All chemicals used in the synthesis and reaction testing are sourced as specified and used as-purchased without further purification. They include tetraethyl orthosilicate (TEOS, 99%, Xilong Science Co., Ltd.) as the silicon source, tetrabutyl titanate (TBOT, AR, Shanghai Aladdin Biochemical Technology Co., Ltd.) as the titanium source, TPAOH (25 wt%, aqueous solution, Shanghai Aladdin Biochemical Technology Co., Ltd.) as the template and base, Sesbania powder (98 wt%, Zhongtuan Biotechnology Co., Ltd.) as a pore-forming agent, silica sol (30 wt%, Qingdao Ocean Chemical Co., Ltd.) as an agglomerant, paraffin (99.8 wt%, aqueous solution, Qingdao Ocean Chemical Co., Ltd.), H₂O₂ (30 wt%, Beijing Chemical Works), methanol (99%, Sinopharm Chemical Reagents Co., Ltd.), and allyl alcohol (98%, Shanghai Aladdin Biochemical Technology Co., Ltd.).

In this study, nanosized TS-1 was synthesized in a TPAOH hydrothermal system following the reported procedure by Wang et al. (2001) In a typical synthesis, TEOS and TBOT were firstly hydrolysed separately and then mixed in a flask. The mixture was then heated at 90 °C for 40 min to remove the alcohols and form a gel with the molar composition of SiO₂: 0.025 TiO₂: 0.33 TPAOH: 40 H₂O. The gel was transferred to a Teflon-lined autoclave and crystallize at 170 °C for 72 h. The resulted solid suspension was separated by centrifugation, washed with distilled water, dried at 80 °C for 12 h, and calcined at 540 °C for 6 h to produce the TS-1 powder.

The above-synthesized powder was then extruded following a reported method (Li et al., 2002). Firstly, the TS-1 powder was mixed with Sesbania powder, silica sol, and paraffin in sequence to form a wet mixture with weight composition of TS-1: Sesbania powder: SiO₂: paraffin = 1: 0.05: 0.30: 0.03. It was then extruded into cylindrical rods of 2 mm diameter, dried at room temperature, calcined at 540 °C for 6 h, and finally cut into 2 mm × 2 mm pellets. This extruded catalyst is denoted as TS-1-Null.

The TPAOH post-treatment was conducted in a Teflon-lined autoclave, with TS-1-Null incubated in a dilute TPAOH solution (1 g: 10 mL) at 170 °C for a specified time.After the treatment, the solids were recovered, washed and calcined at 540 °C for 6 h. The treated extrudates are denoted as TS-1-C-t. Here C refers to the molar concentration of TPAOH in the post-treatment solution, which was varied from 0.04 to 0.10 mol/L, but denoted as C = 4 to 10, and t refers to the post-treatment time (h). For instance, TS-1-TPAOH-6-48 indicates that extruded TS-1 was treated in a 0.06 mol/L TPAOH solution for 48 h. Samples with the suffix “D” represent the spent catalysts after reaction, while those with the suffix “R” represent the spent catalysts regenerated by calcination at 540 °C for 6 h.

X-ray powder diffraction (XRD) patterns were obtained on a Rigaku Corporation SmartLab 9 X-ray diffractometer using Cu Kα radiation. The relative crystallinity (RC) was calculated by comparing the total intensity of the five characteristic MFI diffraction peaks to that of TS-1-Null, which is taken as 100%. Fourier transform infrared (FTIR) spectra were recorded on a Bruker EQUINOX55 spectrometer from 4000 to 400 cm^-1, and via a stainless steel vacuum chamber facing the spectrometer through KBr windows. Ultraviolet resonance Raman (UV-Raman) spectra were obtained using the DL-1 UV-Raman spectrometer at 244, 266 and 320 nm excitation laser lines, respectively. Ultraviolet-visible diffuse reflectance (UV/vis) spectra with wavelengths from 190 to 500 nm were obtained on a Jasco UV-550 spectrometer, using pure BaSO₄ as a reference. The fractions of Ti species were determined by the deconvolution of UV/vis spectra using Gaussian fitting of the PeakFit program. Note that, to ensure the consistency of deconvolution, the bands of tetra-, penta- and octa-coordinated Ti and anatase TiO₂ were fixed at 216, 238, 277, and 320 nm, respectively. Inductive Couple Plasma-optical emission spectroscopy (ICP-OES) analysis was carried out on a Perkin-Elmer OPTIMA 2000DV ICP optical emission spectrometer. X-ray photoelectron spectra (XPS) were acquired with a Thermo VG ESCALAB250 instrument using Al Kα radiation and operating at a constant power of 260 W. The obtained spectra were simulated using the XPSPEAK program to determine the relative contents of differently coordinated Ti species. Nitrogen physisorption measurements were performed at liquid nitrogen temperature (−196 °C) on a Quadrasorb SI physical sorption apparatus. Before the measurements, the samples were degassed at 300 °C for 6 h. The total surface area and pore volume were calculated according to the Brunauer-Emmett-Teller (BET) and t-plot methods, respectively. A Hitachi S-4800 scanning electron microscope (SEM) and a Tecnai G220 S-Twin transmission electron microscope (TEM) were used to observe the microstructure of the TS-1 samples. Solid-state ²⁹Si magic-angle-spinning nuclear magnetic resonance (MAS NMR) analysis was carried out to ascertain the local bonding of Si in the TS-1 samples. MAS NMR spectra were obtained on a Bruker 500 WB nuclear magnetic resonance spectrometer at ambient temperature, using tetramethyl silane (TMS) as an external standard. The chemical shift was referred to as an external standard of tetramethyl silane (TMS). The spin rate of the rotor was set at 4.0 kHz, and a typical π/4 pulse length of 2.73 µs was adopted for ²⁹Si resonance. The spectra were deconvoluted using a Gaussian–Lorentzian mixed model.

The epoxidation of allyl alcohol in a fixed-bed reactor was employed to evaluate the catalytic performance of the TS-1 samples. In a typical reaction test, 3.5 g of extrudate catalyst was placed in the middle zone of a stainless steel tube reactor with an inner diameter of 8 mm and packed with carborundum (SiC) balls below and above the catalyst. The system pressure of 1.0 MPa was maintained by nitrogen. Aqueous H₂O₂ (30 wt%) was mixed with methanol to form a 1.0 mol/L solution. The methanol was the solvent, the molar ratio of allyl alcohol/H₂O₂ was 3:1, and the weight hourly space velocity (WHSV) of allyl alcohol was set at 0.5 h^-1. The flow rates of allyl alcohol and H₂O₂/CH₃OH were 2.1 mL/h and 10 mL/h, respectively controlled by two separate pumps, were fed to the reactor. The reaction was then carried out at 60 °C. The product effluent from the tube reactor was cooled in a storage tank placed inside a refrigerator. The sampling was done at a time interval of 4 h.

The initial and residual concentration of H₂O₂ was determined with iodometric titration, and the products were analyzed by an Agilent 7890B gas chromatograph with a flame ionisable detector (FID) and a capillary column (HP-INNOWAX, 50 m × 0.25 mm × 0.5 µm). The conversion of hydrogen peroxide (X(H₂O₂)), conversion of allyl alcohol (X(AA)), selectivity of glycidol (S(GDL)), yield of glycidol (Y(GDL)), and utilization of hydrogen peroxide (U(H₂O₂)) were calculated with the listed Equations 1–5, respectively. X H 2 O 2 = n 0 H 2 O 2 − n H 2 O 2 /   n 0 H 2 O 2 (1) X AA = n GDL + n Others /   n AA + n GDL + n Others (2) S GDL = n GDL   /   n GDL + n Others (3) Y GDL = X AA × S GDL (4) U H 2 O 2 = 3   X AA   /   X H 2 O 2 (5)

The n₀(H₂O₂) and n(H₂O₂) stand for the initial and final mole numbers of H₂O₂, respectively. The n(AA) and n(GDL) are the molar number of AA and GDL, respectively. The n(Others) represents the overall mole number of the by-products.

Figure 1a depicts the XRD patterns of the TS-1 samples before and after TPAOH treatment, as well as the RC data. The five characteristic diffraction peaks corresponding to the MFI topology (2θ = 7.8°, 8.8°, 23.0°, 23.9°, 24.4°) are observed in each sample; however, the intensity of these peaks varies with the treatment conditions, which is mirrored in the significant changes in the RC values of the treated samples. When the concentration of TPAOH increases from 0.04 to 0.10 mol/L, the RC increases first and then decreases slightly, and has a maximum at 0.08 mol/L. When the concentration of TPAOH is 0.06 mol/L, the RC increases gradually with the prolonged treating time.

Treatment with a base causes silica dissolution, Wang and Tuel (2008) and the dissolved silica can undergo recrystallization when a template such as TPAOH is present (Wang et al., 2007; Jiao et al., 2020). There is an equilibrium between these two processes—dissolution and recrystallization. Specifically, the crystallinity will decrease if dissolution occurs faster than recrystallization, and it will increase in the opposite case. Thus, the reason for the decreased RC after being treated with TPAOH is attributed to the relative rates of dissolution and recrystallization (Yang et al., 2023; Wang et al., 2007).

The Fourier transform infrared spectroscopy (FTIR) spectrum of TS-1 sample is shown in Figure 1. Five characteristic FTIR bands associated with the MFI topology of TS-1 are observed at λ = 450, 550, 800, 1074 and 1226 cm^-1. However, the intensity ratio of I₅₅₀/I₄₅₀ of the samples, which indicates the crystallinity, Xue et al. (2014) appears rather different. When the TPAOH concentration increases from 0.04 to 0.10 mol/L, the value of I₅₅₀/I₄₅₀ first decreases marginally and then increases slightly, and reaches its highest at 0.10 mol/L. On the other hand, a longer treating time leads to a slightly larger I₅₅₀/I₄₅₀. These agree with the findings of the XRD patterns (Figure 1a).

In the FTIR spectrum of TS-1, a characteristic band is usually present at 960 cm^-1; most researchers believe this band corresponds to the stretching vibration of the Si-O bond affected by neighboring titanium ions (Yang et al., 2023; Tozzola et al., 1998). In other words, the presence of this band can be regarded as a fingerprint for titanium incorporation into the MFI framework (Zhang et al., 2016). The spectra indicate that all samples contained Ti^IV. The ratio of the relative intensities of the 960 cm^-1 and 800 cm^-1 bands (I₉₆₀/I₈₀₀) is frequently utilized to quantify the relative content of Ti^IV (Yang et al., 2023; Zuo et al., 2015). As shown in Table 1, the I₉₆₀/I₈₀₀ values of the treated TS-1 samples are slightly lower than those of the pristine TS-1-Null. This can be rationalized by the partial breakdown of the TS-1 framework, leading to the cleavage of Ti-O-Si bonds and the consequent loss of Ti^IV.

Figure 2 shows the UV-Raman spectra of the samples taken at the excitation wavelength (λ_ex) of 244, 266 and 320 nm. Figure 2a shows the spectra excited at 244 nm. The peaks at 380 and 800 cm^-1 associated with the MFI topology are observed in all the samples. The peak at 1125 cm^-1 in the samples is attributed to the Ti-O-Si bond, Jin et al. (2014), Li et al. (2001), Zhang et al. (2001), Wang Y. et al. (2023) and the simultaneous appearance of three peaks at 490, 530, and 1125 cm^-1 in the 244 nm-excited UV-Raman spectrum corroborates the presence of tetrahedrally coordinated Ti(OSi)₄ sites in TS-1, each comprising four Ti-O-Si bonds (Wang Y. et al., 2023; Hao et al., 2025; Song et al., 2015). Table 1 also lists the intensity ratio of I₁₁₂₅/I₃₈₀, which indicates that the samples have different amounts of Ti(OSi)₄. The variation trend of I₁₁₂₅/I₃₈₀ is the same with that of I₉₆₀/I₈₀₀.

Figure 2b is the UV-Raman spectra of the samples excited at 266 nm. The observed band at 510 cm^-1 in the spectra for the TPAOH-treated samples are ascribed to the presence of a pentahedrally coordinated Ti (Ti(OSi)₃(OH)₂, Ti-V) species (Wang Y. et al., 2023). Furthermore, the peak at 700 cm^-1 suggests the existence of Ti-OH group from the octahedrally coordinated Ti namely TiO₆ (Ti(OSi)₂(OH)₂(H₂O)₂) or Ti-VI (Ti(OSi)₂(OH)₄) (Wang Y. et al., 2023; Yao et al., 2023). It has been reported that Ti-V and Ti-VI species demonstrate turnover frequencies (TOFs) for propylene epoxidation that are significantly higher than those of “Ti(OSi)₄” species (Wang Y. et al., 2023). In contrast, no peak at 700 cm^-1 was detected in the UV-Raman spectrum of TS-1-Null, suggesting the absence of Ti-OH groups.

Figure 2C presents the UV-Raman spectra of the samples excited at 320 nm. The bands at 144, 390, 520, and 637 cm^-1 are observed in the spectra, which serves as direct evidence for the existence of anatase TiO₂ (Zhang et al., 2016; Li et al., 1999).

Figure 3 presents the UV/vis spectra of the TS-1, which exhibit three prominent absorption bands centered at 210-220, 240-290 and 310–330 nm. The band at 210–220 nm arises from the electron transition of O 2p to Ti⁴⁺ 3d orbitals, and this absorption feature indicates the presence of tetrahedrally coordinated Ti species (Jorda et al., 1997; Perego et al., 2001; Zuo et al., 2020). The band at 240–290 nm is is attributed to octahedrally coordinated Ti species (Gordon et al., 2020). The band at 310–330 nm is assigned to anatase TiO₂ (Perego et al., 2001; Zuo et al., 2020). The multibands observed in the UV/vis spectra were deconvoluted using the Gaussian fitting method within the PeakFit program, guided by the identification of Ti species from UV-Raman spectroscopy. As seen in Figure 3, there are three bands in TS-1-Null and four bands in the different concentrated TPAOH-treated samples. The relative contents of Ti species are thus determined by the deconvoluted peak areas.

The contents of different coordinated Ti are provided in Table 2 (Columns 8-11). After treatment, the content of Ti^IV decreases significantly, but it increases again as the treatment time prolonged. With the increase of TPAOH concentration, the content of Ti^IV first increases and then decreases. The content of Ti^V shows an opposite trend. The content of Ti^VI increases slightly after the treatment, but it change unclearly with the treating time and TPAOH concentration. These phenomena indicates the dynamic balance of Ti species during the treatment. The dissolution of silicon in the treatment causes the generation of Ti-OH and thus the Ti^V and Ti^VI species, while the recrystallization leads to the condensation of Ti-OH and Si-OH, which enhance the generation of Ti^IV.

Table 2 also presents the elemental composiyion of the TS-1 samples, including both bulk (from ICP-OES) and surface (from XPS) measurements, as well as the Si/Ti molar ratios in the bulk (n_B(Si/Ti)) and on the surface (n_S(Si/Ti)). As seen, n_B(Si/Ti) of the TPAOH-treated samples are all lower than that of TS-1-Null, due to the loss of Si during the treatment. In contrast, after the TPAOH treatment, n_S(Si/Ti) of the samples increases noticeably compared to TS-1-Null. This increase is mostly due to the surface enrichment of Si by the recrystallization of the dissolved Si from the interior of TS-1 and the SiO₂ agglomerant. Furthermore, the n_S(Si/Ti) values of TPAOH-treated samples increase with the prolonged treating time, indicating that more recrystallization occurred.

Figure 4a depicts the Ti 2p XPS spectra of the samples. As shown, the Ti 2p_3/2 peaks can be deconvoluted into three component peaks. The peaks at 460.3, 459.3 and 458.4 eV are attributed to the Ti^IV, Ti^VI and anatase TiO₂, respectively (Lassaletta et al., 1995; Contarini et al., 2002; Vetter et al., 1994; Vogel et al., 2015). The content of Ti^IV on the surface does not change much, while Ti^VI slightly increases with the extension of treating time. When the concentration of TPAOH increases, the content of Ti^IV first decreases and then increases, while that of Ti^VI changes little. This indicates that the concentration of TPAOH can affect the dissolution and recrystallization rates of Si and Ti. An increased concentration initially accelerates the dissolution rate, but further increasing the concentration enhances the structure-directing effect, promoting the recrystallization. Figure 4b shows the ²⁹Si MAS NMR spectra of post-treated TS-1 samples. As seen, the spectra of all the samples are nearly the same. In other words, no significant difference in the silicon state of the TS-1 catalyst was found under varying conditions (TPAOH concentration or treatment time).

Figure 5 shows the nitrogen physisorption isotherms of the samples. A typical type I isotherm is presented for the samples. The presence of the H4 hysteresis loop according to IUPAC classification, this feature is observed in each isotherm of the treated samples owing to the formation of cavities within the materials. However, the hysteresis loop in the treated samples is quite small because the dissolution of silica agglomerant consumes the OH⁻, leading to the decrease of the dissolution of framework Si in crystals.

The surface areas and pore volumes of the treated samples, calculated from the nitrogen physisorption-desorption isotherms, are summarized in Table 3. Both the total surface area (S_BET) and the microporous surface area (S_micro) exhibit a trend of first decreasing and then increasing as the TPAOH concentration increases. The same trend was observed in the samples subjected to prolonged treatment time. However, the external surface area shows an opposite trend. This indicates that in the initial stage of treatment or at a low TPAOH concentration, the dissolution rate of Si is faster than the recrystallization rate, resulting in the destruction of micropores and the formation of cavities. After a certain period of treatment, the content of Si monomers in the liquid phase increases, promoting the recrystallization process. Under the structure-directing effect of TPAOH, the content of micropores increases and the specific surface area accordingly increases. For pore volume, the total pore volume (V_T) and mesoporous volume (V_meso) vary with the treating conditions obviously. This is because the intercrystalline space formed during the extrusion are mainly mesopores and macropores, and the TPAOH treatment causes the silica agglomerant to dissolve and recrystallize, destroying these intercrystalline pores. Therefore, both the V_T and V_meso increase gradually with the prolonging of the treating time.

Figure 6 shows the TEM images of the samples. The TS-1 particles consist of nanosized grains in the size range of 150–200 nm. A considerable amount of silica was introduced during the extrusion process and deposited on the external surface of the TS-1 particles. The silica deposits disappeared following the treatment. The reason is that they were dissolved in the alkaline solution and crystallized on the surface of the TS-1 particles. Since the amorphous silica transforms to crystals, the crystallinity increases after the treatment, which is proved by the XRD (Figure 1). Furthermore, several irregular cavities are observed in all the treated samples, which corroborates the presence of hysteresis loops in the nitrogen physisorption-desorption isotherms. The Si monomers dissolved from the generation of cavities were recrystallized on the external surface of TS-1 particles, which increased the Si/Ti molar ratio on the surface (Table 2). These cavities can shorten the diffusion pathways of both reaction substrates and products, thereby enhancing accessibility to the active sites.

Figure 6 also shows the appearance of the TS-1 samples treated with different concentrated TPAOH solutions. Much silica particles can be seen in the SEM image of TS-1-Null, which is absent in the treated samples, indicating that these silica particles undergo crystallization during the treatment process. Meanwhile, the morphology and particle size of the treated samples exhibit no obvious changes with variations in the treatment conditions.

Figures 7, 8 show the catalytic performances of TS-1-Null and TPAOH-treated TS-1 for allyl alcohol epoxidation in the fixed-bed reactor. At the initial stage, each catalyst exhibited a high H₂O₂ conversion (X(H₂O₂)). For TS-1-Null, all the parameters, including the X(H₂O₂), allyl alcohol conversion (X(AA)), glycidol yield (Y(GDL)) and H₂O₂ utilization (U(H₂O₂)) decrease sharply as the reaction time is prolonged to 150 h (Figure 7), except for the glycidol selectivity (S(GDL)), which remains constant during the time on stream (Figure 7c).

After the treatment, the H₂O₂ and AA conversions, H₂O₂ utilization and glycidol yield are improved obviously. They first increase and then decrease with the increased TPAOH concentration from 0.04 to 0.10 mol/L, and have a maximum H₂O₂ conversion at 0.06 mol/L. However, the glycidol selectivity does not change with the variation of TPAOH concentration.

Figure 8 presents the catalytic performance of the TS-1 treated for different time. The glycidol selectivity does not change with prolonged treating time (Figure 8c). All the other parameters show a maximum when treating for 48 h.

The catalytic performance of TS-1 in the epoxidation of AA is affected by two factors, which are the Ti coordination state and the diffusion property. The former decides the intrinsic catalytic performance, while the latter affects the pathway of substrates to active centres. In terms of Ti coordination state, the Si content decreases while the Ti content increases after the treatment. Furthermore, part of Ti^IV species are transformed into highly activity Ti^V and Ti^VI species, which is beneficial for the epoxidation. For the diffusion property, TPAOH treatment leads to the crystallization of SiO₂ covering on the surface of TS-1 particles, releasing the blockage of the channels and improving the diffusion property. The H₂O₂ and AA conversions over the TS-1-TPAOH-6-48 are the highest, because it possesses the largest external surface area and mesoporous volume, which is more favorable for diffusion.

Figure 9 shows the TG-DTG curves and UV-Raman spectra of the spent catalyst and the regenerated catalyst. As seen in Figure 9a, the total weight loss of each sample is ∼15%, suggesting that there are some substrates leaving in the channels of TS-1. The weight loss of TS-1-TPAOH-6-48D is the lowest, although its operating time is the longest. The DTG curves (Figure 9b) show that the weight loss is located at 250 °C–300 °C, which is attributed to the desorption of by-products (methylin, dimethylin and trimethylin). As seen, the peak of TS-1-TPAOH-6-48D appears at the lowest temperature, indicating that the occurrence of side reactions over it is the least.

Figures 9c,d exhibit the UV-Raman spectra (excited at 266 nm) of the spent and regenerated TS-1, respectively. The intensity of the band at 700 cm^-1 weakens in each deactivated sample, while it recovers after regeneration. This implies that the by-products may adsorb on the octa-coordinated Ti species, and the regeneration leads to the desorption of by-products, recovering the octa-coordinated Ti species.

Therefore, the deactivation of TS-1 in the allylic alcohol epoxidation reaction can be attributed to two primary factors: (i) the blockage of reaction pathways by macromolecular byproducts, and (ii) the adsorption of byproducts onto the octa-coordinated Ti sites during the reaction.

In addition, the regenerated TS-1-NullR and TS-1-TPAOH-6-48R samples were evaluated under identical conditions to the fresh catalysts (Figure 10). For TS-1-NullR, the initial X(H₂O₂) and X(AA) showed a slight recovery compared to the deactivated catalyst, but remained lower than the initial activity of the fresh catalyst, and deactivation recurred after 40 h. In contrast, TS-1-TPAOH-6-48R exhibited an initial X(H₂O₂) comparable to the fresh catalyst, yet it also deactivated after 150 h. Furthermore, while the S(GDL) of the regenerated catalysts remained at 100%, their Y(GDL) and U(H₂O₂) did not recover to the initial levels of the fresh catalyst. These results indicate that micropore blockage during reaction is only partially responsible for deactivation, which is also attributable to irreversible changes in the Ti species.