Gold nanoparticles (AuNPs) have attracted great interest owing to their valuable potential in many important catalytic reactions, such as hydrogenation of olefins, water gas shift reaction, direct synthesis of hydrogen peroxide and selective alcohol oxidation [1]. The catalytic activities of AuNPs are largely associated with their sizes, atom-packing configuration, crystal facets, and the interface types with supports [2, 3]. However, the thermodynamic tendency of AuNPs to grow into larger crystallites, which minimizes their surface energy, is a key obstacle to achieving efficient and sustainable catalytic activity [4-9]. To maintain AuNPs within the desired size range at high temperatures (generally > 300 ℃), both atom migration (Ostwald ripening) and particle migration need to be controlled during thermal treatment [10, 11].

Porous materials have long been used as supports for the preparation of supported AuNP catalysts to improve their thermal stability. There are numerous proposed methods directed at hindering sintering effects caused by particle or atom migration [12-23]. Generally, the narrow pore dimensions of porous supports enhance the physical barriers that mitigate the rate of particle growth of supported AuNPs [14, 24, 25]. Additionally, high metal loading should be avoided to suppress particle aggregation by increasing inter-particle distances [26]. Conversely, pore connectivity and structure have been reported to significantly impact NP stability [14, 27, 28]. It has been shown that one-dimensional (1-D) pores are more efficient in restricting particle growth than three-dimension (3-D) pore systems by restraining metal NP transportation [26, 27]. However, mass diffusion limitations and pore blockage are more of an issue in 1-D porous systems than their 3-D counterparts.

Prior to this work, we have shown that excellent anti-sintering properties can be achieved at high AuNP loading levels when using the extra-large mesoporous silica, EP-FDU-12 (cage size > 25 nm), as the support [29]. The 3-D porous structure of EP-FDU-12 allows two or more AuNPs to reside in every extra-large mesopore at high loading concentrations ( > 9 wt%). Thus, inter-particle interactions are remarkably strengthened via significant overlapping of the diffusion-spheres of AuNPs and local vapor-particle equilibrium in each cage is possible. In this case, a self-focusing mechanism dominates the particle growth, leading to an obvious suppression of atom migration and therefore outstanding anti-sintering properties. Nevertheless, traditional EP-FDU-12 having larger entrance sizes ( > 7 nm) is not able to suppress sintering at low loading concentrations. Following the classic Ostwald ripening mechanism, the AuNPs grow larger after high-temperature calcination.

Herein, we study the sintering resistance properties as a function of FDU-12 pore entrance size. Simple adjustments to the type of organic template and reaction temperature enable the successful synthesis of FDU-12 with controllable entrance sizes ( < 3, 3-5 and 7 nm). Different NP growth mechanisms are proposed for each of the different entrance sizes. It is worth noting that stabilized AuNPs having the desired size range over a narrow distribution (4.5-5.0 nm) form after calcination (550 ℃ in air for 5 h) when supported on FDU-12 having a sub-5-nm entrance size. This stabilization is achieved over a wide loading concentration (1.0-8.3 wt%) in contrast to EP-FDU-12 having a larger entrance size (7 nm). Catalytic performance of the supported AuNPs is evaluated from gas-phase alcohol selective oxidation reactions. It is shown that the supported Au catalysts with the desired uniform size distribution exhibit excellent mass-specific activity (1544 mmol g_Au^-1 h^-1) and selectivity ( > 99%) at 230 ℃ for the gas-phase selective oxidation of cyclohexanol, which significantly surpasses previously reported Ag-or Mn-based catalysts (784 mmol g_Ag^-1 h^-1, 50% selectivity and 879 mmol g_Mn^-1 h^-1, 70% selectivity, respectively) [30, 31]. Furthermore, AuNP/FDU-12 catalysts show superior long-term stability under the applied reaction conditions.

Cubic mesoporous silica (FDU-12) with controllable sub-5-nm entrance sizes was synthesized as a function of changes to the type of organic template and synthesis temperature, following previously reported methods [32, 33]. In a typical synthesis, 0.50 g of triblock copolymer EO₁₃₂PO₅₀EO₁₃₂ (Pluronic F108, BASF), 0.60 g of 1, 3, 5-trimethylbenzene (TMB, CP grade, Sinopharm Chemical Reagent Co. Ltd.) and 1.25 g of KCl (AR grade, Sinopharm) were dissolved in 50 mL of 1 mol/L hydrochloric acid at 25±0.1 ℃. After stirring for 1 h, 2.08 g of tetraethyl orthosilicate (TEOS, AR grade, Sinopharm) was added to give a molar sol-gel ratio of 0.00136 F108 : 0.62 KCl : 0.37 TEOS : 0.185 TMB : 1.85 HCl : 100 H₂O. After stirring for 24 h at 25±0.1 ℃, the mixture was transferred into an autoclave and heated at a desired temperature for 24 h. For FDU-12(18, 4) (18 and 4 indicate the pore and entrance size, respectively), the hydrothermal temperature was set to 140 ℃ and lowered to 100 ℃ for FDU-12(15, < 3). The as-made product was obtained by filtration and air-dried at room temperature. The organic template was removed by calcination in air at 350 ℃ for 5 h. EP-FDU-12 having larger entrance sizes was synthesized according to our previous report [29].

The synthesis of AuNPs was similar to the reported method [34]. In a typical synthesis, 100 mg of AuPPh₃Cl was mixed with 400 μL of dodecanethiol (≥98%, Sigma-Aldrich) in 20 mL of toluene (AR grade, Sinopharm) to form a clear solution, to which 84 mg of NaBH₄ (96%, Sinopharm) was added in one portion. The mixture was heated with stirring at 55 ℃ for 7 h before the reaction system was cooled to room temperature. AuNPs were precipitated out from the reaction mixture as a black solid powder by addition of 40 mL of ethanol. The precipitate was separated by centrifugation before washing with ethanol and natural drying.

AuNPs were loaded into mesoporous silicas by a colloid deposition method [35]. The desired amount of AuNPs was suspended in 25 mL of chloroform. To this solution, the desired amount of mesoporous silica was added. After stirring for 1-24 h, the solid product was centrifuged and air-dried. The AuNP/mesoporous silica samples were thermally treated at 550 ℃ in air for 5 h to evaluate their anti-sintering properties and catalytic performance. The preparation of AuNPs on a commercial γ-Al₂O₃ (99.9%, Aladdin) was identical to the AuNP/mesoporous silica method, except that the calcination temperature was lowered to 350 ℃. AuNP particle size was analyzed by transmission electron microscopy (TEM) and X-ray diffraction (XRD) measurements.

The oxidation of cyclohexanol was performed within an electronically controlled furnace in a fixed bed vertical glass reactor (length=250 mm, i.d.=12 mm), fitted with a glass frit carrying the catalyst mixed with quartz sand. The oxygen stream (4 mL/min) was controlled by a mass flow instrument and the liquid reagent (0.45 mL/h) was supplied through a syringe pump. Liquid vaporization occurred on the reactor wall prior to the catalytic bed. The products were analyzed by gas chromatography (GC). Selectivity was calculated as moles cyclohexanone product per mole of reacted cyclohexanol and the carbon mass balance was > 99% in the case where selectivity was > 99%.

Small-angle X-ray scattering (SAXS) patterns were recorded using a Xenocs Xeuss-saxs/waxs system. Nitrogen adsorption isotherms were measured at-196 ℃ on a Micromeritics ASAP 2020 adsorption analyzer. Before the adsorption analysis, calcined samples were outgassed under vacuum at 200 ℃. TEM micrographs were recorded on a Hitachi HT-7700 microscope operated at 100 kV and a FEI Tecnai G² F20 S-TWIN microscope operated at 200 kV. The samples were embedded in epoxy resin and microtomed for TEM imaging. Scanning electron microscopy (SEM) images were taken on a Hitachi SU-8010 microscope operated at 1 kV. XRD patterns were recorded on a Ragaku Ultima Ⅳ diffractometer using Cu K_α radiation. The average crystal size is derived from the Scherrer equation based on the peak width of the Au (111) reflection.

The ordered 3-D cubic mesostructure of FDU-12 having a narrow entrance size was first confirmed by SAXS measurements (Fig. 1). A typical SAXS pattern of sample FDU-12(18, 4) exhibits six peaks, which can be indexed to the (111), (200), (220), (311), (331) and (531) reflections of an fcc structure (Fm3m). The cell parameters of FDU-12(18, 4) and FDU-12(15, < 3) are calculated to be 34.5 and 32.3 nm, respectively.

Nitrogen sorption analysis was further utilized to measure their porous structures and textural properties. As shown in Fig. 2, the isotherms of two samples both show a type H4 hysteresis loop with a steep capillary evaporation step at p/p₀=0.45 because of the tensile strength effect [36, 37], indicating that the samples have an entrance size smaller than 4.9 nm. The physicochemical properties of the mesoporous silica supports discussed in this work are listed in Table 1.

TEM images of FDU-12(18, 4) taken along the [110] and [233] directions are shown in Fig. 3(a) and (b), respectively. The cell parameter calculated based on the value of d(1-11) and d(3-1-1) is 34.8 nm, in good agreement with the value derived from SAXS results. Furthermore, the pore size estimated from the diameter of the white disks in Fig. 3(c) is also in agreement with the N₂ sorption measurement (18.0 nm). As shown in the SEM image in Fig. 3(d), FDU-12(18, 4) displays a hexagonal prism-like morphology with open pores.

To further determine the entrance size of FDU-12(15, < 3) and FDU-12(18, 4), which is beyond the detection limit of the nitrogen sorption measurements ( < 4.9 nm), we tactfully use 3-nm AuNPs as a probe (Scheme 1) [38]. The pre-synthesized AuNPs have a uniform size of 3.0±0.4 nm (Fig. 4). Entrance size can be determined by observing if the 3-nm AuNPs have ingress to the cage-like pores or not. As shown in Fig. 5(a) and (b), the overwhelming majority of AuNPs are located on the external surface of FDU-12(15, < 3), indicating that the entrance size of FDU-12(15, < 3) is smaller than 3 nm. In contrast to FDU-12(15, < 3), 3-nm AuNPs disperse evenly within the cages of FDU-12(18, 4) (Fig. 5(e) and (f)), suggesting that its entrance size is large enough to allow full ingress into the large cages. TEM observations together with the textural properties derived from sorption measurements strongly suggest that the entrance size of FDU-12(18, 4) is in the range of 3-5 nm. Difference between the entrance sizes of the two FDU-12 samples is also reflected in their AuNP adsorption capacity [39]. AuNPs can only be adsorbed on the external surface of FDU-12(15, < 3), whereas they are readily accessible to the large internal surface of FDU-12(18, 4) and are accommodated within the cages. As a result, at the same loading level (~6 wt% AuNPs), uptake onto FDU-12(18, 4) is stoichiometric (loading concentration 6.0 wt% determined by ICP element analysis), while a final product loading level of 1.0 wt% is observed for FDU-12(15, < 3) even after vigorous stirring for 24 h.

The location of the AuNPs on the FDU-12 supports has a significant influence on their thermal stability (Scheme 1). After thermal treatment at 550 ℃ in air for 5 h, the AuNPs residing in the cages grow from 3 to 4.5 nm (Fig. 5(g)), whereas the size of the AuNPs on the external surface is~6.0 nm (Fig. 5(c)). Their different thermal stability is also confirmed by XRD analysis (Fig. 6). The AuNPs outside of the cages are 2.8 nm larger than those entrapped within the cages.

The thermal stability of AuNPs within FDU-12(18, 4) is studied as a function loading concentration (1.0-8.3 wt%). After thermal treatment, the AuNP size is maintained in the range of 4.5-5.0 nm regardless of the initial loading concentration (Fig. 7). All observed sizes are smaller than the corresponding AuNPs supported on FDU-12(15, < 3). In addition to TEM analysis, XRD measurements on AuNP/FDU-12(18, 4) reveal a consistent AuNP size irrespective of the loading level as calculated by the Scherrer equation based on the peak width of the Au (111) reflection (Fig. 8), confirming that the thermal stability of AuNPs entrapped within FDU-12(18, 4) is not influenced by AuNP loading concentration. This behaviour is completely different from most traditional porous supports, where NPs are observed to grow larger as a function of increased loading concentration because of reduced particle-to-particle distance [40, 41].

Fig. 9 summarizes the statistical size distribution results of AuNPs supported on FDU-12 as a function of entrance size and loading concentration. At a high loading concentration ( > 9 wt%), the self-focusing mechanism narrows the particle distribution and reduces the driving force for Ostwald ripening, thereby assisting the AuNP/FDU-12(27, 7) system to quickly reach dynamic equilibrium, ultimately resulting in the promising thermal stability of AuNPs within FDU-12(27, 7) [29]. At a low loading (e.g. 2.1 wt%), the majority of the FDU-12(27, 7) cages are vacant. Au vapor or small AuNPs residing within a cage can diffuse through the large entrance of the cage, which significantly hinders the system from reaching saturation vapor pressure as given by the Gibbs-Thompson equation, leading to uncontrolled particle growth (14.2±5.5 nm). However, when FDU-12(18, 4) is used as the support, AuNPs larger than 5 nm will not be able to diffuse to another cage because of the dimension limitations of the small entrance size. Additionally, the existence of large cages increases, to a degree, inter-particle distances and small entrance sizes weaken atom migration. As a result, the AuNPs entrapped within FDU-12(18, 4) can be stabilized over a tight size distribution (4.5-5.0 nm) across a wide loading concentration (1.0-8.3 wt%).

Supported AuNPs are promising catalysts for selective alcohol oxidation [42, 43]. For some alcohol molecules (e.g. benzyl alcohol, ethanol), only AuNPs within a certain size range will exhibit promising catalytic activity [44, 45]. Thus, stabilization of AuNPs within a desirable size range is critical for specific catalytic applications. In our previous work, we observed that 5-6 nm AuNPs show optimum activity for gas-phase selective oxidation of cyclohexanol to cyclohexanone [46].

In this work, activity was studied as a function of AuNP loading onto FDU-12 supports and a commercial γ-Al₂O₃ for comparison. Catalytic reactions were evaluated by probing the gas-phase selective oxidation of cyclohexanol. Each reaction was duplicated, however, using a different Au catalyst mass, each within a fixed bed vertical glass reactor having the same cross sectional area. As shown in Fig. 10, the similar activities for the selected two catalysts at the equivalent contact times (0.267-1.425 g s/mL) suggest that the diffusion effect has been eliminated (1.069 g s/mL).

Fig. 11 shows the catalytic performance of four Au-supported FDU-12 catalysts as a function of entrance size. 4.5-nm AuNPs entrapped within FDU-12(18, 4) having a controlled size distribution exhibit optimal activity (1544 mmol g_Au^-1 h^-1at 230 ℃). Activity almost doubles (825 mmol g_Au^-1 h^-1) for AuNPs located on the external surface of FDU-12(15, < 3), which have an average particle size as large as 6.0 nm. At low loading concentrations, the AuNPs supported within FDU-12(27, 7) have a much larger size of 9.9 nm and present the lowest activity (171 mmol g_Au^-1 h^-1at 230 ℃), which corresponds to only 11% of the AuNP/FDU-12(18, 4) activity. A similar trend is observed at lower reaction temperatures (180 and 210 ℃). Furthermore, the catalytic data provides direct evidence to optimize activity by tuning of the AuNP size as a function of calcination temperature. In this study, 550 ℃ is the optimal temperature for thermal treatment. If the thermal treatment temperature is lowered to 450 ℃, the AuNP size within FDU-12(18, 4) decreases from 4.5 to 3.9 nm. As a result, the catalytic activity of the resulting catalyst lowers to 663 mmol g_Au^-1 h^-1at 230 ℃. Catalyst development herein further proves that stabilizing active metal particles within a desired size range is vitally important to fully realize their catalytic function. Furthermore, the mass-specific activity of 4.5-nm AuNPs supported on tailored FDU-12 silica significantly surpasses previously reported Ag-and Mn-based catalysts (784 mmol g_Ag^-1 h^-1and 879 mmol g_Mn^-1 h^-1, respectively) for the gas-phase selective oxidation of cyclohexanol [30, 31] and Ru-and Au-based catalysts (224 mmol g_Ru^-1 h^-1and 645 mmol g_Au^-1 h^-1, respectively) for the liquid-phase selective oxidation of cyclohexanol [47, 48].

To compare with the AuNP/FDU-12 system, the catalytic activity of 1.0 wt% AuNP/γ-Al₂O₃ was tested. After calcination at 350 ℃ for 5 h, AuNPs on γ-Al₂O₃ were observed at 4.1±1.4 nm (Fig. 12(a)). Catalytic conversion and selectivity of the obtained AuNP/γ-Al₂O₃ catalyst are shown in Fig. 12(b). Although cyclohexanol conversion corresponding to the 1.0 wt% AuNP/γ-Al₂O₃ and 1.0 wt% AuNP/FDU-12(18, 4)-4.5 nm systems is similar (23.7% vs. 22.2% at 210 ℃, respectively), the selectivity of 1.0 wt% AuNP/γ-Al₂O₃ is significantly worse (34.0% by-products at 210 ℃). Increasing the reaction temperature to 230 ℃, dehydration of cyclohexanol is increasingly favored (52.9% by-products), which likely relates to the Lewis acid sites of γ-Al₂O₃[49-51].

Furthermore, we investigated the catalytic stability of 1.0 wt% AuNP/FDU-12(18, 4)-4.5 nm and 1.0 wt% AuNP/FDU-12(15, < 3)-6.0 nm under the applied reaction conditions. As shown in Fig. 13(a), no obvious deactivation is observed in both catalysts during a time-on-stream of 100 h at 210 ℃ (only 3.4% and 4.1% activity loss after 100 h for 4.5-nm and 6.0-nm AuNPs, respectively). Additionally, the particle sizes of the fresh and used AuNPs derived from the Scherrer equation taken from XRD analysis (Fig. 13(b)) are near identical (5.1 vs. 5.3 nm and 8.5 vs. 8.6 nm), which further indicates that AuNP/FDU-12 catalysts are very stable under reaction conditions for alcohol oxidation.

We have successfully synthesized cubic mesoporous silica (FDU-12) with controllable entrance sizes and studied their impact on AuNP anti-sintering. Additionally, our work reveals the entrance size for 3-D cage-like mesoporous supports has significant influence on the particle growth. FDU-12(27, 7) shows promising anti-sintering properties at high loading concentrations because of the self-focusing mechanism, which suppresses atom migration; however, at low loading concentrations poor performance is observed. In contrast, FDU-12(18, 4) is able to stabilize AuNPs within the desired size range over a wide loading concentration. The unique 3-D cage-like pores with tailored narrow entrance size act as excellent physical barriers to particle mobility and therefore hinder particle coalescence efficiently. Furthermore, 4.5-nm AuNPs stabilized within FDU-12(18, 4) exhibit excellent activity and long-term stability in the gas-phase selective oxidation of cyclohexanol. It is expected that the outcome of this work will help in the design of highly efficient and durable catalysts for applications where strict control of the metal particle size is key to application optimization [52-56].