Photocatalytic hydrogen generation of monolithic porous titanium oxide-based glass–ceramics

A large relative surface area is crucial for high catalytic activity. Monolithic catalysts are important catalytic materials because of minimal self-degradation. Regarding large surface area catalysts, the glass–ceramics (GCs) with high formability, obtained by heat-treatment of the precursor glass, are plausible candidates. This study examines the photocatalytic behaviour of porous GCs obtained after acid leaching of MgO–TiO2–P2O5 GCs. After heat-treatment, anatase TiO2 was precipitated along with other phases. The diffraction intensity ratio between anatase and other phases was the maximum for a heat-treatment temperature of 900 °C. After acid leaching of the GCs, the relative surface area decreased with increasing TiO2 fraction; the surface area was also affected by the sample morphology. H2 generation was observed from porous GCs, while GCs without etching exhibited approximately zero activity. Thus, it was demonstrated that high surface area and prevention of the reduction reaction to Ti(III) are important for tailoring monolithic photocatalytic materials.

www.nature.com/scientificreports/ the volume fraction of TiO 2 -precipitated GCs with TiO 2 precipitated as a single phase [29][30][31] is not sufficient to make a porous skeleton network. Hosono et al. reported several functional crystallites, such as anatase TiO 2 , NASICON-type CaTi 4 (PO 3 ) 4 22 and CaTi 4 (PO 4 ) 6 23 from Na-doped CaO-TiO 2 -P 2 O 5 glasses 19 . In these reports, various applications using the porous structure and the constituent cations were discussed 19 . Although some TiO 2 -SiO 2 porous materials were obtained 23,24 , these multi-component materials were either phase-separated or not amorphous after quenching, and, therefore, were not suitable for the design of TiO 2 precipitation. On the contrary, as Ca 2+ and Mg 2+ play a similar role in the vitrification of glass, it is expected that TiO 2 -precipitated GCs can also be obtained from MgO-TiO 2 -P 2 O 5 (MTP) ternary glasses. However, there are relatively few reports on ternary MTP glasses 32 . MTP glasses have been prepared with the glass forming region, and their density and the molar volume have been reported by Kishioka 32 . Although it was reported that the minimum P 2 O 5 and maximum TiO 2 fractions in the MTP glasses without precipitation of the crystallites were 30 mol.% and 35 mol.%, respectively, there was no detailed information about the precipitated phase and the thermal stability 32 . As vitrification of glass using a melt-quenching method depends on its mode of preparation, it is expected that broader chemical compositions of MTP glass can be obtained by tuning the quenching process.
In the present study, the focus was on porous TiO 2 -containing GCs using the MTP glass system via the GC route. In addition to conventional X-ray diffraction (XRD) and scanning electron microscope (SEM) technique, the structural change was examined, depending on chemical composition, using X-ray absorption fine structure (XAFS), 31 P magic angle spinning (MAS) NMR, and elastic modulus measurements. For establishing the guidelines for photocatalytic hydrogen generation of monolithic materials, physical parameters and structure of porous GCs have been examined using a combination of various analytical methods in addition to photocatalytic activities. Figure 1 shows the differential thermal analysis (DTA) curves of (70-x) MgO-xTiO 2 -30P 2 O 5 (MTPx) glasses. These DTA curves were similar and independent of the TiO 2 fraction. The temperature of glass transition, T g , of these glasses was approximately 640 °C, whose error bars were comparable to those of the DTA measurements. The temperatures of crystallization onset, T x and the crystallization peak T p are also shown in Fig. 1. Considering the thermal stability of the glasses was estimated using the ΔT (= T x − T g ) values 33 , the MTP40 glass was highly unstable against crystallization. However, as the T g and the T p were similar, it was assumed that same heat-temperature was suitable for comparison of each sample.

Results and discussion
The obtained MTPx glasses were brownish, as the colouration evolved with an increase in the TiO 2 fraction. Figure 2 shows the optical absorption spectra of the MTPx glasses containing different TiO 2 fractions. The optical absorption edge existed at approximately 3.3 eV, corresponding to the band gap of anatase 34 . In addition, there was an optical absorption band at 2.5 eV, due to the Ti 3+ species 35 . The absorption coefficient linearly increased with increasing TiO 2 fraction. Therefore, it was expected that the Ti 3+ :(Ti 3+ + Ti 4+ ) ratio would be approximately constant and independent of the chemical composition.
To examine the valence state of the Ti cation, X-ray absorption near edge structure (XANES) was measured. Figure 3 shows the Ti K-edge XANES spectra of the MTPx glasses with Ti 2 O 3 , anatase and rutile. Compared with these references, it was assumed that the main valence state of Ti was tetravalent and was independent of the TiO 2 fraction. Although the existence of Ti 3+ was confirmed by the optical absorption, the concentration was not high enough to appear in the K-edge XANES spectra. It was notable that the spectral shapes of Ti in the MTPx glasses were similar to those of anatase. The similarity to anatase could be understood from the dip at 4.99 keV, although rutile TiO 2 was used as a starting material. Thus, in the present MTP glasses, TiO 2 existed with partial characteristics of anatase. Recently, it was reported that crystallization from a glass was affected by the local coordination of the mother glass, even though the chemical composition was not the stoichiometric www.nature.com/scientificreports/ chemical composition of a crystal 36 . As discussed below, it was found that the coordination of Ti in the mother glass affected the precipitated phase in the GCs. Figure 4a shows the Brillouin shifts ν B of the MTPx glasses obtained by fitting the Brillouin peaks, as shown in the inset. The density and theν B increased with increasing TiO 2 fraction. On the contrary, the longitudinal sound velocity v L and the longitudinal elastic modulus c 11 indicated that MTP40 was the inflection point as shown in Fig. 4b. These parameters are summarized in Table 1. In the binary ZnO-P 2 O 5 glass, the elastic modulus of the Zn-rich glass was higher than that of the ZnO-poor glass owing to the closed packed network consisting of ZnO x polyhedral 37 . As TiO 2 also belongs to the intermediate group in glass science, and can be assumed as either glass network-forming or network-modifying 38 , it was expected that the TiO 2 network consisting of TiO 6 anatase-like structure would be formed in TiO 2 -rich glasses. Although the first coordination states of Ti 4+ cation are similar in these glasses, as shown in Fig. 3, the inflection point at the MTP40 suggested that the network structure had been changed. Therefore, the network consisting of TiO 6 polyhedral and PO 4 tetrahedron are discussed for PO 4 units using 31 P MAS NMR measurement. Figure 5 shows the 31 P MAS NMR spectra of the MTPx glasses. In their study, the spectra were decomposed into different building units, denoted by Q n . These described the number of oxygen atoms (n) of the PO 4 tetrahedral interlinked to other cations. Although P-O-Ti bonds were observed in the present NMR spectra, for simplicity, these observed peaks were assumed to have only covalent P-O-based bonds based on a previous study 39 . It  www.nature.com/scientificreports/ is notable that the peak widths were constant, indicating that PO 4 and TiO 6 units were homogenously dispersed at the first and second coordination. According to previous studies [39][40][41] , MTPx glasses can be decomposed with individual signal components: Q 1 and Q 2 , indicated with dashed curves. It was assumed that the slight chemical shift was due to the increase in the number density of the Ti-O-P bond, whose shielding was smaller than that of the P-O-P bonds. Therefore, no remarkable structural change in the PO 4 units could be observed from the P MAS NMR spectra 31 . Figure 6a shows the XRD patterns of powdered MTPx GCs heat-treated at 900 °C for 3 h with several references: TiO 2 anatase, Mg 2 P 2 O 7, MgTi 4 (PO 4 ) 6 and Mg 0.5 (TiO) (PO 4 ). The obtained diffraction patterns showed that the anatase TiO 2 crystallites co-precipitated with the other phases in all the GCs, and the precipitation of TiO 2 as a single phase was not observed. It is often observed that metastable crystalline phase is precipitated after heat-treatment. Similar to previous results 28,29 , metastable anatase was precipitated from the present glass system. In addition, such co-precipitation of TiO 2 and other phases was observed in the Na 2 O-doped CaO-TiO 2 -P 2 O 5 GC systems 19 . To examine the relationship between the TiO 2 precipitation and the heat-treatment temperature, the intensity ratio using the highest diffraction intensity was used for the evaluation. Figure 6b shows the TiO 2 precipitation ratios, which were the relative intensities to the maximum diffraction intensities of Mg 2 P 2 O 7 or   6 , as a function of the heat-treatment temperature. The anatase TiO 2 ratio achieved the maximum with heat-treatment at 900 °C, suggesting that an optimized heat-treatment temperature exists. The average particle diameters estimated from the Scherrer equation 42 were approximately 27 nm.
To obtain the porous GCs containing TiO 2 crystallites, acid leaching of MTPx GCs heat-treated at 900 °C for 3 h using 1 M HNO 3 was performed at 90 °C for 4 d. The results are summarized in Table 2. After acid leaching, the weight losses of MTP35, MTP40 and MTP45 glasses were calculated as 42 ± 5%, 31 ± 3%, and 8 ± 6%, respectively. This indicated that the chemical durability against acid leaching was enhanced with an increase in the TiO 2 fraction. In addition, results of the eluted cations in the etched solution agreed with the results of the weight loss. This indicated that the TiO 2 -rich glass possessed a high chemical durability against acid solution. This narrowly correlated with a speculation based on the results of the elastic modulus (see Fig. 4). Figure 7a shows the XRD patterns of the MTPx porous GCs obtained after heat-treatment at 900 ºC for 3 h and their acid leaching. In acid-treated samples, the relative intensity of anatase increased compared with other phases, especially in the MTP35 system. Figure 7b shows the relative diffraction intensity ratio I TiO2 :I MgTi4(PO4)6 as a function of TiO 2 fraction. In the MTP35 system, the relative intensity of anatase increased compared with other systems. Crystallization of glass is a kind of thermally stabilization behavior of glass melt above the T g , and the precipitated crystalline phase and the residual amorphous region depend on the chemical composition. In the case of Mg-rich glass, it is expected that TiO 2 can be precipitated easily because of better stability of residual Mg-rich phosphate amorphous region, which was easily removed by acid etching. Considering the results of weight loss shown in Table 2, it was assumed that several precipitated crystallites were removed due to acid leaching in the MTP35 system. Figure 8 shows the volumes of adsorbed N 2 at standard temperature and pressure of MTPx porous GCs as a function of relative pressure. To understand the hysteresis shape, the vertical axis was plotted on a logarithmic scale. The hysteresis curves indicate the connected pores in the etched samples. Although the shapes of hysteresis curves of MTP35 and MTP40 were similar, the shape of MTP45 differed from the others. Additionally, the relative surface areas of porous MTPx GCs calculated from the Brunauer, Emmett and Teller (BET) method are shown in Table 2. The surface area decreased with increasing TiO 2 fraction, and there was a large difference between the MTP40 and MTP45 porous GCs.
Ti K-edge XANES spectra of the MTP40 glass, the GC and the porous GCs after acid leaching are shown in Fig. 9. In addition, the spectrum of anatase is shown for comparison. Compared with the mother glasses, characteristic absorption of anatase was clearly observed in the GC and the porous GC. Considering structural www.nature.com/scientificreports/ similarity, the anatase-like local structure in the glass network may have been the origin of the precipitation of the anatase TiO 2 phase. For discussion, the morphology of porous ceramics was checked after leaching by taking SEM images of the outer and inner surfaces. Figure 10 shows the SEM image of the MTPx porous GCs obtained after HNO 3 (aq.) leaching of the GC heat-treated at 900 °C for 3 h. At the surface of the porous sample, dendrite-like structures were observed. Meanwhile, pillar-like structures were observed inside the MTP40 and the MTP45 porous  www.nature.com/scientificreports/   www.nature.com/scientificreports/ ceramics, which was different from those of the MTP35 ceramics. Considering the values of c 11 , a change in the network structure affected the morphologies of the precipitated crystalline phases. Based on these results, the hydrogen evolution of the TiO 2 GCs was examined using sacrificial molecules as electron donors for improvement of H 2 production 43 . Figure 11 shows the hydrogen evolution rate from TiO 2 GCs with and without acid leaching. Here, the H 2 evolution rate per irradiated areas of the bulk samples (μmol⋅h −1 ⋅cm −2 ) are used to compare the photocatalytic activities of these monoliths. In the GCs without etching, the H 2 evolution was approximately zero, independent of the chemical composition, indicating that the surface of GCs was covered with an amorphous region of low photocatalytic activity. To the contrary, the porous GCs exhibited H 2 evolution activity without metal deposition [44][45][46][47] , with a maximum obtained with the MTP40 porous GCs. It should be noted that the relative surface areas of the porous GCs increased with decreasing TiO 2 fraction ( Table 2). The XRD patterns suggested that the relative intensity of precipitated anatase increased with decreasing TiO 2 fraction. Therefore, from these results, it could be assumed that the MTP35 GCs should exhibit the highest photocatalytic activity. However, the result without metal deposition did not correspond to the expectation. It is notable that the sample surface turned blue in colour after the catalytic reaction, as shown in the inset of Fig. 11. As it was expected that the reduction in Ti(IV) to Ti (III) 47,48 generally inhibited catalytic performance, the generation of Ti(III) species should have been prevented, although an optimized Ti(III):Ti(IV) ratio is proposed 49 . In addition, to prevent the reduction reaction of Ti(IV), photocatalytic activity following Pt nanoparticle deposition was examined according to previous reports 45,46 . Notably, the photocatalytic activity of MTP35 was highly enhanced due to Pt deposition, as shown in Fig. 12. It was suggested that the H 2 evolution rates broadly depended on the relative surface area. Thus, it was concluded that relative high surface area and prevention of the reduction reaction to Ti(III) were the most important factors in tailoring monolithic photocatalytic materials.
Here, we have compared the H 2 generation rate of the present study with those of previous cases, as shown in Table 3. In Table 3, we only focused on the TiO 2 GCs materials that could be treated as monolithic plates on the water. The present MTP GC exhibited the best H 2 generation rate among the monolithic glass-ceramics that can be prepared to larger sizes. Although the present result was not optimized, there should be great potential for improving H 2 generation or environmental cleaning properties using GC routes.
To understand the hydrogen generation rates and the structure at the surface, a transmission electron microscopy (TEM) image was taken of the porous MTP35 GC after acid etching with Pt deposition. Figure 12a shows a cross-sectional bright TEM image of the porous MTP35 GC with Pt deposition. At the surface, there are small islands in addition to fibrous pillar structure. This is also confirmed in Fig. 10. The precipitation of anatase at the surface was also confirmed from the nanobeam electron diffraction pattern shown in Fig. 12b (the morphology on a larger scale is shown in the Supplemental Information). Because Pt nanoparticles are difficult to observe in the bright TEM image, a dark scanning TEM (STEM) image of the porous MTP35 GC was also taken, as shown in Fig. 12c. In the dark STEM image, heavy elements can be detected as bright spots. One can observe white spots both at the surface (at the left side of the figure) and at the fibrous pillar region. The energydispersive X-ray spectroscopy (EDX) profile of the white spot -circled region in Fig. 12c-is shown in Fig. 12d. The profile shows that Pt particles with diameters of approximately 5 nm are randomly dispersed. However, it was confirmed that fibrous pillar-like structures consist of several crystalline phases (see the Supplementary  Information). Therefore, the sparse structure at the surface, which consisted of anatase regions and fibrous pillars, affects the hydrogen generation. It has been reported that the morphology of the sample also contributes to the H 2 production, because the interior microporous channels provide an easy path for the electrons for the effective surface charge transfer 51 . In principle, photocatalytic activity increases with increasing surface area. If the surface areas of these samples were equal, the MTP45 with the highest TiO 2 fraction might exhibit the best performance. However, in the present case, there is an inverse relationship between the relative surface area of www.nature.com/scientificreports/ porous GCs and the TiO 2 fraction of the samples. Since the porous MTP35 GC has the highest surface area (see Fig. 8), we conclude that photocatalytic activity of the porous MTP GCs are dominated by the relative surface area. Because nanostructures at the surface are important to enhance the specific surface area, the glass-ceramic route combined with chemical etching is important to attain high photocatalytic activity. Although the present materials are not the optimized photocatalytic materials, it was demonstrated that monolithic bulk materials have the potential to be catalytic materials that can contribute to a sustainable society.  www.nature.com/scientificreports/ To summarize, photocatalytic activity has been demonstrated with and without Pt nanoparticle deposition. The Pt deposition is effective for enhancement of the catalytic activity as it prevents a reduction in TiO 2 . For application in sustainable energy conversion using porous TiO 2 -precipitated materials, tailoring the TiO 2 crystallites in addition to the skeleton network is required. However, as the present materials are monolithic and effective for surface tailoring using the vapour reaction, further improvement using surface treatment is required 52,53 . To obtain porous GCs with good catalytic activity, the precipitated morphology of crystallites along with the nature of residual amorphous phases should be considered during materials design. For materials design, an understanding of the nucleation and crystal growth processes in glass is necessary, and structural analysis combined with several different measurement techniques will be helpful in this regard. For better formability, this study emphasizes that porous GC will be a candidate for energy harvesting following further performance improvements.  54 . The obtained solids after calcination were then melted in a platinum crucible in an electric furnace at 1,300 °C for 30 min. The glass melt was quenched on a stainless steel plate at 180 °C, then annealed at the temperature of glass transition, T g , for 30 min. After mechanical polishing to obtain a mirrorlike surface, the glass sample was heat-treated on an alumina plate in an ambient atmosphere to obtain the corresponding GCs. The heat-treatment strategy consisted of three steps. The heating rate was 10 °C/min from room temperature (RT) to 30 °C below the target temperature, and then reduced to 1 °C/min to the target temperature. After heat treatment at the holding temperature for 5 h, the furnace was cooled down without temperature control.

Methods
preparation of porous Gcs. The GCs were leached in 1 M HNO 3 (50 mL) at 90 °C for 4 d without stirring. After leaching, the samples were rinsed with pure water at 90 °C for 2 h, then rinsed in an ultrasonic bath using pure water for approximately 10 min. After rinsing, the samples were heated at 150 °C for 2 h to obtain the porous GCs.
Analysis methods. The T g , crystallization onset, T x and the crystallization peak T p were measured using differential thermal analysis (DTA) operated at a heating rate of 10 °C/min using TG8120 (Rigaku). The densities were measured by applying the Archimedes method using water at RT. The refractive indices (error bars ± 0.0001) at 452 nm, 633 nm, and 832 nm were measured using a prism coupler (Metricon, NJ). The refractive indices at 532 nm were calculated using lambda-in-Cauchy fitting [A + B/(l 2 ) + C/(l 4 )] of these three values. Here, l was the wavelength, and A, B and C were the fitting parameters. The Brillouin shifts, ν B , of the glasses were measured using a high-resolution modification of a Sandercock-Fabry-Perot system 55 . The longitudinal sound velocity, V L , was calculated using the equation V L = ν B λ/2n 532 , where ν B , λ and n 532 , are the Brillouin shift, the wavelength of incident light (532 nm), and the refractive index at 532 nm, respectively. The c 11 values were calculated using the equation c 11 = ρV L 2 , where ρ is the density. The absorption spectra were measured with a spectrometer UV-4150 (Hitachi High-Tech). X-ray diffraction (XRD) using UltimaIV (Rigaku) was used for examining the precipitated phases.
The 31 P NMR of the precursor glasses and the GCs were measured using DELTA (JEOL) under 14.1 T. A frequency of 161.80 MHz, a spin rate of 10 kHz and a pulse delay of 5 s were used in the measurements. The chemical shifts were estimated with respect to 85% H 3 PO 4 aqua solution (0 ppm) and the conventional notation for phosphorus sites, Q n , was used for the analysis. The n value denotes the number of bridging oxygen atoms per PO 4 tetrahedron.
The Ti K-edge (4.98 keV) XANES spectra were measured at the BL01B1 and BL14B2 beamlines of the SPring-8 synchrotron radiation facility (Hyogo, Japan). The measurements were performed using a Si (111) double-crystal monochromator in the transmission mode (Quick Scan method) at RT. Pellet samples for the measurements were prepared by mixing the granular sample with boron nitride. The corresponding analyses were performed using Athena software 56 .
To examine the durability against acid solutions, a sample weight was measured before and after 1 N HNO 3 leaching. In addition, Inductively Coupled Plasma-Atomic Emission Spectrometry (ICP-AES) measurement of the leaching solution was performed using SPS7800 (Hitachi High-Tech) to check the eluted composition. The surface areas of the samples were measured using NOVA3200 (Quantachrome). The morphology of the GCs 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 TEM images and perform EDX with an acceleration voltage of 200 kV.

photocatalytic hydrogen evolution from methanol aqueous solution. A gas chromatography vial
equipped with an open-top screw cap bottle sealed with a butyl rubber septum (SVG-12, Nichiden-Rika Glass Co. Ltd., internal volume of 15.6 mL) was employed as the reaction vessel. A GC sample was placed at the bottom, and a 50 vol.% aqueous solution of methanol (5 mL) was added. Nitrogen gas was bubbled into the solution for 30 min to remove dissolved oxygen. The reactor was irradiated from bottom with a 100 W high pressure Hg lamp (HL100G, SEN Lights Corp.) and cooled using a fan. The irradiation intensity was 342 mW/cm 2 . The headspace gas was sampled (0.2 mL) with a gas-tight syringe at 20 min intervals and analysed using a TCD gas chromatograph (GC320, GL Sciences Co., Ltd.) equipped with a MS5A column (2 m) of flowing Ar as the carrier gas 57  www.nature.com/scientificreports/ platinum deposition on the surface of glass-ceramics samples. Spontaneous reduction of platinum ions on the surface of the glass-ceramics sample was performed in accordance with the reported method 45,46 using the sample with oxygen vacancies obtained after photocatalytic hydrogen evolution. After the photocatalytic reaction, the bluish-gray glass-ceramics sample was washed with distilled water several times. A H 2 PtCl 6 (10 mM) aqueous solution of 0.31 mL was added to 5.0 mL of methanol (50 vol. %) aqueous solution.
The glass-ceramics sample with oxygen vacancies was added to the solution, and was left to stand for 15 min with occasional stirring. The resulting sample with Pt deposition was washed, and used again for photocatalytic H 2 evolution.

Data availability
The datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request.