Densification in transparent SiO2 glasses prepared by spark plasma sintering

Recently, spark plasma sintering (SPS) has become an attractive method for the preparation of solid-state ceramics. As SPS is a pressure-assisted low-temperature process, it is important to examine the effects of temperature and pressure on the structural properties of the prepared samples. In the present study, we examined the correlation between the preparation conditions and the physical and structural properties of SiO2 glasses prepared by SPS. Compared with the conventional SiO2 glass, the SPS-SiO2 glasses exhibit a higher density and elastic modulus, but a lower-height first sharp diffraction peak of the X-ray total structure factor. Micro-Raman and micro-IR spectra suggest the formation of heterogeneous regions at the interface between the SiO2 powders and graphite die. Considering the defect formation observed in optical absorption spectra, reduction reaction mainly affects the densification of SPS-SiO2 glass. Hence, the reaction at the interface is important for tailoring the structure and physical properties of solid-state materials prepared by the SPS technique.

www.nature.com/scientificreports/ between the structure and the physical properties of the prepared SiO 2 glass. However, a detailed study of the structure of SPS-SiO 2 glass is lacking. In this study, we performed a structural analysis of SiO 2 glass prepared by the SPS method and compared its characteristics with those of conventional SiO 2 glass. In addition, spaceselective microscopic analysis was used to determine the spatial heterogeneity of SPS-SiO 2 glass.

Results and discussion
Optical and physical properties. The obtained SPS-SiO 2 glasses were transparent to the naked eye. In the previous report, SiO 2 glasses were prepared in the temperature range of 727-1427 °C at 100 MPa pressure, but transparent bulk material was obtained with the sintering temperature above 1250 °C 6 . Since transparent SPS-SiO 2 glass is the target of the study, four preparation conditions were used: 6 MPa, 1300 °C; 6 MPa, 1400 °C; 70 MPa, 1300 °C; and 70 MPa, 1400 °C. The physical properties of the prepared glasses are listed in Table 1. All the SPS-SiO 2 glasses exhibit higher densities than that of conventional SiO 2 glass. In addition, the dense SPS-SiO 2 glasses possess a higher G 0 , E 0 , and K 0 than those of conventional SiO 2 glass. Despite the same preparation conditions, the densities of the SPS-SiO 2 glasses were changed depending on the weight of starting chemicals. The lighter the starting material, the heavier the density. Notably, the dependence of the preparation conditions (temperature and pressure) on the elastic properties is obscure. Therefore, it is expected that not only the temperature, pressure, and heating program but also the volume of the starting material affects the nature of the obtained samples. Considering the mechanism of the SPS method, the distance between the graphite punches might affect the sintering efficiency and the resulting properties [14][15][16] . Figure 1 shows the optical absorption spectra of the glasses along with that of conventional SiO 2 glass. Small absorption bands are observed below 350 nm, which originate from the generation of defects, such as oxygen deficiency centres and dangling oxygen bonds [11][12][13] . It is notable that absorbance of the samples prepared at 6 MPa is higher than that of ones prepared at 70 MPa under the same preparation temperature. We also found that the increasing absorbance, i.e. defect formation, is suppressed by applying lower temperature under the same preparation pressure.
Positron annihilation spectroscopy (PAS) analysis. For structural analysis of the SPS-SiO 2 glasses, we performed PAS, which is used to quantify the cavity size in materials [17][18][19][20] . Figure 2 shows the positron decay curves of the SPS-SiO 2 and conventional SiO 2 glasses; there is no significant difference in the curves. For insulators, the decay constant of ortho-positronium (the third component in fitting the decay curve) is used for size calculation. The cavity radius of the SPS-SiO 2 glass was calculated to be 0.245 (± 0.002) nm, which is almost identical to that of standard SiO 2 glass (0.247 nm) (Supplementary Table 1). This result is in accordance with  20 . As reported previously 19 , PAS tends to detect larger cavities (rather than smaller ones) in SiO 2 glass. Hence, the small changes observed in the density cannot be explained using the cavity sizes obtained by PAS as it is based on preferential annihilation of positrons in cavities. A microscopic analysis is expected to provide additional information to explain the densification of SPS-SiO 2 glasses.

Microscope observation.
We performed Scanning Electron Microscope (SEM) and Transmission Electron Microscope (TEM) imaging of the surface and the interior to analyse the morphologies of the SPS-SiO 2 glasses. Figure 3 shows the SEM images of SPS-SiO 2 glasses prepared at different conditions. Inset shows the surface of the samples by mechanically polishing. Inside the samples prepared by ion-milling, there are no grain boundaries originating from the SiO 2 powder, resulting in a uniform morphology. The fully sintered images at 1300 and 1400 °C are very similar to those in the previous paper 6 . It is natural that the obtained SPS-SiO 2 glasses without grain boundary exhibit the transparency over a wide wavelength range. Figure 4 shows the TEM images of SPS-SiO 2 glasses prepared at different conditions. Both the surface and the interior of the samples are homogenous with no precipitation of crystallites. Hence, all the SPS-SiO 2 glasses are expected to be amorphous, which was confirmed by conventional XRD (discussed later). Because there is no clear evidence of densification in the TEM results, other analytical approaches are required.
High energy XRD (HEXRD) analysis. Based on the relationship between the first sharp diffraction peak (FSDP) of the X-ray total structure factor S(Q) and the density of SiO 2 glasses (as reported in our recent study) 16 , we focused on the FSDP profile of the SPS-SiO 2 glasses. The FSDP height at Q = 1.55 Å −1 strongly correlates with the structural disordering of glasses [21][22][23][24][25] . Figure 5a shows the S(Q) of the SPS-SiO 2 glasses prepared at 70 MPa, 1400 °C along with that of SiO 2 glass. The X-rays were irradiated at the centre of the samples. The spectral shapes   Figure 6b shows the enlarged microscopic IR spectra obtained from different positions of the SPS-SiO 2 glass prepared at 6 MPa, 1400 °C. The spectra obtained from the edges of the SPS-SiO 2 glass are different from that obtained from the centre of the sample; the Si-O-Si peak shifts to lower wavenumbers when moving from the left edge to the right. A shift to a lower wavenumber can be assigned to a higher T f , i.e., not annealed. However, the peak shift in the present data is too large to explain this phenomenon from the viewpoint of merely the T f of SiO 2 glass 16,29 . In addition, although the SPS-SiO 2 glasses were obtained by cooling without temperature control, the quenching rate was slower than the water-quenching rate of SiO 2 glass with a high T f . Therefore, it is expected that this shift originates from the reaction at the interface between the SiO 2 powder and the surrounding graphite die.
In Raman spectroscopy, we focused on the boson peak and the vibrational modes at 490 cm −1 (D 1 ) and at 600 cm −1 (D 2 ) [34][35][36][37][38][39] . Therefore, we discuss Raman spectra measured with HH (parallel nicol) polarisation (Supplementary Fig. 1(a)). Although the origins of both these vibrational modes are not completely understood, the boson peak has been correlated with the free volume of glasses 40 . On the other hand, D 1 and D 2 are assigned to the vibration of four-and three-membered rings of SiO 4 tetrahedral units, respectively 39 . In contrast to conventional SiO 2 glass, SPS-SiO 2 glass exhibits fluorescence upon laser irradiation, suggesting the formation of defects in the matrix (Supplementary Fig. 1(b)). The expected defect generation is consistent with the results of optical absorption shown in Fig. 1. Considering the results of micro-IR spectroscopy, the Raman spectra were also obtained from three different points (shown in the inset of the photograph) on the samples. The broad baseline due to fluorescence was removed by applying an extended multiplicative signal correction (EMSC) algorithm, adopting the bulk SiO 2 glass spectrum without fluorescence as a reference. 41 Figure 7a shows the micro-Raman spectra of the SPS-SiO 2 (prepared at 6 MPa, 1400 °C) and conventional SiO 2 glasses; these spectra were recorded with HH polarisation. Although the spectral shapes are roughly similar, there is a slight difference between the spectra of SPS-SiO 2 and SiO 2 glasses. Figures 7b-d show the enlarged Raman spectra highlighting the boson, D 1 , and D 2 peaks, respectively. The height of the boson peak of SPS-SiO 2 glass is comparable to that of the SiO 2 glass, and a remarkable peak shift is not observed. On the contrary, the intensities of the D 1 peak at 490 cm −1 (Fig. 7c) and the three-membered-ring (D 2 ) peak at 600 cm −1 (Fig. 7d) of SPS-SiO 2 glass increase. It is suggested that D 2 structures exist at the vicinity of SiO 2 surface 39 . Considering the result of PAS, in which a large cavity is not diminished by SPS, it is suggested that small silica units are formed at the edges of the sample (near the interface with the mould). Because the decrease in the height of FSDP in HEXRD (Fig. 5) also suggests a less ordered glass network, it can be concluded that a defect-like structure is generated in the SPS-SiO 2 glass. Notably, the height of the boson peak is comparable, although the shift in the Si-O-Si vibrational peak is the largest at the left edge of the sample (interface with the mould). The correlation between the Boson peak of silica glass and the average of distribution of Si-O-Si bonding angle is well established 42,43 . Since there is no visible correlation between the boson peak in the Raman spectrum and the Si-O-Si vibrational peak in the IR spectrum, it is expected that Raman spectroscopy is less sensitive to small density changes than IR spectroscopy.
The results of the present study suggest that the interface with the mould affects the properties of ceramics obtained by pressure-assisted fabrication under reduced atmosphere. We assume that the highest density of SPS-SiO 2 glass prepared at 6 MPa and 1400 °C is mainly due to reduction reactions, rather than conventional densification by applying pressure above several GPa. In addition, it is expected that higher pressure and lower temperature are effective to prevent reduction reaction. Even in the aerodynamic levitation method 44 , where www.nature.com/scientificreports/ samples are not in contact with any container, the interface between the materials and surrounding atmosphere is considered important. We believe that the influence of the interface on the structure and physical properties should be considered not only in SPS but also other manufacturing methods. Thus, ceramics with core-shell-like structures can be spontaneously prepared by selecting the fabrication method. In addition, we emphasise that different probes are required for different scales of constituents in a solid-state matrix. For example, because the relationship between elastic (macroscopic) properties and spectroscopic (microscopic) approaches is not straightforward, a combination of analytical methods is necessary for complete characterisation [45][46][47] . A deeper understanding of the structure of the materials is required for more precise control of the properties.

Conclusion
Transparent SPS-SiO 2 glasses without grain boundary or precipitation of crystallites were successfully prepared. The results of HEXRD, micro-IR spectroscopy, and micro-Raman spectroscopy suggest the formation of irregular silica species in the SPS-SiO 2 glass near the interface with the mould. Since the highest density of SPS-SiO 2 glass was attained prepared at lower pressure and higher temperature, it is expected that that the densification of SiO 2 glass is mainly due to reduction reactions, rather than conventional densification by applying GPa pressure. Although sintering can produce novel solid-state matrices, spatial analysis and macroscopic properties are important for understanding the nature of the sample.

Methods
Preparation. SiO     www.nature.com/scientificreports/ Characterisation. A high-energy X-ray diffraction (XRD) experiment was performed using a two-axis diffractometer dedicated to the study of disordered materials at the BL04B2 beamline of the SPring-8 synchrotron radiation facility (Hyogo Japan) 48 . The energy of the incident X-rays is 61.43 keV. The raw data were corrected for polarisation, absorption, and background, and the contribution of Compton scattering was subtracted using a standard data analysis software. The morphology of the samples was measured using a scanning electron microscope (SEM), where SEM images were taken using a JSM-6510(JEOL). An HF-2000 (Hitachi) microscope was used to obtain transmission electron microscopy (TEM) images. Transmittance measurements were performed at 25 °C using a spectrophotometer (UH-4150; Hitachi, Ltd.). The ultrasonic velocities of the longitudinal (V L ) and transverse (V T ) waves were measured at room temperature using the ultrasonic pulse-echo method (DPR300, JSR Ultrasonics). The frequencies of the longitudinal and transverse transducers are 10 and 5 MHz, respectively. The Young's modulus (E 0 ), instantaneous shear modulus (G 0 ), bulk modulus (K 0 ), and Poisson's ratio (ν) were calculated according to previously reported methods 47 . The errors in E 0 , G 0 , and K 0 were less than ± 0.1 GPa.
The micro-Raman spectroscopy was performed in backscattering geometry using a single-frequency diodepumped solid-state laser operating at 532 nm (Oxxius LCX-532S-300) and a custom-built microscope with ultra-narrowband notch filters (OptiGrate). The incident laser was attenuated to 7 mW and focused using a 50 × objective lens. The scattered light, collected by the same lens, was analysed using a single monochromator (Jovin-Yvon, HR320, 1200 grooves/mm) equipped with a charge-coupled device camera (Andor, DU420). The measurement system for Raman spectra is shown in a previous study 49 . We employed a multivariate analysis software Unscrambler 11 (Camo Analytics), which provides EMSC, to remove fluorescence-like broad baselines for the SPS samples.
Positron annihilation lifetimes were measured using a PSA TypeL-II system (Toyo Seiko Co., Ltd.) with an anti-coincidence system 50 . The 22 Na source, with a diameter of 15 mm, was encapsulated in a Kapton film. The accumulated count for each sample was 10 7 .
Infrared spectra (in the mid-infrared (IR) region from 600 to 8000 cm −1 ) were measured using the IR beamline BL43IR at the SPring-8 synchrotron facility (Hyogo, Japan). A Fourier transfer infrared (FTIR) microspectrophotometer (BRUKER model HYPERION IR microscope with a VERTEX70 FTIR spectrometer) was used with IR synchrotron radiation. The microscope has a motorised xy-stage, which is used to specify the measurement position. The magnification of the objective mirrors is 36 × . The spatial resolution is approximately 20 µm at 2000-3000 cm −1 . The wavenumber resolution is 1 cm −1 , and the number of accumulations is 4000 times. The sample was set on a stainless steel mesh with a honeycomb array with a hole diameter of 2 mm. All the measurements were performed at room temperature. The infrared optical path was purged with dry air to remove water and CO 2 molecules (FT-IR purge gas generator; Parker Co., Ltd.).