Proton exchange membrane fuel cells (PEMFCs) have attracted significant attention due to their outstanding energy efficiency and high power density (Sun et al., 2019a). According to the operating temperature, PEMFCs are divided into low-temperature PEMFCs (LT-PEMFC) working below 80°C and high-temperature PEMFCs (HT-PEMFC) that work over 150°C (Søndergaard et al., 2018). In comparison with LT-PEMFC, the HT-PEMFC has advantages including improved resistance towards CO poisoning, fast electrode reaction kinetics and flexible fuel supply (Li et al., 2009). HT-PEMFCs are promising in a wide range of applications, including combined heat and power, uninterrupted power supply and cars based on fuel cells. One of the core materials of HT-PEMFCs is the high-temperature proton exchange membrane (HT-PEM) that requires high proton conductivity under high operating temperature and anhydrous conditions (Nambi Krishnan et al., 2019). The proton conductivity of the HT-PEM tends to increase with the introduction of highly ordered meso-channels. Jiang et al. (Aili et al., 2016; Zhang et al., 2017) significantly enhanced the proton conductivity of a phosphoric acid-doped polybenzimidazole membrane at 200°C by introducing mesoporous silica into the polymer membrane.

Compared to mesoporous silica, hollow mesoporous silica (HMS) has both meso-channels that accelerate proton transport and hollow voids for electrolyte storage, and has received significant attention in HT-PEMs (Zhang and Jiang, 2016; Zhang et al., 2019; Zhao et al., 2020a). Jiang et al. (Zhang et al., 2018) increased the proton conductivity of the pristine Nafion^® membrane at 140°C and low relative humidity (0.55%) by two orders of magnitude by adding a small amount of ionic liquid-impregnated HMS to the membrane. In addition, the introduction of amine-functionalized mesoporous silica into a phosphoric acid-doped polyethersulfone-polyvinylpyrrolidone polymer also increased the proton conductivity of the membrane at 180°C and anhydrous conditions by 24.6% (Zhang et al., 2016). However, the synthesis of HMS usually requires the hard template method, which not only leads to a complicated synthesis process, it is not also conducive to scaling up the production of HMS (Soltys et al., 2018).

Several new strategies have recently been developed for the fabrication of HMS (Qiu et al., 2019). Teng et al. (Teng et al., 2013) transformed mesoporous silica microspheres obtained from the Stöber method into HMS via a selective etching process, which is based on the principle that the condensation degree of the outer layer is higher in comparison to the core. Furthermore, the combination of mesoporous silica with multiple interfaces and a one-step hydrothermal treatment leads to the formation of multi-shell hollow mesoporous silica microspheres (Teng et al., 2015). Although the transformation procedure does not require any sacrificial template or surface protection agent, these methods either cause disorder of the mesoporous channels in the shell or require harsh reaction conditions (hydrothermal reaction, etc.) (Li et al., 2011). While only a few theories have been established for the formation mechanism of HMS, which includes surface protection etching, Ostwald ripening and cationic surfactant-assisted selective etching (Fang et al., 2011), the formation mechanism of HMS is rarely studied and verified through microscopy. Consequently, it is still a significant challenge to develop a facile and controllable method for the synthesis of HMS and verify the formation mechanism.

Herein, we demonstrated a novel solid-to-hollow transformation approach for the fabrication of HMS via an ion exchange-induced selective etching method. The formation mechanism of the HMS was verified by scanning transmission electron microscopy (STEM) combined with energy dispersive X-ray spectroscopy (EDS). In addition, the HMS was impregnated by phosphoric acid to fabricate an inorganic composite HT-PEM (PA/HMS). The proton conductivity of the as-synthesized composite membrane under high temperature conditions was investigated as a proof-to-concept for fuel cell applications. The outstanding proton conductivity of the PA/HMS composite membrane achieved in this study is promising for HT-PEMFCs.

The fabrication of the PA/HMS membrane is shown in Figure 1. First, mSiO₂ particles with a mesoporous structure were synthesized by a modified Stöber method. Then, the mSiO₂ spheres were transformed into HMS spheres by an ion exchange-induced selective etching method. PA was doped into HMS by vacuum impregnation method to achieve the PA/HMS composite.

The obtained mSiO₂ particles have a regular spherical shape with a uniform size distribution, as shown in Figure 2A. The average size of the particles is 281 ± 54 nm. In addition, the mSiO₂ particles show a solid structure with worm-like mesopores oriented radially outwards from the center of the particle (Figure 2B). However, after immersing the mSiO₂ in a 0.2 M Na₂CO₃ solution at 60°C for 10 h, the solid core of the particles was etched to a hollow structure (Figure 2C), whereas the edge of the particles remained intact with a worm-like mesoporous structure. In addition, the average diameter of the HMS particles was 276 ± 32 nm, which is close to the average size of the solid mSiO₂. The results demonstrate that it is only the core of the mSiO₂ that is etched instead of the edge of the particles. Furthermore, the HAADF-STEM images in Figure 2D show that various HMS particles have different thicknesses, indicating the extension of the etching process from the core to the outer layer of mSiO₂ (Hao et al., 2015). Moreover, the homogeneous distribution of O (Figure 2E) and Si (Figure 2F), in the whole hollow microsphere shell, confirms that the HMS was composed of SiO₂.

Figure 3A shows the SAXS curves of mSiO₂ and HMS particles. There is a broad peak in the SAXS curve of mSiO₂ centered at 0.133 Å, which corresponds to a d-spacing of 4.7 nm (Zhang et al., 2014). When mSiO₂ is transformed into HMS, a broad peak is also observed centered at 0.139 Å in the SAXS curves of HMS, corresponding to a d-spacing of 4.5 nm. The similarity in the d-spacing values of mSiO₂ and HMS indicates that the mesoporous structure of HMS is the same as that of mSiO₂. In addition, the specific surface areas and pore sizes of mSiO₂ and HMS were characterized by the N₂ isotherm curves, as shown in Figure 3B. The N₂ isotherm curves of mSiO₂ and HMS are both type IV curves, confirming the mesoporous structure of both mSiO₂ and HMS (Hao et al., 2015). The BET surface area of mSiO₂ was found to be 1,423 cm²g⁻¹. After the etching process, the BET surface area of HMS reduced to 896 cm²g⁻¹, which is 37% lower than that of the mSiO₂. This is due to the dissolution of the center of mSiO₂ (Li et al., 2015). In addition, the pore sizes of HMS and mSiO₂ are 3.2 and 3.4 nm, respectively, indicating that the mesoporous structure of mSiO₂ remained the same before and after etching. The results verify that only the inside of the solid mSiO₂ was dissolved during the formation of HMS via the solid-to-hollow structure transformation.

To observe the transformation process of mSiO₂ to HMS, the morphology of the mSiO₂ particles that were stirred in a 0.2 M Na₂CO₃ solution at 60°C, for different times, was observed, as shown in Figure 4. The mSiO₂ particles show a uniform solid structure with no void in their center. When it was reacted in a 0.2 M Na₂CO₃ solution at 60°C for 2 h, a void of size of 40 ± 16 nm was formed in the center of the mSiO₂ particle (Figure 4B). When the reaction time was extended to 15 h, the size of the void increased to 110 ± 30 nm (Figure 4C). In addition, the SAXS characterization was conducted on the mSiO₂ samples with different etching times (Figure 4D). In the etching time range of 0–20 h, a broad peak was observed in the SAXS curve of all the samples located at 0.135 Å. Similar q values for each sample indicate that the size of the mesoporous structure of mSiO₂ remained the same during the entire etching period. Consequently, on increasing the reaction time, the mSiO₂ was gradually etched from the inside to the outside, and finally, a hollow mesoporous silica microsphere, with a certain shell thickness, was formed.

As shown in Figure 5A, significant weight loss (33.1 wt%) occurred in mSiO₂ starting from 180°C, which is assigned to the decomposition of CTAB. When mSiO₂ was refluxed in an ethanol solution containing HCl for 3 h, the weight loss of the particles at temperatures above 180°C was only 5.2 wt%, indicating that CTAB was almost completely removed in mSiO₂-r. In addition, the spherical and mesoporous structure of mSiO₂-r is clearly observed (Figure 5B), indicating that the removal of CTAB in mSiO₂ did not affect the mesoporous structure of the particles. Generally, the mSiO₂ is formed by self-assembly and condensation of Si-OH on the surface of CTAB micelles which act as a soft template (Li et al., 2017; Sun et al., 2019b). In addition, solid SiO₂ spheres can also be transformed into HMS with the assistance of CTAB through the interaction and self-assembly of Si-OH that derives from the dissolution of solid SiO₂ and CTAB (Ghasemi et al., 2017; Su et al., 2019). Thereby, CTAB tends to play a critical role in the transformation of mSiO₂ to HMS in this case. To verify this inference, the mSiO₂-r was etched in the 0.2 M Na₂CO₃ solution at 60°C for 10 h. The spherical structure of mSiO₂-r transformed into an amorphous lamellar structure with an absence of the mesoporous structure (Figure 5C), and HMS was not obtained. Therefore, CTAB plays a critical role in the transformation of mSiO₂ particles to HMS in this case.

Figure 6 shows the HAADF-STEM and elemental mapping of mSiO₂ after immersion in the 0.2 M Na₂CO₃ solution at room temperature for 10 min. The mSiO₂ particle shows a solid structure, and the Si (Figure 6B) and O (Figure 6C) are shown to be evenly distributed in the particle. In addition, the uniform distributions of C (Figure 6D) and N (Figure 6E) in the entire mSiO₂ particle confirm the homogeneous distribution of CTAB in the mesoporous channels of mSiO₂. On the other hand, the intensity of the peak of Na in the center of the mSiO₂ particle was higher than that at the edge of the particle (Figure 6F). As Na⁺ ions are paired with OH⁻ ions, the EDS result implies that the OH⁻ ion concentration in the center of the mSiO₂ particle is higher than that of the OH⁻ ion at the edge of the particle. The OH⁻ ions tend to attack the Si-O-Si bond in SiO₂ and lead to the dissociation of Si-O-Si to Si-OH, resulting in the dissolution and etching of SiO₂(Khoeini et al., 2019). Therefore, an ion exchange-induced selective etching mechanism is proposed for the solid-to-hollow transformation of mSiO₂ to HMS. The mSiO₂ derived from the modified Stöber method shows uniform distribution of CTAB in the worm-like mesopores of mSiO₂ that are oriented from the center to the edge of the particle. When mSiO₂ was placed in a weakly alkaline Na₂CO₃ solution, Br⁻ ions in CTAB and OH⁻ ions in the solution were exchanged, leading to the accumulation of OH⁻ ions in the mesopores of mSiO₂. Thereby, the concentration of OH⁻ in the center of the particle was higher than the concentration of OH⁻ ions in the solution. In addition, the OH⁻ ions in the mesoporous channels on the edge of mSiO₂ could easily exchange with the OH⁻ ion in the solution under a dynamic equilibrium, leading to dilution of the OH⁻ ions in the mesoporous channels close to the surface of the mSiO₂ particle. In other words, the OH⁻ concentration in the mesoporous channels of the mSiO₂ particle gradually decreased from the center to the edge of the particle. Thereby, the center of the mSiO₂ was dissolved by the high concentration of the OH⁻ ion, leading to the formation of HMS. The void became larger in size along with the concentration gradient of OH⁻ ion.

The HMS is employed as the HT-PEM for HT-PEMFC. The HMS particles were hot-pressed into the inorganic membrane of diameter and thickness of 38 and 0.51 mm, respectively, as shown in Figure 7A. Then, the HMS membrane was doped with phosphoric acid to obtain a dense PA/HMS inorganic HT-PEM (Figure 7B). However, some pinholes in the PA/HMS membrane were found due to the existence of certain pores among the HMS particles, which may lead to higher gas permeability than that of pure polymer electrolyte membranes (Zhao et al., 2020b). In addition, the isothermal curve of PA/HMS in N₂ still shows a type IV curve, indicating the existence of mesoporous structure in PA/HMS. On the other hand, the BET specific surface area of PA/HMS substantially decreased to 29 m²g⁻¹ which is 98.0% lower than that of the pristine HMS particle. The drastically reduced BET surface area of HMS after PA doping demonstrates that a large number of phosphoric acid molecules entered the mesopores of HMS (Zeng et al., 2013). In addition, compared with the HMS that shows slight weight loss in the temperature range from 100–800°C, PA/HMS exhibited a significant weight loss (34.8 wt%) at temperatures above 100°C, with two peaks centered at 123°C and 205°C. The weight loss at 123°C is likely attributed to the volatilization of water molecules in the phosphoric acid solution, while the weight loss at 205°C is due to the dehydration of phosphoric acid molecules (Hoffmann et al., 2012). The abovementioned results indicate that a large number of phosphoric acid molecules have been successfully incorporated into HMS, which is beneficial to the proton conduction of the PA/HMS membrane.

The proton conductivity of the PA/HMS membrane is shown in Figure 7E. Typically, at 0% relative humidity (RH), the proton conductivity of a Nafion^®112 membrane tends to decrease from 1.5 × 10⁻² S.cm⁻¹ at 80°C to 2.9 × 10⁻⁴ S.cm⁻¹ at 180°C. This is due to the evaporation of free water from the membrane as the temperature rises (Zhang et al., 2014). Compared with the Nafion^®112 membrane, the anhydrous proton conductivity of PA/HMS increased from 1.6 × 10⁻² S.cm⁻¹ at 120°C to 2.5 × 10⁻² S.cm⁻¹ at 180°C. This is likely due the fast proton conduction between H₂PO₄ ⁻ and HPO₄ ²⁻ through the Grotthuss mechanism (Asensio et al., 2010). Furthermore, when the RH increased to 10%, the proton conductivity of PA/HMS at 120°C increased to 4.0 × 10⁻² S.cm⁻¹, which is 1.5 times higher than that of the membrane under 0% RH. In addition, the cell performance of the HT-PEMFC based on the PA/HMS membrane was measured at 160°C and 0% RH, as shown in Figure 7F. The open circuit voltage of the fuel cell under H₂/O₂ atmosphere is only 0.83 V, which is likely due to the gas permeation through the pinholes in the PA/HMS. However, the maximum output power of the fuel cell under 160°C reached 55.6 mW.cm⁻². Therefore, this successful result demonstrates the feasibility of the PA/HMS inorganic membranes in HT-PEMFCs.

Hollow mesoporous silica particles were synthesized by etching mSiO₂ microspheres in a weakly alkaline solution via an ion exchange-induced selective etching strategy. It was found that the mesoporous silica microspheres were etched from the center, and the volume of the void increased with the extension of the etching time. For the mSiO₂ microspheres with CTAB in the mesopores, the Br⁻ ions in CTAB were exchanged, with OH⁻ ions in the alkaline solution, so that the OH⁻ concentration in mSiO₂ was higher than the concentration of OH⁻ ions in the solution. As a result, the inside of the mSiO₂ microspheres was etched first to form HMS. In addition, when HMS was impregnated by phosphoric acid via the vacuum impregnation method, the proton conductivity of PA/HMS at 160°C and 0% RH reached 2.5 × 10⁻² S.cm⁻¹, and the peak power density of the HT-PEMFCs based on the PA/HMS membrane reached 55.6 mW.cm⁻² under the same conditions. Therefore, the PA/HMS inorganic membrane is a promising candidate for applications in HT-PEMFCs.