosphate-based glasses (PBGs) could be promising materials for biomedical applications such as radioembolisation therapy (48). However, from the glass formulations investigated, compositions above 5 mol% yttrium oxide content showed strong tendency towards crystallisation. The crystallisation kinetics and activation energy can provide a better understanding about the glass formability and thermal stability of glasses. In this study, the effect of Y2O3 incorporation into PBGs (upto 5 mol%) on crystallisation kinetics was investigated. Glasses containing more than 5 mol% yttrium oxide content (in this series) were not considered for analysis as these formulations rapidly crystallised (see Figure 1) and confirmed via XRD as shown in Figure 2.
Compositional analysis was confirmed via EDX and revealed that the respective oxide contents for the glasses investigated were within 2 mol% of their predicted values (see Table   2). The reduction of small amount of P2O5 content was observed and this was attributed to the effect of its volatilisation (49).
In general, for PBGs the density and molar volume are usually inversely related (50). However, in the present study for the glasses investigated, the density and molar volume increased with increasing yttrium oxide from 0-5 mol% (see Figure 3). Shaharuddin et al. reported that the density of phosphate glass was dependent on the atomic weight of the components added as well as compactness of the glass structure (51). The increase in density was attributed to the replacement of lower molar mass (61.97 g/mol) and density (2.27 g/cm 3 ) of Na2O with higher molar mass (225.81 g/mol) and density (5.01 g/cm 3 ) of Y2O3 (11,14). Ebrahim et al. also found an increase in density (from 2.78 -3.01 g/cm 3 ) for calcium magnesium aluminosilicate glass due to the addition of Y2O3 (from 0 to 3 mol%) in place of Al2O3 (22). Structural NMR analysis on phosphate based glasses were studied by different authors in the literature (52)(53)(54) and reported that Q 1 and Q 2 structures were found for compositions with fixed P2O5 at 45 mol%. The thermal profiles of PBG are strongly dependent on their structure and composition (50).
A gradual increase in glass transition temperature (Tg) for glass particles were observed with increasing yttrium oxide (from 0 to 5 mol%) content and with different heating rates (5°, 10°, 15° and 20°C / min) as shown in Figure 4 (57,58).
In this study, the Tg values were also found to increase with increasing heating rates (see Table 3). This was attributed due to the decrease of structural relaxation which subsequently increase the glass transition temperature with increasing heating rates. The onset crystallisation (TX) and crystallisation peak (Tc) temperatures of the glass particles were also seen to increase with incorporation of yttrium oxide at different heating rates (see Table 3).  (64). Similar results were also found by Clupper and Hench (65) and Bretcanu et al. (66).
The thermal stability of the glass particles in terms of their processing window increased with the addition of yttrium oxide up to 3 mol% (see Table 3). This was suggested to be due to the increasing Y2O3 content from 0 to 5 mol% (see Figure 9). Eg is directly related with the relaxation energy of glass structure. The relaxation energy is the amount of energy needed to overcome the barrier between different quasi-stable regions during the preparation of glass.
Mitang et al. studied yttrium oxide containing silicate glass and reported that Eg was linked with to the structural relaxation energy and as such, Eg was inversely proportional to the stability of the glass structure (23). In addition, they stated that Y2O3 acts as a network modifier and its introduction into the glass structural network creates non-bridging oxygen's (NBOs), which can increase the disorder and entropy of the glass structure. Three types of activation energy can be involved in the crystallisation process, including activation energy for onset of crystallisation (EX), for peak crystallisation (Ec) and for the overall crystallization (E KJ/mol) with increasing Y2O3 content via Kissinger method (see Figure 11). The reduction of Ex or Ec could be attributed to an increase in the number of non-bridging oxygens by the addition of Y2O3 acting as a network modifier (7,73). Therefore, the movement of atoms in glass structure was easier for the formation of crystal nucleation and growth. Zarabad et al.  (40). In our study, the nucleation or Avrami parameter (n) was found to be approximately 2 for all the glass formulations investigated (see Table 4). This suggested that one-dimensional crystal growth was occurring for the glasses investigated, which could be explained by the small addition of yttrium oxide in the glass system (78). SEM analysis confirmed that the crystalline morphologies observed resembled one-dimensional structures in the form of needle like arrays, for all glass formulations investigated (as shown in Figure 8 (a-d). Mitang et al. studied yttrium doped zinc borosilicate glasses and found that one-dimensional crystal growth formed from the surface to the inside of the glass structure (23). From Figure 12d, deviations from the straight lines for Y5 glass formulation was observed at high temperature (550°C, 554°C and 558°C) which could be attributed to the saturation of nucleation sites during the latter stages of the crystallisation process (79). In addition, restriction of crystal growth by small size particles may also be responsible for the deviations observed at high temperature (80). The overall crystallisation activation energy (E) also decreased from 253 to 205 KJ/mol with increasing yttrium oxide content. Similar trends were also observed for other Tg, Tx and TC activation energy values. The decrease of overall activation energy could also be attributed to the network modifying effect of Y2O3 forming NBOs.
In general, the Eg was higher than Ec due to the molecular motion around the Tg (81). A similar behaviour was noted for lithium ions in bismuthate glasses (82). In this study higher Ex value was found compared to Ec (see Figures 10 and 11) and this was attributed to the effect of bulk nucleation and crystal growth processes. Huan et al. also found higher Ex value compared to Ec for amorphous Zr-Cu-Ni alloys and suggested that the nucleation process of amorphous Zr70Cu20Ni10 alloy was more difficult than the growth process (41).
In summary, the incorporation of Y2O3 into the glass system investigated decreased Eg, EX and EC and increased their thermal stability. Differences of activation energy values using Moynihan, Ozawa, Augis and Bennette, Kissinger and Matusita-Sakka plots were observed and attributed to the different temperatures involved in the various equations. The reduction in activation energy with incorporation of Y2O3 into the glass system implied that Y2O3 acted as a nucleating agent promoting crystallisation effects. On the basis of activation energy values for this glass family series (with fixed 45 mol% P2O5), it could be suggested that yttrium containing PBGs with more than 5 mol% Y2O3 would crystallise as the activation energy reduced to lower than 200 KJ/mol.

Conclusions:
In this study, crystallisation kinetic studies were carried out for 45P2O5-(30-x) Na2O-25CaO -xY2O3 glass system using non-isothermal DSC measurements, which indicated that the glass transition and crystallisation temperatures were influenced by the heating rates. The activation energy of Tg, Tx and Tc was determined by five different methods. The Tg, TX and TC increased with the addition of Y2O3 into glass system while the Eg, Ex, Ec and E decreased. The Avrami parameter (n = 2), revealed the crystallisation kinetics to be controlled by bulk nucleation with one-dimensional growth predicted and observed via SEM analysis. It is suggested that Y2O3 acted as a nucleating agent in this glass family series and glasses with yttrium oxide content above 5 mol% would crystallise. The overall activation energy data also revealed that values close to or lower than 200 KJ/mol would very likely results in crystallisation of phosphate-based glasses in this family series. :   Table 1: Glass codes, drying, melting and casting temperature used throughout the study

Y8
Glass became crystalline during cooling (see Figure 2)