Tetraethyl orthosilicate (TEOS, 98%), 3-(trimethoxysilyl) propylmethacrylate (MPS, 98%), styrene (St) (>99%, Aldrich), divinylbenzene (DVB) (80%, Aldrich), 2,2'-azobisisobutyronitrile (AIBN) (98%, Aldrich), methacrylic acid (MAA) (>98%, Aldrich), hydrofluoric acid (HF) (48% in water, Aldrich), doubly distilled water (DDW), ammonia (25% in water, Aldrich) and analytical grade acetonitrile and ethanol were used for the experimental processes.

MPS modified silica templates were prepared according to the Stöber method [26,27]. About 4.5 mL of TEOS was added to a mixture of 75 mL ethanol, 7.5 mL DDW, and 1.5 mL ammonia. The mixture was vigorously stirred at room temperature for 24 h. A total of 1 mL MPS was then injected into the silica sol dispersion and the dispersion was stirred at room temperature for another 24 h. After the reaction, the resultant MPS modified silica particles SiO₂-MPS were purified by centrifuge.

About 0.2 g SiO₂-MPS powder was dispersed into 80 mL of acetonitrile under ultrasonication in a 100 mL flask, equipped with a fractionation column, a condenser, and a receiver. A mixture of St (1.15 mL, 0.01 mol), DVB (0.267 mL, 0.0015 mol, 15 mol% relative to styrene), and 2,2'-azobisisobutyronitrile (AIBN, 0.02 g) was then added into the flask. The polymerization then proceeded under reflux and removing of solvent by gradual distillation. Through this distillation process, the P(St-DVB) network formed in the solution phase underwent effective grafting to the MPS surface anchors, leading to a polymer shell on each SiO₂ microsphere. The reflux lasted 1.5 h and 40 mL acetonitrile was distilled off. The resultant SiO₂-P(St-DVB) core-shell microspheres were purified using soxhlet extraction procedure in which the remaining monomers and oligomers were removed by a binary solvent of acetonitrile and ethanol. The resulting SiO₂-P(St-DVB) powder was vacuum dried at 50 °C. The particles were then sulfonated in a concentrated sulfuric acid at 50 °C for 3 h and the mixture was poured into a cold ethanol to allow sedimentation of pale yellow particles. After centrifuging and washing with ethanol, the collected powder was vacuum dried. Similarly, the preparation of SiO₂/P(MAA-DVB) was prepared by the distillation polymerization except sulfonation. Hence two different types of hydrophilic SiO₂/polymer core-shell particles were synthesized.

The SiO₂/polymer core-shell microspheres were treated in 8% HF for 48 h by gentle agitation in a plastic flask. The unused HF and reaction product (SiF₄) in liquid phase were separated from the hollow microspheres by centrifuging and the microspheres were washed in ethanol and water for several cycles. The hollow polymer spheres prepared are labeled by s-HPS and c-HPS, respectively, with the former bearing –SO₃H pendant groups while the later one –CO₂H pendant groups. The equivalent weights of s-HPS (5.7 mmol SO₃H/g) and c-HPS (4.8 mmol CO₂H/g) were determined by titration. In a typical run, 0.01g sample was gently stirred in 20 mL 0.01M NaOH for overnight. The mixture was centrifuged and the clear decant was titrated by 0.01M HCl. In addition, the sample of hollow spheres with the pristine P(St-DVB) was prepared as control.

As a typical procedure, Nafion solution (5 wt % in a mixture of low aliphatic alcohols/H₂O (20%), EW 1100) was mixed with HPS (0.5 wt % on the basis of dry Nafion resin). The suspension was ultrasonicated until a homogeneous sol was obtained. The sol was casted onto a petri dish and dried at 80 °C for 12 h. The membrane obtained was soaked in 1 M H₂SO₄ for at least 24 h before use. The thickness of membranes was fixed at about 50 μm. Three composite membranes obtained are named by N/s-HPS, N/c-HPS and N/n-HPS. As defined above, the first two denote the composite membranes containing s-HPS and c-HPS in Nafion matrix respectively, while the last one denotes the composite membrane containing the HPS that has neutral or hydrophobic P(St-DVB) shell, which was prepared as the control sample.

The synthesis of submicron s-HPS employed the bottom-up strategy as depicted in Scheme 1 and the respective FTIR spectrum was shown in Figure 1. Initially SiO₂-MPS seed particles with diameters of 100–150 nm were synthesized using the Stöber process [26,27]. The FTIR absorption spectrum of the SiO₂-MPS particles shows a peak at 1630 cm⁻¹ (Figure 1a), which corresponds to the double bond (vinyl groups) stretching of the MPS units attached to the SiO₂ microspheres. This peak shows a shoulder at about 1730 cm⁻¹ representing the carbonyl group. Vinyl groups on the surface of SiO₂-MPS microspheres then served as anchor to the radical chains produced in the solution upon polymerization, and the accompanying distillation drove the interfacial grafting to generate the SiO₂-P(St-DVB) core-shell microspheres. The spectrum (Figure 1b) proves realization of this structure according to benzene ring stretching at 1500 cm⁻¹, aromatic C–H stretching at 3010 cm⁻¹ and the main chain C–H stretching at 2900 cm⁻¹. The subsequent sulfonation introduced –SO₃H groups to benzene rings in the P(St-DVB) shell. The characteristic double S=O stretching peaks at about 1041 and 1180 cm⁻¹ could be observed only after silica core was removed since the broad –Si–O–Si– symmetric stretching peak appears at 1000~1200 cm⁻¹ (Figure 1b).

Figure 2 shows the TEM image of the core-shell structures of the SiO₂/P(St-DVB) and SO₂/P(MAA-DVB) spheres. The resultant core-shell microspheres of both types had a 20–30 nm thick shell. Hollow polymer spherical structure with two types of functional groups, –SO₃H (s-HPS) and –CO₂H (c-HPS), were securely retained after the silica core was removed. It is noteworthy that the P(St-DVB) must be sulfonated in the presence of SiO₂ core in order to etch the SiO₂ core more readily and to avoid splintering of the P(St-DVB) shell while it is sulfonated. The hollow structures of the s-HPS and c-HPS spheres exhibited well-defined capsular structure and an integrity shell. Also, the sufficiently high DVB crosslinking extent assured a free-standing shell and prevented severe agglomeration among HPS of both types.

The cryofractured cross-sections of the four membranes of interest exhibited a clear transition from a dense and homogeneous matrix of the pristine Nafion membrane to a micro-rupturing matrix of the composite membranes, i.e., N/s-HPS and N/c-HPS. This change in morphology is typically due to interaction between Nafion matrix and the filler used. The composite membrane N/n-HPS, containing neutral HPS, nevertheless still held a rather homogeneous matrix, analogous to its pristine counterpart (Figure 3a). Assimilating either s-HPS or c-HPS into Nafion matrix was to trigger hydrophilic interaction between –SO₃H groups of Nafion molecules with s-HPS or with c-HPS [33]. Compared to the microstructure of the pristine Nafion (Figure 4), consisting of perfluoro-polymer domains and a hydrophilic and continuous boundary phase, hence the hydrophilic HPS associated with the boundary composed of pendant –SO₃H groups to form a particle-domain packing morphology since both have roughly comparable submicron particle sizes (Scheme 2). As a result, the association of the Nafion domains became insignificant or was weakened due to spatially distancing effect. After drying, such composite membrane is mechanically weaker than the denser pristine Nafion membrane matrix. Hence, numerous micro-ruptures could be observed across the cross-section (Figure 3c,d). Owing to considerably low specific density, 0.5 wt % loading of either c-HPS or s-HPS would have taken up a large volume fraction of the hydrophilic phase of Nafion. However, the smaller volume fraction of hydrophilic phase, relative to the hydrophobic perfluoropolymer phase, in Nafion could not separate the filler particles extensively and thereby the structure of composite can be illustrated by Scheme 2, which was actually verified by many loosely bound hydrophilic HPS and the broadened gap between the Nafion domains in Figure 3c,d.

These hydrophilic HPS aggregates functioned as strong moisture reservoir with prominent effect on proton transport, which will be elaborated in the following section. Despite the positive effect of hydrophilic HPS, a loading exceeding 0.5 wt % would further weaken the cohesive strength of membrane and hence be irrelevant. On the contrary, being hydrophobic in nature, n-HPS particles were likely to be entangled by perfluoro-chains upon drying. As the hydrophobic perfluorocarbon chains take up the major volume fraction and this hydrophobic combination does not or only slightly perturb the ionic clustering of –SO₃H throughout the membrane, a dense morphology could still remain after being cryofractured. In summary, the hydrophilic HPS filler acted to broaden proton conducting channel in the composite Nafion membrane.

It is believed that proton transport occurs in hydrophilic channel consisting of acidic groups (typically –SO₃H) and proton-water clusters (such as H₃O⁺, H₅O₂⁺, and H₉O₄⁺) [34,35,36]. Therefore, besides the concentration of sulfonic acid groups, both physical states and content of the water entrapped in hydrophilic channel are influential to the proton transportation. As proposed above, the assimilation of hydrophilic HPS into Nafion was to enhance water uptake. The temperature effect on water uptake at 100% RH showed the sequence as expected (Figure 5a) where N/s-HPS revealed the largest water-absorption capacity of the three membranes under investigation. Furthermore, their water absorption capability was examined by equilibrating the membranes with different RH% at 20 °C (Figure 5b). The composite membranes revealed stronger moisture capture capability than the pristine cast membrane. Fundamentally, this implies the existence of a denser distribution of hydrophilic sites in the composite membranes. Furthermore, the condensed states of water entrapped in a membrane can be described by “free water” freezing at 0 °C, “freezable bound water” or “freezable water” freezing below 0 °C, and “non freezable bound water” or “non freezable water” due to absence of hydrogen bonding among water molecules, and hence, no phase transition can be detected.

Both free water and freezable water can be identified by the low temperature DSC profiles [37,38] as shown in Figure 6. The four membranes displayed only freezable water existing below 0 °C. Assuming that the specific fusion heat of freezable water would remain approximately the same as that of free water (4.18 cal/g) regardless of the shift of freezing point, the amount of this type of water could be calculated (Table 1). Subsequently, the non-freezable water would be the difference between water uptake and the amount of the freezable water. In contrast to the freezing point of freezable water in the Nafion membrane, which is identical to the reported value [38,39], the freezable water in the three composite membranes showed higher freezing point as expected due to the lodging of water inside the chamber of HPS. Unlike some hydrophilic polymer matrices, such as poly(vinyl alcohol) [40] and sulfonated polystyrene [41], these three composite membranes contain no free water. This observation can be attributed to the substantially small sizes of dispersed water droplets and a high percentage of water molecules associating with the pendant –SO₃H groups. Hence, the bulk properties of the water droplets significantly faded. By this interfacial effect, the stronger the interfacial interactions between water droplets and hydrophilic groups, the lower the freezing point. Indeed the water freezing points of the composite membranes showed the order: N/s-HPS < N/c-HPS < N/n-HPS, which is determined by the hydrophilic affinity of HPS, i.e., the stronger affinity led to stronger interfacial interactions following the order of s-HPS > c-HPS >> n-HPS in terms of hydrophilicity. This order also reflects the same trend of freezable water% described in Table 1.

Of the three composite membranes, the N/n-HPS membrane contained hydrophobic n-HPS particles that associated with perfluorochains as ascribed above. The analysis of water state further suggests that this association reduce the packing density of perfluoro chains because of the occurrence of the interface between perfluorochain and P(St-DVB). Such perturbation occurring in the hydrophobic domains might also affect the clustering of –SO₃H groups in the hydrophilic channels to a small extent. This is supported by the increase in freezable water content and the reduction in non-freezable water content in the N/n-HPS relative to the pristine Nafion membrane.

The measured IEC values (Table 1) unveil not only the equilibrium accessibility to the pendant acid groups but also structural response in connection with the above assignment on hydrophilic channel. Both Nafion and N/n-HPS show the same IEC, implying no blockading action imposed by the n-HPS particles to the hydrophilic channels. Moreover, the composite membranes, N/c-HPS and N/s-HPS, showed IEC values of 0.82 and 0.88 mmol/g, respectively. The calculated value of IEC of N/s-HPS can be made using the measured IEC of Nafion membrane (0.75 mmol/g) and the measured equivalent mass of s-HPS (5.7 mmol/g) by the formula: 0.5% × 5.7 mmol/g + 99.5% × 0.75 mmol/g = 0.775 mmol/g. Obviously, this calculated value is well below the experimental IEC of N/s-HPS. This difference of about 13.5% validates the role of s-HPS particles in broadening hydrophilic channel, which opened accessibility of those dead ends in the pristine Nafion matrix. A similar discrepancy also existed in N/c-HPS. In addition, the difference in IEC between N/s-HPS and N/c-HPS is 0.06 mmol/g, a far greater discrepancy than that between s-HPS and c-HPS (i.e., 0.0045 mmol/g). As found previously, the IEC value is largely affected by a strong acid rather than a weak acid [42].

The above study on water uptake, condensed state of water, and IEC comes to the conclusion: both s-HPS and c-HPS participated in restructuring hydrophilic channel in Nafion matrix, entrapped additional amount of water in the tiny cavities, and functioned as supplement pendant acid sites. Similar phenomena were also observed previously [43,44]. The enhancement of water reservation and reachability of acid sites on proton transport and electrochemical polarization was examined in this section. Figure 7 investigated variation of proton conductivity of the three membranes with temperature at 100% RH. A clear sequence of N/s-HPS > N/c-HPS > Nafion was presented over the temperature range (25–95 °C). This sequence is consistent with the water uptake, which favored the Grotthuss or hopping mechanism, and with the density of acid groups, which enhanced proton flux and possibly the vehicle mechanism [45]. Regarding the superior conductivity of N/s-HPS over N/c-HPS, the stronger –SO₃H (based on ionization constant K_a) than –CO₂H is an additional factor besides the above two.

Following the examination on the water uptake under various RH% (Figure 5), the three membranes revealed different rising trends of proton conductivity with the increase in temperature from 20 to 90 °C under constant 10% RH (Figure 8a). In such a chamber, the membranes captured just 6–9 wt % water at 20 °C and the bulk moisture levels would be lower than this range at higher temperatures. This test demonstrated the impact of the bound acid groups as well as the structure of hydrophilic channel on transport of protons because the Grotthuss mechanism became subsidiary at such moisture level. The conductivity values though show the order of magnitude of −3 under 10% RH, the discrepancies of them tell the capability of shipping protons by pendant acid groups, –SO₃H and –CO₂H. The N/s-HPS membrane presented the strongest capability of shipping protons through contact of spherical surface (Scheme 2) because this two-dimensional mould is far more efficient than transport along tangled polymer chains. A similar test was conducted to collate the proton-conductivity decreasing profile with the measurement time under 10% RH and at 20 °C (Figure 8b).

In this test, a fully hydrated membrane sample was employed at the beginning of test. The capability of deferring moisture evaporation as well as maintaining proton transport with decreasing of bulk humidity was assessed. Taking the conductivity values at 110 min as example, N/s-HPS held a conductivity of at least 5 times greater than N/c-HPS, while the Nafion membrane became almost non-proton conductive. This sequence is consistent with that displayed in Figure 5, in which water-retention power originates from the content of freezable water (Table 1) since it reflects the impact of micro-environment, including size of cavity and the type of acid group the hydrophilic shell bears, on water-retention capability of the membrane.

Unlike the proton conductivity measurement that reveals the structural effect on small proton flux, the electrochemical polarization curve of a single H₂-PEMFC manifests the capability of the loaded membrane to sustain a larger stream of protons. Figure 9a displays the polarization curves of the four membranes including the commercial Nafion N-112 membrane at 20 °C without external humidification. The two composite membranes held higher cell voltages with the increase in current density than the two pristine Nafion membranes. It is considered that the differences in the water uptakes amid the membranes (at 20 °C in Table 1) played the main role. The hydrophilic channels, constructed by the –SO₃H groups of Nafion and either s-HPS or c-HPS, can sustain higher proton flux than that of its pristine counterpart. With raising the single cell operation temperature to 70 °C (Figure 9b), the N/s-HPS membrane sustained the maximum power density of 525 mW/cm², representing an increment of 106% against the utmost value of 255 mW/cm² obtained at 20 °C. In contrast to this, the Nafion membrane achieved only an increment of 73% through this temperature rise.

As shown in the water-uptake/temperature measurement (Figure 5a), these two membranes had about 11 wt % difference in water uptake at 70 °C, responsible for the different energy output. The other important factor is the different proton transport efficiency as observed in examining conductivity under 10% RH. The N-112 membrane also possessed a lower power density than that of the composite membranes, validating the impact of the above two factors that originates from the structure of hydrophilic channel.