High Elastic Moduli of a 54Al2O3-46Ta2O5 Glass Fabricated via Containerless Processing

Glasses with high elastic moduli have been in demand for many years because the thickness of such glasses can be reduced while maintaining its strength. Moreover, thinner and lighter glasses are desired for the fabrication of windows in buildings and cars, cover glasses for smart-phones and substrates in Thin-Film Transistor (TFT) displays. In this work, we report a 54Al2O3-46Ta2O5 glass fabricated by aerodynamic levitation which possesses one of the highest elastic moduli and hardness for oxide glasses also displaying excellent optical properties. The glass was colorless and transparent in the visible region, and its refractive index nd was as high as 1.94. The measured Young’s modulus and Vickers hardness were 158.3 GPa and 9.1 GPa, respectively, which are comparable to the previously reported highest values for oxide glasses. Analysis made using 27Al Magic Angle Spinning Nuclear Magnetic Resonance (MAS NMR) spectroscopy revealed the presence of a significantly large fraction of high-coordinated Al in addition to four-coordinated Al in the glass. The high elastic modulus and hardness are attributed to both the large cationic field strength of Ta5+ ions and the large dissociation energies per unit volume of Al2O3 and Ta2O5.

due to their high melting temperatures. These issues have limited the fabrication of bulk glasses with high elastic moduli and high hardness values. Recent progress in containerless processing has, however, allowed the vitrification of low glass forming materials, including those without added network formers such as TiO 2 -based, Nb 2 O 5 -based, WO 3 -based, and Al 2 O 3 -based compositions, because heterogeneous nucleation from the melt can be avoided with this technique [13][14][15][16][17][18] . Thus, R 2 O 3 -Al 2 O 3 glasses containing large quantities of Al 2 O 3 have been prepared and found to exhibit superior mechanical properties as expected 19,20 . As a result, Al 2 O 3 -based glasses have attracted interest as high elastic moduli and high hardness materials. The properties of such glasses should be enhanced through the incorporation of additional components other than Al 2 O 3 with high dissociation energies and high packing volumes. Herein, we describe the preparation of the new 54Al 2 O 3 -46Ta 2 O 5 glass, which exhibits high elastic moduli and hardness values, using containerless processing. The thermal, optical, and mechanical properties of the glass are also reported. In addition, an approach to the design of glasses with higher elastic moduli and higher hardness is proposed on the basis of the results of the local structure analysis around aluminum performed using 27 Al MAS NMR spectroscopy. Figure 1 shows the Differential Thermal Analysis (DTA) curve for the 54Al 2 O 3 -46Ta 2 O 5 glass. The glass transition temperature T g is located at 858 °C, and the first T P1 and second T P2 crystallization peak are observed at 912 °C and 1054 °C, respectively. The difference between T P1 and T g (Δ T = T P1 − T g ) a measure of the thermal stability of the glass, is 54 °C, indicating the difficulty for vitrifying the glass using a conventional melting process. X-ray Diffraction (XRD) analysis confirmed that glass was totally amorphous and that the main phase of the crystallized sample after DTA was AlTaO 4 . The density of the annealed glass was ρ = 6.01 g/cm 3 . The composition of the glass samples measured by x-ray fluorescence (XRF) showed that the changes with respect to the nominal composition were less than 1 mol%. The microstructure of the fabricated glasses investigated through high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) is shown in Fig. 2. Observation through the HAADF-STEM has the advantage of achieving chemical contrast at the nanometric scale because it is very sensitive to the atomic number 21 . From the figure it can be observed that the glass is homogeneous at different scales and no phase-separation is observed. The randomly distributed bright points at the highest magnification are associated with the Ta atoms which have a much larger atomic number compared with the Al atoms (dark regions). Figure 3 shows the transmittance spectrum of the 54Al 2 O 3 -46Ta 2 O 5 glass in the ultraviolet-visible (UV/vis) region. The glass was transparent in the visible region and had a maximum apparent transmission of 81%. The maximum theoretical transmittance was also estimated to be 81% using the equation 2 , and the experimental refractive index n d value of the glass which was found to be 1.94. The estimated value was similar as that of the experimental result, indicating that the apparent transmittance value was to the result of losses only due to sample surface reflection, and no light scattering occurred in the glass 22 . As observed in the inset of Fig. 3, the glass is colorless and transparent, which confirms that the valence state for all of the Ta ions is five, and no Ta 4+ ions are present 23 . The optical bandgap energy was estimated to be 4.3 eV using the UV absorption edge located at 288 nm.

Results
The measured longitudinal velocity V P and transversal velocity V S of the 54Al 2 O 3 -46Ta 2 O 5 glass were 5.86 km/s and 3.20 km/s, respectively. From these values and the experimental density, it was found that the Young's modulus E was 158.3 GPa, the bulk modulus K was 124.1 GPa, the shear modulus G was 61.5 GPa, and the Poisson's ratio v was 0.29. These values for the elastic moduli are considerably high and comparable to the maximum values in oxide glasses such as 40Y 2 O 3 -55Al 2 O 3 -5SiO 2 and 28.5La 2 O 3 -71.5Al 2 O 3 , whose Young's moduli were determined using Brillouin spectroscopy (169 GPa); however our measurement system showed that the Young's modulus of those glasses were 145.5 GPa and 123 GPa respectively 1,9,10 . The Vickers hardness of the 54Al 2 O 3 -46Ta 2 O 5 glass was 9.10 ± 0.05 GPa, which   Figure 4 shows indentation imprint for the 54Al 2 O 3 -46Ta 2 O 5 glass at a load of 2.942 N. Extensive lines due to shear deformation on each face of the imprints are observed. In addition, at the same load, some of the imprints exhibited radial crack behavior 25,26 . No cracks were observed in any indentation below 1 N. The indentation cracking resistance (CR) was estimated to be 2.50 ± 0.13 N, which is comparable to a commercial Vycor glass 27 .
The 27 Al MAS NMR spectrum of the 54Al 2 O 3 -46Ta 2 O 5 glass is presented in Fig. 5. Although the spectrum is broad due to the amorphous nature of the glass, two distinctive peaks and a small shoulder were observed. These peaks and the shoulder were assigned to 4-coordinated Al (Al [4] ), 5-coordinated Al (Al [5] ), and 6-coordinated (Al [6] ), respectively [28][29][30] . The spectrum was decomposed into the three components using the "dmfit" program applying a simple Czjzek model 31,32 . The thin dotted lines in the spectrum correspond to each of the components. The fitting, values for δ iso (isotropic chemical shift), dCSA (width of the Gaussian distribution of δ iso ), and vQ* (quadrupolar product in kHz) were determined to be 64.8 ppm, 15 ppm, and 1134 kHz for Al [4] ; 36.7 ppm, 12 ppm, and 985 kHz for Al [5] ; and 10.3 ppm, 15 ppm, and 973 kHz for Al [6] , respectively 33 . Based on the integration of the peak areas, the fractions of Al [4] , Al [5] , and Al [6] were estimated to be 44.1%, 41.9%, and 14.0%, respectively. The estimated average oxygen coordination number for Al was 4.7. The fractions of Al [5] and Al [6] were considerably larger than those observed in other aluminate glasses; Al typically forms AlO 4 tetrahedra in MO-Al 2 O 3 (M = Ca, Sr and Ba) glasses 30 . While Al [5] and Al [6] have been observed in some Al 2 O 3 -containing glasses, such as and CaO-Al 2 O 3 -SiO 2 , the fraction of Al [5]   The total fitting curve corresponds to the sum of three Czjzek fitting curves: blue dots (Al [4] ), red dashes and dots (Al [5] ), and orange dashes (Al [6] ).
Scientific RepoRts | 5:15233 | DOi: 10.1038/srep15233 has generally ranged from 3 to 30%, and that of Al [6] from 1 to 2% 24,34,35 . The structure of the 54Al 2 O 3 -46Ta 2 O 5 glass may therefore be due to not only the presence of AlO 4 networks but also result in part from the high oxygen coordination of Al. The mechanism of glass formation with retention of large fractions of Al [5] and Al [6] is interesting and thus will be the subject of further investigations.

Discussion
These combined results indicate that the 54Al 2 O 3 -46Ta 2 O 5 glass have good mechanical properties, high transparency and a high refractive, with an unconventional amount of Al [5] and Al [6] species. In order to understand the origin of the good mechanical properties of the glass the results are analyzed within the context of the Makishima and Mackenzie model e.g. the atomic packing density and dissociation energy per unit volume of the glass components.
The atomic packing density C g was calculated from the density using the formula  36 . The coordination numbers for Ta and O in the 54Al 2 O 3 -46Ta 2 O 5 glass were assumed to be 6 and 2, respectively, and the fractions of the coordination numbers for Al estimated from the results of the 27 Al MAS NMR were used. The atomic packing density C g was found to be 0.586, which is significantly larger than those for conventional oxide glasses (i.e., for SiO 2 glass, C g is 0.452). The small molar volume of Ta 2 O 5 and the large fraction of highly coordinated Al are thought to contribute to the high packing density of the 54Al 2 O 3 -46Ta 2 O 5 glass. It has been suggested that the formation of highly coordinated Al in aluminate glasses is promoted by the large cationic field strength, as observed in R 2 O 3 -Al 2 O 3 -SiO 2 glasses [8][9][10][11] . Ta 5+ also has large cationic field strength because of its small ionic radius and high valence state. Accordingly, Ta 2 O 5 likely contributes to the high packing density of the 54Al 2 O 3 -46Ta 2 O 5 glass via the formation of a large number of highly coordinated Al atoms.
A high content of Ta 2 O 5 is also characteristic of the 54Al 2 O 3 -46Ta 2 O 5 glass. The dissociation energy of Ta 2 O 5 is substantially large (95.6 kJ/cm 3 ) 11 . The elastic moduli of the glass were estimated using the Makishima

and Mackenzie equation given by
Here G i is the dissociation energy of each component oxide. Values of 131 kJ/cm 3 , 125 kJ/cm 3 and 119.2 kJ/cm 3 were used for G Al2O3 with Al in coordination of 4, 5, and 6 respectively 11,37,38 . The calculated Young's modulus E of the 54Al 2 O 3 -46Ta 2 O 5 glass was 131.9 GPa, which was approximately 17% less than the experimentally determined value, but still in relatively good agreement. A more accurate model may be necessary for estimation of the atomic packing density that includes the real contribution of the more highly coordinated cations. The energy contribution ratios of Al 2 O 3 and Ta 2 O 5 to the Young's modulus were also estimated using the Makishima and Mackenzie equation and found to be 62% and 38%, respectively. It should be noted that the contribution of Al 2 O 2 is not that high, while that of Ta 2 O 5 is considerably high, which is unlike most other binary aluminate glasses with high elastic moduli. For example, a 28.5La 2 O 3 -71.5Al 2 O 3 glass, which has one of the highest reported Young's modulus values among the oxide glasses, has the following contribution: 16.71% from La 2 O 3 and 83.3% from Al 2 O 3 . It has been previously accepted that a large contribution by Al 2 O 3 is necessary to achieve a high elastic modulus for binary aluminate glasses, such as R 2 O 3 −Al 2 O 3 . However, a simple estimation of the energy contribution of the components in 54Al 2 O 3 -46Ta 2 O 5 glass revealed that an appropriate component, like Ta 2 O 5 , can increase the elastic modulus even if the dissociation energy contribution of Al 2 O 3 is small.
In summary, a glass with composition 54Al 2 O 3 -46Ta 2 O 5 was fabricated using an aerodynamic levitation technique. Its glass transition temperature T g was 858 °C, and crystallization occurred at 54 °C above T g , indicating a low glass forming ability. The glass is colorless and highly transparent in the visible region and has a refractive index n d of 1.94. The Young's modulus E, bulk modulus K, shear modulus G, and Poisson's ratio v of the 54Al 2 O 3 -46Ta 2 O 5 glass were determined via ultrasonic pulse-echo overlap analysis and were found to be 158.3 GPa, 124.1 GPa, 61.5 GPa, and 0.29, respectively, while the Vickers hardness of the glass was found to be 9.1 GPa. These elastic moduli and Vickers hardness values are quite high and comparable to the maximum values of conventional oxide glasses. In addition, an indentation cracking resistance of 2.5 N was estimated from the indentation experiments. Furthermore, 27 Al MAS NMR spectroscopic analysis revealed that the fractions of Al [4] , Al [5] , and Al [6] in the 54Al 2 O 3 -46Ta 2 O 5 glass were 44.1%, 41.9%, and 14.0%, respectively, and the average oxygen coordination number of the Al cations was 4.7. Notably, the fractions of Al [5] and Al [6] are considerably large compared to those observed in conventional oxide glasses, and may form because of the large cationic field strength of Ta 5+ . These results indicated that Ta 2 O 5 was a key contributor to the high elastic moduli and high hardness values of the glass because the addition of Ta 2 O 5 increases the packing density via formation of Al atoms that are highly coordinated with oxygen and because the Ta 2 O 5 itself has a large dissociation energy. Moreover, a simple estimation of the energy contributions of the components in the 54Al 2 O 5 -46Ta 2 O 5 glass using the Makishima and Mackenzie equation also suggested that the use of appropriate components can increase the elastic moduli even if the contribution of Al 2 O 3 is small. These results provide insight into the design and fabrication of harder glasses based on both the local structure and the dissociation energies of the components.
Scientific RepoRts | 5:15233 | DOi: 10.1038/srep15233 Methods Glass synthesis. Glasses were fabricated using an aerodynamic levitation furnace described elsewhere 23 . High-purity (99.99%) α-Al 2 O 3 and Ta 2 O 5 powders were mixed stoichiometrically with the chemical composition 54Al 2 O 3 -46Ta 2 O 5 , pelletized using a hydrostatic press, and annealed at 1050 °C for 12 h in air. Pieces obtained from the crushed pellets were levitated in an oxygen gas flow and melted using two CO 2 lasers at approximately 2000 °C. The melt was rapidly solidified by shutting off the lasers at a cooling rate of approximately 300 °C/s in order to obtain fully vitrified samples. The obtained spherical glasses (2 mm in diameter) were colorless and transparent. Glass formation was confirmed via Cu Kα XRD analysis (Rigaku, RINT 2000). In order to rule out any compositional changes of the glass during the melting process, X-ray fluorescence experiments (JEOL, JSX-3100RII) were performed on glass samples under vacuum conditions. Glasses with composition 40Y 2 O 3 -55Al 2 O 3 -5SiO 2 , 28.5La 2 O 3 -71.5Al 2 O 3 , and 29.3Al 2 O 3 -50.2SiO 2 -20.5Sc 2 O 3 were also fabricated using the levitation technique for comparative purposes.
Scanning transmission electron microscopy observation. In order to verify the homogeneity of the fabricated glasses observation with a scanning transmission electron microscope (JEOL, ARM-200CF) coupled with a high-angle annular dark field (HAADF) detector was performed. The microscope was equiped with a spherical aberration corrector (Ceos, Gmbh) and a cold field emission gun was used. The probe-forming aperture angle was 24.5 mrad while the HAADF and bright field (BF) detectors spanned through 68-280 and 0-17 mrad respectively. The spatial resolution of the present observation was approximately 0.1 nm. Glass powders were dropped into a perforated amorphous carbon films supported in Cu grids. No sputtering or heating was applied to the samples prior to the observation.
Thermal and physical properties. The glass transition temperature T g and crystallization temperature T p were determined via DTA at a heating rate of 10 °C/s (SII, TG6300). Prior to the analysis of the physical and structural properties, the glasses were annealed at 10 °C above the T g in order to relax the stress introduced during quenching. The density ρ was determined using gas pycnometry (Micrometrics, AccuPyc II 1340). The experimental error associated with the density measurements was smaller than 0.01 g/cm 3 3  Indentation behavior. Indentation experiments were performed on a Shimazu DUH HMV-1 Vickers tester at 23 °C and approximately 60% relative humidity. Optical-grade polished samples with a thickness of approximately 500 μm were used. The dwell time was set at 15 s. The Vickers hardness values H V were calculated from the diagonal lengths of the imprints at a load of 0.980 N. At least 20 indents were made for measuring H V . The indentation cracking resistance CR values were estimated from cracking probability curves using the method proposed by Wada et al. 40 . Here, CR is defined as the load required to generate two radial cracks on average or to achieve a 50% cracking probability. Each data point on the cracking probability curves developed in the present study also represents 20 indentation imprints. The reported value of CR was obtained by averaging the CR values determined from sigmoidal fittings of the cracking probability curves for three different samples. The imprints were observed by optical microscopy.
Al local structure. 27 Al MAS NMR spectroscopy of the glass was performed on a JEOL JNM-ECA 500 spectrometer equipped with MAS probe head at 11.74 T (500 MHz). The spinning rate was 15 kHz, and a 4-mm-diameter zirconia rotor was used. The NMR spectra were recorded using π/6 pulses (0.4 μs) and a relaxation delay of 1 s, and 4000-12000 signals were accumulated. The 27Al chemical shift δ iso in parts per million (ppm) was referenced to an external 1 M AlCl 3 solution (− 0.1 ppm).