Carbon nanotubes (CNTs) are composed of graphene layers with a tubular morphology [1], which distinguishes them from other carbon materials such as activated carbon and carbon nanofibers. The tubular morphology leads to the π electron density shifting from the interior to the exterior surface [2]. Therefore, the nanochannels of CNTs are frequently used as a container and they provide an intriguing confinement environment for catalysis [3, 4]. A number of substances and materials have been introduced into the channels of CNTs, and they exhibit different properties and behavior from those in the bulk and on the exterior walls of CNTs [5, 6].

It was reported that the interaction of transition metal atoms with CNT walls significantly differs from their interaction with planar graphite layers with regard to the bonding sites, magnetic moments, and charge-transfer directions [7], which could modify their physiochemical properties. For example, reduction of metal nanoparticles such as Fe [6] and Ru [8] is facilitated when they are confined within CNT channels compared with when they are dispersed on the exterior walls of CNTs. Reduction is facilitated more within smaller nanotubes owing to more distorted sp² hybridization and hence stronger confinement effects [5]. This could affect the catalytic activity, particularly for reactions involving reduction and oxidation. For example, the CNT-confined Fe catalyst exhibits higher activity in Fischer-Tropsch synthesis [9, 10]. The CNT-confined RhMn catalyst also exhibits higher activity than nanoparticles dispersed on the exterior walls of CNTs in C₂ oxygenate synthesis [11]. Confinement within the CNT channels also provides protection for metallic nanoparticles [12] and their oxidation is retarded. This allows generation of highly active and durable oxidation catalysts. For instance, CNT-confined Pt nanoclusters exhibit high activity and stability for oxidation of methylbenzene compared with conventional Al₂O₃, zeolite, and high surface area activated carbon supported Pt catalysts [13]. Although many studies have reported different activities by introducing metal nanoparticles into CNT channels, few studies have discussed the effect on the product selectivity. Serp and co-workers [14] reported enhanced activity for confined PtRu nanoparticles in selective hydrogenation of cinnamaldehyde (CAL) within CNTs with an average inner diameter as large as 40 nm. They found that the product selectivity was related to the location of the PtRu nanoparticles (inside or outside the CNTs) [14]. We wonder if similar effects are present in smaller nanotubes.

In this study, we chose CNTs with an inner diameter of 1-3 nm and considered selective hydrogenation of CAL as a probe reaction to explore the effect of confinement on the product selectivity. Because CAL contains conjugated C=C and C=O double bonds, CAL hydrogenation could possibly lead to three different products: cinnamyl alcohol (COL), hydrocinnamaldehyde (HCAL), and hydrocinnamyl alcohol (HCOL) (Scheme 1). Therefore, we encapsulated Ru nanoparticles within the channels of CNTs and also deposited them on the outer walls of the same CNTs to investigate how the product distribution is influenced by confinement.

All the chemicals used in this study were analytical grade and they were not subjected to further treatment.

Raw CNTs were purchased from Chengdu Organic Chemicals and purified in HNO₃ and H₂SO₄ (Tianjin Kermel Chemicals Co., Ltd., China) (volume ratio = 1:3) for 4.5 h at 40-50 ℃, followed by thorough washing with deionized water. The CNT aqueous dispersion was then subjected to freeze drying for 84 h. Subsequently, the sample was heated to 1000 ℃ at a rate of 5 ℃/min in Ar atmosphere and kept at this temperature for 4 h.

Volatile bis(2, 4-dimethylpentadienyl)ruthenium (99%, Strem Chemicals) was used as the precursor of Ru. The CNTs were evacuated to 10^-4 Pa for 16 h at 450 ℃ before exposure to the precursor vapor. The mixture was kept at 120 ℃ for 48 h to allow diffusion of the precursor into the CNT channels. The catalyst was then reduced in H₂ (60 mL/min) at 230 ℃ for 3 h. The obtained catalyst is denoted Ru@CNT.

For comparison, Ru was also dispersed on the exterior walls of CNTs by the impregnation method. Bis(2, 4-dimethylpentadienyl)ruthenium dissolved in toluene was used. After drying at 60 ℃ for 12 h, the catalyst was subjected to reduction under the same conditions as those used to produce Ru@CNT. The resulting catalyst is called Ru/CNT.

Hydrogenation of CAL was performed in a 300-mL Parr high-pressure reactor equipped with a stirrer, a thermocouple, a gas inlet, and a vent. In each experiment, 1 g of CAL (Sinopharm Chemical Reagent Co., Ltd, China), 100 ml of isopropanol (Tianjin Kermel Chemicals Co., Ltd., China), and 10 mg of catalyst were added into the reactor. After the reactor was purged three times with high purity H₂ (99.999%), it was heated to 100 ℃. The reactor was then pressurized to 2.0 MPa with H₂. The reaction was performed for 2 h. The products were analyzed by gas chromatography (GC)-mass spectroscopy (MS) and quantified by a gas chromatograph (Agilent 7890 A) equipped with a flame ionization detector and a DB-5 capillary column, and 7000 triple quadrupole MS.

Transmission electron microscopy (TEM) was performed with an FEI Tecnai F30 microscope operated at an accelerating voltage of 300 kV. The catalyst was well dispersed in ethanol under ultrasonic treatment before the measurements. Energy dispersive X-ray (EDX) analysis of the catalysts was performed using the same instrument.

The Ru loading of the catalysts was measured by inductively coupled plasma-optical emission spectroscopy (ICP-OES) on a PerkinElmer ICP-OES 7300DV. The catalysts were first oxidized in air at 600 ℃ for 2 h and Ru was then dissolved in aqua regia (HNO₃:HCl = 1:3) in a microwave reactor (MARS, CEM Inc.).

Raman spectroscopy was performed with a LabRam I confocal microprobe Raman instrument (Dilor) at an excitation wavelength of 632.8 nm.

H₂-temperature programmed desorption (H₂-TPD) was performed with a Micromeritics Autochem Ⅱ chemisorption analyzer. The catalyst charge was 8 mg for all tests. Prior to tests, the pre-reduced catalysts were further reduced in situ in H₂ at 230 ℃ for 3 h. After exposure to H₂ for 1 h at 20 ℃, the catalysts were purged by He until a stable baseline was obtained. They were then heated to 800 ℃ at a rate of 5 ℃/min with the effluent monitored by an online thermal conductivity detector (TCD). Temperature programmed reduction (TPR) was also performed with the Micromeritics Autochem Ⅱ chemisorption analyzer. Before the tests, the catalysts were first dehydrated at 120 ℃ in flowing He for 1 h. The catalysts were then reduced in 10% H₂-90% Ar. Consumption of H₂ was monitored by TCD while the temperature was increased to 550 ℃ at a rate of 5 ℃/min.

The N₂ adsorption-desorption experiments were performed with a Quantachrome Autosorb iQ2 automated gas sorption analyzer. The samples were dehydrated at 300 ℃ for 6 h before analysis.

Because pristine CNTs have capped ends and contain impurities, they were subjected to acid treatment before introducing the metal nanoparticles into the channels. The Raman spectra in Fig. 1 show that the purified CNTs contain the two characteristic D and G bands of CNTs, i.e., the peaks at about 1355 and 1585 cm^-1, respectively. The D band is related to disorder and defect structures, and the G band is from vibration of sp² hybridized carbon atoms. The intensity ratio of the D band to the G band (I_D/I_G) is frequently used as an indicator of the content of defects or disordering degree of a sample. It is estimated to be 0.04 for the pristine CNTs and 0.09 for the CNTs after purification, indicating that some defects were introduced during the purification process even though the purified CNTs were treated at 1000 ℃ in Ar. This is consistent with a previous study [15].

Table 1 shows that the specific surface area of the purified CNTs is significantly larger than that of the pristine CNTs because of the open tubes and defects generated upon acid treatment. Interestingly, upon deposition of Ru nanoparticles on the exterior walls, the surface area decreases to 402 m²/g while the surface area of Ru@CNT further decreases to 270 m²/g, indicating that more surfaces become inaccessible when Ru nanoparticles are introduced into the CNT channels.

TEM characterization indicates that the CNTs are shortened to small segments with lengths in the range 5-17 µm by acid treatment, and most of the ends are open. In a previous study [16], we showed that it was fairly easy to introduce noble metal nanoparticles into the channels of nanotubes with an inner diameter larger than 3 nm using a wet chemistry method based on capillary forces. However, the wet chemistry method is not applicable for the CNTs in this study, which have inner diameters only 1-3 nm. A large number of the metal nanoparticles were deposited on the exterior walls of the CNTs. Therefore, we adapted our previously reported method and used a volatile organo-metallic molecule as a precursor [17]. By evacuating the sample before exposure to the vapor of the organo-metallic molecule, the precursor can be driven into the channels. After reduction in H₂, the catalysts were characterized. EDX analysis shows the presence of Ru species in both samples (Fig. 2(a) and (b)). The TEM images in Fig. 2(c) show that discrete Ru nanoparticles have been introduced into the channels of the CNTs. The particle sizes are in the range 1.0-2.0 nm (Fig. 2(e)), which are smaller than the CNT inner diameter. This ensures sufficient space for transport of the reactants and products. However, randomly scanning across the specimen, there are still some similarly sized particles remaining on the exterior walls of CNTs. In comparison, for the Ru/CNT catalyst, the Ru particles disperse well on the exterior walls of the CNTs (Fig. 2(d)). They are slightly larger and have a wider size distribution than Ru@CNT, but most of them are still in the range of 1.0-2.0 nm (Fig. 2(f)).

Interestingly, the Ru/CNT catalyst only catalyzes hydrogenation of the olefinic C=C bond, leading to formation of HCAL, as shown in Fig. 3. Although hydrogenation of C=C is expected to be more facile because of the lower bond energy [18, 19], it is still surprising that almost no other products are detected over Ru/CNT. Note that a blank experiment in the absence of any catalyst under the same reaction conditions did not produce any detectable products. In comparison, Galvagno et al. [20] used activated carbon supported Ru for this reaction and observed a variety of products, including HCAL, COL, and other side products. Castillejos et al. [14] also observed formation of considerable amounts of COL (59% selectivity) and HCOL (8%) in addition to HCAL (33%) over a PtRu catalyst with 90% particles located on the exterior walls of CNTs.

When the Ru@CNT catalyst is used, the reaction yields all three products (i.e., HCAL, COL, and fully saturated HCOL) with comparable selectivity. This indicates that locating Ru within the CNTs facilitates hydrogenation of the C=O bond and also enhances the complete hydrogenation. Although it has been reported that less dispersed Ru (i.e., larger particles) favors C=O hydrogenation [20], this effect can be excluded here because Ru/CNT and Ru@CNT have similar Ru particle sizes. Furthermore, Fig. 2 shows that the particle sizes of the catalysts after the reaction do not significantly change compared with those of the fresh catalysts. Facilitated hydrogenation of C=O within the CNT channels compared with that on the outside of the CNTs is consistent with the finding of Castillejos et al. [14], although they used much larger nanotubes. They found that both the selectivity of COL and the activity increased with increasing percentage of PtRu bimetal nanoparticles inside the CNT channels. They attributed the enhanced activity to enrichment of the reactant [14]. It seems reasonable to understand formation of HCAL over Ru@CNT by comparing the performance of Ru@CNT and Ru/CNT because there are still some Ru nanoparticles remaining on the exterior walls as shown by TEM analysis. These Ru nanoparticles outside the CNTs should be as active for hydrogenation of C=C bonds as those on Ru/CNT. However, the effect of confinement within the small CNTs on the product selectivity needs to be further investigated.

The TPR profiles for Ru/CNT and Ru@CNT are shown in Fig. 4(a). There are several reduction peaks in both profiles, which indicates the presence of different types of Ru species. The profile for Ru@CNT has a main reduction peak at 170 ℃, which is accompanied by a very broad and weak peak below 300 ℃. In comparison, for Ru/CNT, there is an intense reduction peak at 190 ℃ and two broad peaks below 300 ℃. The lower reduction temperature of Ru@CNT than Ru/CNT indicates a facilitated reduction, which is consistent with our previous findings for Fe [6] and Ru confined in multiwalled carbon nanotubes with inner diameters of 4-8 nm [8].

It has been reported that upon decoration of CNTs with Ru nanoparticles, the work function of the composite is lowered because of electron transfer from Ru to the CNTs [21]. In a previous study, we showed that Ru species probably transferred fewer electrons to the exterior walls of CNTs than those located on the interior walls because of the host-guest interaction caused by the unique electronic structures of curved graphene layers [8]. Thus, the outside Ru particles could exhibit higher electron density than the inside Ru species. It is therefore reasonable to assume the same for the present catalysts considering the even larger curvature of the nanotubes used in this study. Electron-deficient Ru can adsorb reactants with electron-donating groups, and vice versa. Several previous studies have shown that the increased electron density on Ru lowers the probability for C=C activation and favors activation of the C =O bond [20, 22]. For Ru@CNT and Ru/CNT, there are likely other factors that also play important roles, in addition to electronic effects. For example, the steric effect may play an important role within the nanotubes with an inner diameter of only 1-3 nm. In particular, CAL molecules may not enter channels with inner diameter smaller than 1 nm. In addition, within such a small space, the steric effect may force the molecule to adsorb in a certain configuration that facilitates hydrogenation of the C=O groups, although this requires further investigation.

The H₂-TPD profiles in Fig. 4(b) show several desorption peaks for both Ru/CNT and Ru@CNT. In addition, desorption occurs at a slightly lower temperature for Ru@CNT than for Ru/CNT. Integration of the peak areas indicates that Ru@CNT adsorbs much more H₂ than Ru/CNT (0.62 for Ru@CNT versus 0.37 for Ru/CNT). This cannot be attributed to the metal loading and dispersion of Ru nanoparticles because the Ru loading of Ru@CNT is 3.04% while it is 3.40% for Ru/CNT, and the two catalysts have similar particle size distributions. It is more likely because of enrichment of H₂ in the CNT nanochannels. In a previous theoretical study [23], we predicted that H₂ could be enriched within CNT channels with a diameter of 1.4 nm by 2 times. This enrichment could become stronger within smaller nanotubes owing to the stronger interaction between the molecules and the interior walls of the CNTs [23]. By TPD, infrared spectroscopy, and theoretical analysis, Kondratyuk and co-workers [24] also showed that 1-heptene was enriched within the CNT channels with an inner diameter of 1.4 nm. The H₂ enrichment could enhance the hydrogenation rate and may explain facilitation of complete hydrogenation to give HCOL.

Ru nanoparticles have been introduced into CNT channels with an inner diameter of 1-3 nm by a simple procedure using a volatile organo-ruthenium complex as the Ru precursor. Considering selective hydrogenation of CAL as a probe reaction, confinement within the narrow channels can modulate the product selectivity. The more accessible Ru nanoparticles dispersed on the exterior walls of the CNTs allows selective formation of HCAL. In comparison, Ru nanoparticles located within the CNT channels facilitate hydrogenation of the C=O bond in addition to complete hydrogenation. Although the underlying mechanism requires further study, the concept of the confinement effect can be used for rational design of catalysts for analogous selective hydrogenation reactions.