The effect of MgO/TiO 2 on structural and crystallisation behaviour of near invert phosphatebased glasses

Varying formulations in the glass system of 40P 2 O 5 -(24-x)MgO-(16+x)CaO-(20-y)Na 2 O-yTiO 2 (where 0≤x≤22 and y=0 or 1) were prepared via melt-quenching. The structure of the glasses was confirmed by X-ray diffraction (XRD), Fourier transform infrared (FTIR), micro Raman and solid-state nuclear magnetic resonance (NMR) spectroscopies. The thermal properties and the activation energy of crystallisation (E c ) were measured using thermal analysis and the Kissinger equation, respectively. The glass forming ability of the formulations investigated was seen to decrease with reducing MgO content down to 8 mol% and the glass stability region also decreased from 106 o C to 90 o C with decreasing MgO content. The activation energy of crystallisation (E c ) values also decreased from 248 (for 24 mol% MgO glass) to 229 kJ/mol (for the 8 mol% MgO content) with the replacement of MgO by CaO for glasses with no TiO 2 . The formulations containing less than 8 mol% MgO without TiO 2 showed a strong tendency towards crystallisation. However, the addition of 1 mol% TiO 2 in place of Na 2 O for these glasses with less than 8 mol% MgO content, inhibited their crystallisation tendency. Glasses containing 8 mol% MgO with 1 mol% TiO 2 revealed a 12 o C higher glass transition temperature, a 14 o C increase in glass stability against crystallisation and a 38 kJ/mol increase in E c in comparison to their non TiO 2 containing counterpart. NMR spectroscopy revealed that all of the formulations contained almost equal percentages of Q 1 and Q 2 species. However, FTIR and Raman spectroscopies showed that the local structure of the glasses had been altered with addition of 1 mol% TiO 2 , which acted as a network modifier, impeding crystallisation by increasing the cross-linking between phosphate chains and consequently leading to increased E c as well as their glass forming ability.


Introduction
Phosphate-based glasses (PBGs) have attracted a lot of interest in the field of biomaterials and tissue engineering due to their controllable degradation profiles and chemical similarity with the inorganic component of natural bone [1][2][3] . PBGs with various modifying oxide such as CuO 4 , ZnO 5 , Ag2O 6 , Fe2O3 7 , TiO2 8 , SrO 9 , have been extensively investigated to adjust for biomedical and tissue engineering applications 10 . For example, the addition of CaO has been reported to improve the bioactivity of these glasses and to enhance haemostatic activity [11][12][13][14] . However, the glass structure could be disrupted due to the addition or replacement of modifying oxide and consequently, the glass forming ability could be decreased 15 . 3 The structure of PBGs is composed of PO4 3tetrahedral units. The tetrahedra can be described in terms of Q n terminology, where n represents the number of bridging oxygens (BOs) per PO4 3tetrahedron. In vitreous P2O5 each tetrahedral unit is connected with three others via bridging oxygen and the fourth oxygen of each tetrahedral unit is known as a non-bridging oxygen (NBOs) 16 . Phosphate tetrahedra with three bridging oxygens (BOs) are referred to as Q 3 species where the fraction O/P = 2.5. Addition of modifying oxides, such as CaO, disrupts the phosphate network and increases the O/P ratio. As the O/P ratio increases from 2.5 to 4 (2.5, 3, 3.5, 4) the phosphate structural group passes from Q 3 to Q 0 (Q 3 →Q 2 →Q 1 →Q 0 ), where Q 0 represents isolated PO4 3units with four NBOs 17 . PBGs with smaller phosphate units can crystallise easily as compared to glasses with longer phosphate chains as the entangled chains of the long chain phosphate glasses increase the viscosity of melt and impede crystallisation 18 .
However, vitreous P2O5 (i.e. with completely Q 3 structural unit) is chemically unstable, therefore modifying oxides have been added to make the PBG glass stable via the formation of P-O-M bonds (where M is a metal cation) 19 . The chemical durability and glass-forming ability of phosphate-based glasses has been reported to be significantly improved with the addition of TiO2 20,21 . Titanium can present in the glass as Ti 4+ or Ti 3+ , with the oxidation state of Ti in the glass dependent on the melting conditions (oxidizing or reducing) and the total amount of titanium present in the glass 22 . Ti 4+ has also been reported to act as a glass former, whilst Ti 3+ serves as a glass modifier 22 . In addition, titanium containing glasses have shown good cell viability, attachment and proliferation 8,[23][24][25][26] . Abou Neel et al. studied the effect of TiO2 on the cytocompatibility of P50Ca30Na20Ti0 glasses and reported that the addition of 5 to 15 mol% TiO2 in place of Na2O was effective in increasing cell viability 8 . It has also been 4 reported that TiO2 in phosphate glasses could help to induce calcium phosphate surface nucleation and thus improve the bioactivity of glasses 27 . However, TiO2 is also act as a nucleating agent and favors devitrification at local level for the silicate glasses [1][2][3][4].
They found that the more ordered local structure of the glasses with increasing TiO2 content from 1.25 to 5 wt% [1].
Crystallisation kinetics is an important topic for glass formation and glass-ceramic synthesis 28 . The glass forming ability of a composition can be assessed by its reluctance to undergo crystallisation 28 . The crystallisation process is usually described by the activation energy of crystallisation (Ec) and the Johnson-Mehl-Avrami (JMA) exponent (n). The Ec gives an indication of the temperature dependence of the crystallisation process, whereas n gives an idea on the crystal growth dimensionality 29 . The activation energy of crystallisation can be calculated using the Kissinger equation 30 , amongst others. The JMA exponent can be calculated using the Augis and Bennett equation 31 . It is also worth noting that the crystallisation parameters can often be influenced by the size of glass particles 32,33 . The resistance of a glass against crystallisation can also be quantified in terms of the glass stability, which is the temperature difference between the glass transition (Tg) and the onset of crystallisation (Tx) 34 .
The aim of this work was to investigate the effect of decreasing MgO content with CaO in the quaternary glass formulation of 40P2O5-(24-x)MgO-(16+x)CaO-20Na2O. However, reducing the MgO content to lower than 8mol% revealed that the glasses crystallised. As such, a further series of glasses were then evaluated incorporating 1 mol% TiO2 (in the series 40P2O5-  x)MgO-(32+x)CaO-19Na2O-TiO2), to investigate if the crystallisation could be arrested.

Glass Preparation
Varying formulations in the glass system 40P2O5-(24-x)MgO-(16+x)CaO-(20-y)Na2O-yTiO2 (where 0≤x≤22 and y=0 or 1) were prepared using sodium dihydrogen phosphate (NaH2PO4), calcium hydrogen phosphate (CaHPO4), magnesium hydrogen phosphate trihydrate (MgHPO4.3H2O), phosphorous pentoxide (P2O5) and titanium dioxide (TiO2) as starting materials (Sigma Aldrich, UK). The required amounts of precursors were weighed, mixed and transferred to a platinum rhodium alloy crucible (Birmingham Metal Company, U.K.) which was then placed into a furnace at 350 o C for 0.5 hours to remove H2O. The mixture was then transferred to another furnace pre-heated to between 1150-1200 o C for 1.5-2 hours depending on the glass composition, as shown in Table 1. The resulting molten glass was poured onto a steel plate and left to cool to room temperature.

X-ray diffraction analysis
X-ray diffraction studies were conducted to confirm the amorphous nature of each glass formulation using a Bruker D8 Advanced diffractometer. The instrument was operated at room temperature and ambient atmosphere with Ni-filtered CuKα radiation (λ=0.15418 nm), generated at 40 kV and 35 mA. Scans were performed with a 5˚ glancing angle, a step size of 0.04° and a step time of 8 seconds over an angular range 2θ from 15° to 50°. 6

Density measurement
The density of the glasses was determined in accordance with the standard ASTMD-792 using the Archimedes method. The measurement was carried out at room temperature using industrial methylated spirit (IMS) as the immersion fluid. Bubble free glass rods (9×20 mm) were used for the experiment, which was repeated three times for each glass formulation.
The density of glasses was calculated using equation 1: Where A and B are the weight of the glass rods in air and in IMS respectively, and ρo is the density of IMS at the given temperature.
The compactness of a glass structure can be measured by the oxygen density of that glass [35][36][37] .The oxygen density considers the effect of oxygen and neglects the effect of other components in the glass 35

Thermal analysis
The thermal properties of the phosphate-based glasses (PBGs), specifically the glass transition (Tg) (measured at the midpoint), onset of crystallisation (Tx), crystallisation peak (Tc), melting 7 peak (Tm) temperatures and glass stability against crystallisation, were characterised using a simultaneous thermal analysis instrument (SDT, TA Instruments SDT Q600, USA).
Approximately 30 mg of glass powder (particle size range between 45-100 µm were used which were obtained by crushing the glass using Ball mill and followed by sieving through stainless still sieves) was placed into a platinum pan and then heated from room temperature to 1100 o C at different heating rates (i.e. 10, 15 and 20 o C min -1 ) under 50 mL min -1 of nitrogen gas flow. An empty pan was also analysed in order to determine the baseline which was then subtracted from the thermal traces using TA Universal Analysis 2000 software. Triplicate was used for this measurements.

Activation energy of crystallisation
The activation energy for crystallisation (Ec) associated with the crystallisation peak of the glasses was calculated using the Kissinger equation 4 30 .
Where β is the heating rate, Tc is the crystallisation peak temperature measured at the different heating rates (i.e. 10, 15 and 20 o C min -1 ) and R is the gas constant. Where n is the JMA exponent, Tc is the crystallisation peak temperature, ΔTFWHM is the full width at half maximum of the crystallisation peak, Ec is the activation energy of crystallisation and R is the gas constant. 8

Fourier transform infrared (FTIR) spectroscopic analysis
Infrared spectroscopy was performed using a Bruker Tensor-27 spectrometer (Germany). All spectra were analysed using Opus TM software version 5.5. The glass powders (particle size range 45-100 µm) were scanned in absorbance mode in the region of 4000 to 550 cm -1 (wave numbers) and the spectra were collected with a resolution of 4 cm -1 by averaging 32 scans using a standard pike attenuated total reflectance (ATR) cell (Pike technology, UK).

Micro Raman spectroscopic analysis
Micro Raman spectroscopy was performed using a Horiba JobinYvonLabRAM HR Raman spectrometer equipped with an automated xyz stage (Märzhäuser). Spectra were acquired over the range 100-1400 cm -1 using a 532 nm laser at 34 mW power, a 100x objective and a 50 µm confocal pinhole. To simultaneously scan a range of Raman shifts, a 600 lines mm -1 rotatable diffraction grating along a path length of 800 mm was employed. Spectra were detected using a Synapse CCD detector (1024 pixels) thermoelectrically cooled to −60 °C.
Before spectra collection, the instrument was calibrated using the zero-order line and a standard Si (100) reference band at 520.7 cm -1 . The spectral resolution in this configuration is better than 1.8 cm -1 .
For single point measurements of all glass formulations, spectra were acquired with an acquisition time of 30 seconds and 8 accumulations to improve the signal to noise ratio, from five random locations and averaged to give a mean spectrum. For M8T1 (as a representative of a 1 mol% TiO2 containing glass), a lateral map of the top surface of the glass was obtained by acquiring spectra at 2 µm steps within a square 40 x 40 µm (a total of 441 spectra). In this configuration, the spatial resolution was ~1 and ~3 µm in the lateral (xy) and axial (z) dimensions, respectively. As each individual spectrum was collected for 10 seconds, repeated 9 once in order to automatically remove the spikes due to cosmic rays, the whole map required approximately 2.5 hours of acquisition time. The intensity (as height) of the band in the region 845-940 cm -1 (a diagnostic vibrational mode the TiO5 unit) was evaluated within the map using univariate analysis within Labspec 6 software. Optical micrographs of the glasses were obtained using a 100x objective lens.
Nuclear magnetic resonance (NMR) spectroscopic analysis 31 P NMR is used to evaluate the structural connectivity of the phosphate glass network 38 .

P MAS NMR spectra were recorded at room temperature on a Varian Chemagnetics
Infinityplus spectrometer operating at a Larmor frequency of 121.468 MHz using a 4 mm MAS probe spinning at 12.5 kHz. The 31 P π/2 pulse duration was 3.0 µs, the spectral width was 100 kHz and the acquisition time was 10.24 ms. Chemical shifts are quoted relative to 85% H3PO4 using Na4P2O7.10H2O as an external secondary reference. Prior to acquiring quantitative 31 P spectra the spin-lattice relaxation time T1 was determined for each sample by saturation recovery. Saturation was achieved by 100 31 P π/2 pulses spaced by delays of between 5 and 20 ms with recovery delays of up to 1000 s. Quantitative 31 P NMR spectra required relaxation delays (5 T1) of the order of between 60 s and 250 s depending on the sample. The resulting spectra were deconvoluted into a set of Gaussian lineshapes which were integrated in order to quantify the proportions of the different Q environments in the sample. First-order MAS sidebands were included in the analysis. Higher-order sidebands contained less than 1 % of the spectral intensity and were neglected.

X-ray diffraction analysis
XRD traces of the glasses are presented in Figure 1. With the exception of M4T0 and M2T0 formulations, a single broad peak between 20 o and 40 o (2θ) was observed for all of the glass compositions. The absence of any sharp crystalline peak confirmed the amorphous nature of these glasses. The amorphous glasses were chosen for further characterisation in this study.
As M2T0 and M4T0 crystallised during manufacture they were not studied further. Figure 2 shows the density of the glasses. As can be seen, the density decreased from 2745 to 2725 kgm -3 with decreasing MgO content from 24 to 8 mol% for the glasses that did not contain TiO2. However, the titanium contianing glasses showed a higher density in comparison to M8T0.

Oxygen density and molar volume
Figures 3a and 3b present the oxygen density and molar volume of glasses with no TiO2 and with 1 mol% TiO2, respectively. As seen from Figure 3a, the oxygen density decreased from 1.30 to 1.25 gcm -3 , whereas the molar volume increased from 32.05 to 33.21 cm 3 mol -1 , as the CaO content increased from 16 to 32 mol% (i.e. the MgO content correspondingly decreased from 24 to 8 mol%). For the glasses with 1 mol% TiO2 (see Figure 3b), the oxygen density was seen to decrease from 1.26 to 1.24 gcm -3 , whereas molar volume was seen to increase from 11 33.1 to 33.5 cm 3 mol -1 , with increasing CaO content from 32 to 38 mol% (or decreasing MgO content from 8 to 2 mol%).

Thermal properties
From Table 2, it has been observed that the glass transition (Tg), onset of crystallisation (Tx), crystallisation peak (Tc), initial melting (Tm) temperatures and the stability (ΔT) of the glasses against crystallisation were found to decrease with decreasing MgO content. It is to be noted that, the Tg, Tx, Tc and Tm values of some of these formulations (except for the M8T1 glass) have been reported previously 39 . However, in this investigation those values were utilised to calculate the glass stability in order to address their crystallisation behaviour, as presented in

Crystallisation behaviour
As stated earlier, PBGs in the glass system 40P2O5- (where 0≤x≤22 and y=0 or 1) with MgO content 8 mol% and above 8 mol% were formed successfully, as confirmed by XRD analysis. However, further replacement of MgO with CaO showed a strong tendency towards crystallisation (i.e. M4T0 and M2T0 glasses, see Figure 1).
On the other hand, the addition of 1 mol% TiO2 in place of Na2O for glasses with less than 8 mol% MgO (i.e. M4T1, M2T1) reduced the tendency of the glass to crystallise. Figure 4a and 4b show 2 mol% MgO containing glasses without and with 1 mol% TiO2 content, respectively. 12 As the crystallisation temperature varied with glass particle sizes (see ESI Figure 1 for M24T0 as an example), a fixed particle size range of 45-100 µm was used to calculate the activation enery of cystallisation for all glass formulations. Figure 5 shows results for the M24T0 glass as an example of the effect of different heating rates on the crystallisation temperature. This process was repeated three times for each glass formulation. As can be seen, the crystallisation peak shifted to higher temperature (from 571 to 587 o C) and the peak intensity increased with increasing heating rate (from 10 to 20 o Cmin -1 ). In order to calculate the activation energy of crystallisation, ln (β/Tc 2 ) over 1/Tc was plotted for all glass formulations (see ESI Figure 2 for M24T0 as an example) and the slope of these lines corresponds to -Ec/R (where, R is the gas constant).  [40][41][42] were observed in the FTIR and Raman spectra, respectively, the position and proposed assignments of which are presented in ESI Table 1.
Of note in the IR spectra (Figure 6a), the intensity of the peaks at ca. 900 and 1100 cm -1 , associated with asymmetric P-O-P stretching vibrations of bridging oxygens in Q 2 and Q 1 tetrahedra, respectively 9,42 , were seen to increase with decreasing MgO content (from 24 to 8 mol%) for glasses containing no TiO2. However, the M8T1 glass showed a lower peak intensity at the same positions relative to M8T0. Moreover, redshifting of bands from 739 to 729 cm -1 (νs(P-O-P), Q 2 ), from 899 to 889 cm -1 (νas(P-O-P), Q 2 ) and from 1119 to 1101 cm -1 (νas(P-O-P), Q 1 ) was observed with decreasing MgO content for the glasses containing no TiO2.
Interestingly, these same vibrational modes were blueshifted to a higher frequency for M8T1 relative to M8T0.
In the Raman spectra (Figure 6b), analogous redshifting of bands, such as those associated with the PO2 symmetric stretching of non-bridging oxygen in Q 1 (from 1046 to 1042 cm -1 ) and Q 2 (from 1161 to 1158 cm -1 ) units, respectively, 43 Table   2 provides peak positions and relative proportions of Q 1 and Q 2 species for the individual glasses. The proportion of Q 2 species for the glasses without Ti was found to be between 50 to 51%. No significant variation in the relative proportions was observed between the M8T0 and M8T1 glasses.

Discussion
Several researchers have investigated the crystallisation behaviour of different types of glasses 35,46-48 and report that crystallisation is strongly affected by the glass composition, heating rate and particle size 35,46 . In this current study, the effect of decreasing MgO (and hence increasing CaO) content and the addition of TiO2 on glass formation ability was investigated for glasses in the system 40P2O5-(24-x)MgO-(16+x)CaO-(20-y)Na2O-yTiO2 (where 0≤x≤22 and y=0 or 1), including assessment of the physical, thermal, structural and crystallisation properties.
Density measurements give an indication of the degree of change in the glass structure with the variation of glass composition. The decrease in glass density from 2745 kgm -3 (for M24T0) 15 to 2725 kgm -3 (for M8T0) (see Figure 2) 8 . They reported that the increase in density was due to the replacement of low density Na with higher density element Ti 8 .
As seen from Figure 3, the oxygen density decreased, whilst the molar volume increased, with the substitution of MgO by CaO in this glass series. This was attributed to the fact that calcium has a larger ionic radius than magnesium, and as a result has lower polarising power due to its small charge to size ratio and subsequently lower attractive force to non-bridging oxygens (NBO) 35  The thermal properties of phosphate-based glasses (PBGs) are strongly dependent on their structural features, such as chain length, cross-linking density and bonding strength. The decrease in Tg, Tx, Tc and initial melting temperature with decreasing MgO content for glasses that did not contain TiO2 (see Table 2) was also attributed to the same reasons as discussed above for the oxygen density, i.e. due to the replacement of Mg-O bonds with Ca-O bonds. 16 The chemical bond strength of Mg-O is known to be higher than Ca-O, as Mg 2+ has a smaller ionic radius (0.65 Å) compared to Ca 2+ (0.99 Å) 49 . Furthermore, the smaller size of Mg results in a high charge to size ratio and thus exhibits higher electronegativity, which creates higher attractive forces to non-bridging oxygens (NBO) as compared to Ca 35 . Therefore, it was anticipated that Mg formed a strong cross-linking with phosphate chains, compare to Ca 50 , which also correlated with a high compact structure for higher Mg content glasses as indicated by the oxygen density data. in Tg with addition of TiO2 was due to the smaller ionic radius of Ti 4+ (0.56 Å) compared to Na + (0.97 Å), where the size of ions (i.e. ionic radius) introduced into the glass network is a key factor in controlling their chemical durability and thermal stability. As TiO2 has a smaller ionic radius and higher electrical charge than Na + it most likely generated stronger cross-linking between the phosphate chains via creating Ti-O-P links [51][52][53] . This finding also correlated with the compactness of the glass structure as the oxygen density increased with the addition of TiO2.
The glass formation tendency of the formulations was seen to decrease with the substitution of MgO with CaO as the glass stability against crystallisation decreased from 106 o C (for M24T0) to 90 o C (for M8T0) (see Table 2). Formulations containing less than 8 mol% MgO in the series investigated showed a strong tendency towards crystallisation (see Figure 4a). This was suggested to be due to the lower field strength of Ca 2+ (0.33) as compared to Mg 2+ (0.45) 54 . According to Dietzel, MgO can behave as an intermediate oxide 55 and in its intermediate state, can potentially act as a network former as well as a cross-linker between the phosphate chains 50,56,57 . On the other hand, Ca acts only as a network modifier and depolymerises the phosphate chains, which consequently increases the mobility of the phosphate structural groups in the melt 34 . As a result, the components can arrange themselves into an ordered crystalline structure more easily 34 . Therefore, for this glass series, a decrease in MgO content lead to an increase in crystallisation tendency. It has also been reported that the addition of TiO2 to PBGs increased the glass stability by the formation of cross-links between the phosphate chains 21,52,53,58 . A number of studies have also suggested that titanium can enter the phosphate glass network as Ti 4+ ions and thus behave as a glass network former 21,22 .
Therefore, the crystallisation tendency may have been inhibited by the addition of 1 mol% TiO2 (see Figure 4b).
Crystallisation is generally described by the activation energy for crystallisation (Ec) and the Johnson-Mehl-Avrami (JMA) exponent 47 . As seen from Figure 5, the crystallisation temperature and peak height increased with increasing heating rate. This could be attributed to the fact that the heat flow to the samples increased with increasing heating rate 35  rate on the crystallisation peak intensity for 50P2O5-40CaO-SrO-10Na2O glasses with particle sizes in the range 300-500 µm 47 . They reported that the crystallisation peak intensity 18 increased from 1.1 to 1.4 Wg -1 with increasing heating rate from 5 to 20 o Cmin -1 47 . The changes in the width (broad vs narrow) of the crystallisation peaks are usually indicative of changes in the crystallisation mechanism 46 . The intensity of crystallisation peaks are often used to determine whether crystallisation has occurred at the surface or within the bulk of the materials 33,61 . A broad crystallisation peak can be associated with surface crystallisation whereas a sharp peak is associated with bulk crystallisation 46 .
The decrease in Ec with decreasing MgO content for glasses containing no Ti content and the higher Ec for M8T1 (compared to M8T0) was attributed to the effect of field strength has been discussed above for oxygen density and thermal analysis. This result also correlated well with the glass stability (ΔT) data (shown in Table 2 interaction as Mg 2+ has higher field strength than Ca 2+ 9 . On the other hand, the Ti containing glasses showed broader peaks as compared to M8T0, which may have been due to Ti 4+ increasing the phosphate-cation bonding interactions compared to Na + 9 . The bands at around 1000 and 1100 cm -1 were assigned to symmetric νs(PO3) 2and asymmetric νas(PO3) 2modes, respectively which are associated with Q 1 species 16,65 . The intensity of the absorption bands for asymmetric stretching modes, νas(PO3) 2of the chain terminating groups of the glasses increased with decreasing MgO content down to 8 mol%, which was an indication for decreasing the cross-links between the glass network and thereby reducing the rigidity of the network 41 . However, the addition of 1 mol% TiO2 (i.e. M8T1) reduced the intensity (compared to M8T0) of the absorption bands for the asymmetric stretching modes, νas(PO3) 2significantly. 20 Similarly, this could be attributed to the increase in degree of polymerisation or creating more cross-linking between the glass network as Ti can act as a network former or network modifier 41 .
From the mean Raman spectra, obtained by averaging point spectra collected from five random locations, a diagnostic peak for Ti containing glasses was observed at 899 cm -1 , associated with a Ti-O stretching vibration in TiO5 (titanyl) units and confirming the role of TiO2 as a network modifier 44 . Although subtle differences in the intensity of the red colouration were noted in the false colour spectroscopic map, associated with differences in surface topography, TiO5 was found to be within every pixel of the map, indicating its uniform dispersion throughout the surface of the glass. No evidence for the presence of TiO6 (titanate) units, which typically are observed at around 640 cm -1 in the Raman spectrum, and known to act as a network former 44 , were found for the Ti containing glasses investigated.
The addition of 1 mol% TiO2 to the phosphate glass network (in the glass series investigated) was shown to be sufficient to increase cross-linking between the phosphate chains, and was sufficient to reduce the crystallisation tendency of the lower Mg containing glasses in the glass system 40P2O5-(24-x)MgO-(16+x)CaO-(20-y)Na2O-yTiO2 (where 0≤x≤22 and y=0 or 1).

Conclusions
The glass forming ability for the composition with the replacement of MgO by CaO in 40P2O5-