1 Introduction

Metal particles supported on various oxides are widely used as catalysts in chemical transformations [1]. As the size of a metal particle decreases, especially below 5 nm, more surface atoms become available for surface reactions with the same amount of material. In addition, as a metal particle becomes smaller, its other properties, such as particle morphology, particle electronic structure and particle-support interaction could also change, which all in turn could affect its catalytic activity [2]. Particle size effects on a variety of reactions have been widely studied because of their fundamental and practical interests. Fischer–Tropsch synthesis on supported cobalt catalysts [3], the hydrogenation of alkyl alcohol on supported palladium catalysts [4], low temperature CO oxidation by gold nanoparticles [5], electro-oxidation of methanol on supported platinum catalysts [6], and CO oxidation on supported ruthenium catalysts [7] are listed as examples.

Supported Ni catalysts, due to their high activity in hydrogenation/dehydrogenation reactions and relatively low price compared to noble metals, have attracted great attentions in terms of studying their particle size effects on catalytic activities. Sinfelt and co-workers [8] found that the specific catalytic activity of nickel decreased as the crystallite size increased in ethane hydrogenolysis reaction. For the same reaction, Goodman and co-workers [9] found that the reaction rates over Ni/SiO2 catalysts increased with increasing particle size, reaching a maximum at 2.5 nm and then decreasing as the particle size further increased. Pina et al. [10] found that the direct hydrogenation of phenol was largely insensitive to Ni particle size. For benzene hydrogenation to cyclohexane reaction, Martin and Dalmon [11] found that, as the Ni particle size increased from 2.5 to 6 nm, their activity increased.

While in earlier studies the dehydrogenation of cyclohexane to benzene was found to be a structure insensitive reaction over supported Pt catalysts [12, 13], Somorjai and co-workers [14] found that it was structure sensitive on Pt single crystal surfaces. Tsai et al. [15] also found that the conversion of cyclohexane to chemisorbed benzene was structure sensitive on different Pt surfaces. On Ni–Cu bimetallic catalysts, Sinfelt et al. [16] showed that cyclohexane dehydrogenation was insensitive over a wide range of alloy compositions. Desai and Richardson [17] found that the cyclohexane dehydrogenation activity slightly decreased by a factor of two when the average Ni particle size increased from 2 to 4 nm. To our best knowledge, a consistent study of the particle size effects on the cyclohexane dehydrogenation on Ni model catalysts is still missing. The structure effects on reaction activity are often observed in the particle size range of 1–5 nm, where the fraction of surface sites with different coordination numbers changes more drastically [18]. Because of the complexity of real catalysts and the sensitivity of Ni catalysts to oxygen and carbon contaminations compared to noble metals, it is difficult to establish a reliable correlation between the observed activity and the Ni particle size. In this letter, we report our recent study of Ni particle size effect on cyclohexane dehydrogenation reaction under elevated pressure conditions, using model Ni/SiO2/Mo(112) catalysts.

2 Experimental

The chamber used for the reaction kinetics, temperature programmed desorption (TPD) and polarization-modulation infrared reflection absorption spectroscopy (PM-IRAS) measurements has been described in detail previously [19]. Briefly, the stainless steel ultrahigh vacuum (UHV) system is equipped with Auger electron spectroscopy (AES), low energy electron diffraction (LEED) and a UTI 100 mass spectrometer. The bottom of the chamber is coupled with a high pressure/IR cell which could be isolated from the main chamber by differentially pumped sliding seals for elevated pressure studies. The base pressure for both main chamber and the high pressure cell is less than 2 × 10−10 Torr. The high pressure cell is equipped with CaF2 windows for PM-IRAS measurements. PM-IRAS (Bruker Equinox 55, Hinds PEM-90 photoelastic modulator) measurements were conducted using 600 scans at a resolution of 4 cm−1. The temperature ramp rate was 3 K/s for TPD measurements.

The Mo(110) sample was mounted to tantalum wires at the bottom of the sample manipulator. The sample temperature could be varied from 80 to 2,300 K by combined resistive/e-beam heating and liquid nitrogen (LN2) cooling. The sample temperature was measured using a W–5 %Re/W–26 %Re C type thermocouple spot-welded on the back of the Mo single crystal. The Mo(110) was cleaned by repeated cycles of annealing at 1,200 K in 1 × 10−6 Torr O2 with a subsequent flash to 2,200 K until no contamination was detected by AES. Ultra-thin SiO2 films (~10 nm) were prepared by depositing Si in 1 × 10−5 Torr O2 at room temperature and subsequently annealed to 1,200 K for 10 min [20]. Ni was evaporated from a home-made doser consisting of a tantalum wire filament wrapped with a Ni wire. The Ni deposition rate was calibrated using Ni TPD on Mo(110) and the saturation of the first layer desorption was taken as one monolayer (1 ML) [21]. All Ni/SiO2 samples were annealed at 600 K for 5 min before any measurement.

C.P. grade CO (Matheson Tri-Gas) was further purified by passing through a LN2 cooling trap. Ultrahigh purity H2 and O2 (Matheson Tri-Gas) were used as received. Cyclohexane was purified by repeated freeze–pump–thaw cycles. For kinetics studies, the cyclohexane dehydrogenation reaction was conducted in a batch reactor, and the conversions were controlled below 5 %. The post reaction products were analyzed using a mass spectrometer. The cyclohexane-to-benzene molar ratio was calculated using corresponding mass spectra calibrated by their relative sensitivities.

Scanning tunneling microscopy (STM) experiments were conducted in a separate UHV chamber equipped with a room temperature Omicron UHV STM-1 system, an Auger electron spectroscopy and a LEED system. All STM images were obtained at the constant current mode with an electrochemically etched tungsten tip. The silica film was prepared via the same method mentioned above. The only difference is the thickness of the silica film. In order to represent bulk silica while make the surface conductive enough for STM studies, multilayer silica films were prepared. In our previous STM studies, one layer of SiO2 films (3 Å) was grown on Mo substrates [22, 23]. The thickness of the silica film in the current study is estimated to be 5 nm, which is about one half as thick as the silica film used for reaction studies.

To accurately determine the Ni particle size, the STM measured nominal particle diameter was corrected for the tip convolution effects that lead to apparent particle enlargement. The detailed information of the correction can be found in our previous publications [22, 23]. In brief, the measured particle diameter was calibrated based on two assumptions: (1) the sticking probability of Ni on the thin silica film is one, and (2) for a given image, all Ni particles’ diameters are exaggerated by the same factor, χ, but this factor was allowed to vary from image to image. The factor χ can be obtained by solving an equation based on the idea that the total calculated volume of Ni particles should equal to the total deposition volume of Ni. The value of χ varies from 0.36 on a surface with the average Ni particle size of 0.96 nm to 1.0 on a surface with the average Ni particle size of 5.2 nm. The trend of χ variation is consistent with the intuition that the apparent diameter of a small particle is exaggerated more than that of a large one. The value of χ on each of four surfaces with different average Ni particle sizes is listed in Fig. 2.

3 Results and Discussion

Ni particle sizes were controlled by varying the vacuum deposition times of Ni onto the SiO2 substrate. The morphologies and particle sizes of Ni on SiO2/Mo(110) were studied by STM with Ni dosages varying from 0.25 to 10.0 ML. The Ni particle size distribution was analyzed based on a randomly chosen 100 × 100 nm area at each dosage. Figure 1 shows the STM images of Ni model catalyst surfaces and Fig. 2 show Ni particle size distributions at Ni dosages of 0.25, 1, 2 and 10 ML, respectively. At 0.25 ML Ni dosage, Ni clusters are highly dispersed on the silica film (Fig. 1a) with the average particle size, ‹dp›, of 0.96 nm and the standard deviation, σ, of 0.33 nm (Fig. 2a). The number density of Ni particles is 8.0 × 1012/cm2. With increasing Ni dosage, the particle number density is increased to 1.0 × 1013/cm2 on a 0.5 ML Ni/SiO2 surface (image not shown), but decreased to 9.2 × 1012/cm2 on a 1 ML Ni/SiO2 surface (Fig. 1b) and further to 8.1 × 1012/cm2 on a 2 ML Ni/SiO2 surface (Fig. 1c). Both size and height of the Ni particles are increased accordingly with increased Ni coverage. The average Ni particle size and standard deviation are 1.7 and 0.57 nm for 1 ML Ni/SiO2 surface (Fig. 2b) and 2.5 and 0.96 nm for 2 ML Ni/SiO2 surface (Fig. 2c). Compared with the surfaces of Rh/SiO2 model catalysts [22], the average Ni particle sizes are smaller and the particle size distributions shift heavily to the small particle size side at the same metal dosages. Even though we can not rule out the effects of SiO2 film thickness on the particle’s surface nucleation and diffusion behaviors, we believe that the variations in particle size distributions for these two metals are most likely caused by the stronger Ni–SiO2 interaction than that of Rh-SiO2. As the Ni coverage was increased to 8.0 and 10.0 ML (Fig. 1d), Ni particles almost covered the whole surface. The morphologies are composed of small Ni clusters on top of large Ni islands. The average Ni particle size and the standard deviation are 5.2 and 1.7 nm on the 10 ML Ni/SiO2 surface (Fig. 2d).

Fig. 1
figure 1

Room temperature STM images of Ni/SiO2/Mo(110) surfaces at Ni dosage of a 0.25 ML, b 1.0 ML, c 2.0 ML, d 10.0 ML, respectively. All the images have the size of 100 × 100 nm2. The thickness of SiO2 films is estimated to be 5.0 nm

Full size image
Fig. 2
figure 2

Ni particle size histograms for a 0.25 ML, b 1.0 ML, c 2.0 ML, d 10 ML Ni on the SiO2/Mo(110) substrate. The average Ni particle size, ‹dp›, the standard deviation, σ, and the correction factor for STM tip convolution effects, χ, for each Ni coverage are reported in the inset in each sub-figure

Full size image

The Ni particle sizes and standard deviations as a function of Ni coverage are plotted in Fig. 3a. The average Ni particle size increases sharply till it reaches 2.5 nm (corresponding to a 2.0 ML Ni dosage), and levels off when it reaches ~5 nm.

Fig. 3
figure 3

a Average Ni particle size versus Ni surface coverage measured by STM for Ni dosage varying from 0.25 ML to 10 ML on SiO2/Mo(110). The error bar represents the standard deviation (±σ) of each particle size. b Relative CO TPD intensity as a function of Ni coverage

Full size image

With Ni coverage increasing from 0 to 10 ML, the total number of surface sites on each Ni/SiO2 model catalyst was calibrated by CO TPD measurements, as shown in Fig. 3b. The integrated CO TPD peak intensity was obtained, then the relative intensity ratio was calculated by normalizing it to the intensity of a 10 ML Ni/Mo(110) film. Since multilayer Ni films on a Mo(110) substrate mainly expose the most stable (111) facet [24], it is reasonable to assume that the surface atomic density of the 10 ML Ni/Mo(110) film equals to that of the Ni(111) surface, or 1.8 × 1015 atoms/cm2. CO TPD results indicate that the number of surface Ni active sites can be tuned continuously by changing the Ni coverage on SiO2 support.

Figure 4 shows the benzene turnover frequency (TOF) (number of benzene molecules formed per active site per second) as a function of Ni particle size for cyclohexane dehydrogenation reaction over Ni/SiO2 model catalysts. The reactions were carried out at 600 K using a 1:50 benzene/hydrogen mixture at a total pressure of 25 Torr in a batch reactor. The benzene TOFs were calculated using the surface active sites estimated from the CO TPD measurements (Fig. 3b). As shown in Fig. 4, the benzene TOF value sharply increases with the decrease of Ni particle size below 2.5 nm. The benzene TOF on the 1.0 nm Ni/SiO2 catalyst is 2.7 s−1, which is about five times higher than that on the 2.5 nm Ni/SiO2 catalyst. The benzene TOF values almost keep constant at ~0.5 s−1 on Ni/SiO2 with particle sizes above 2.5 nm. The benzene TOF value as a function of Ni particle size shows two regions. The cyclohexane dehydrogenation over Ni surface is a structure sensitive reaction with Ni particle below 2.5 nm, and a structure insensitive one with Ni particle size above that. This observation agrees well with the literature reported results. For example, Desai and Richardson [17] reported that the cyclohexane dehydrogenation activity decreased by a factor of two with the average Ni particle size increased from 2 to 4 nm.

Fig. 4
figure 4

Benzene turnover frequency (TOF) as a function of Ni particle size. Reaction conditions: H2:C2H12 = 50:1, total pressure = 25 Torr at T = 600 K for 1 h

Full size image

The electronic properties and surface structures of metal nanoparticles change with decreasing particle sizes, which therefore could influence the reactant adsorption, intermediate formation and product desorption. In order to better understand the Ni particle size effect on cyclohexane dehydrogenation reaction, H2 and CO TPD were carried out on Ni/SiO2 model catalysts at saturated H2 and CO surface coverage, as shown in Figs. 5 and 6, respectively. On the 1.0 nm Ni/SiO2 surface, two H2 desorption peaks were observed at 230 and 350 K. The effects of defect sites on the thermal desorption of H2 were studied on a Ni(111) surface [25]. Two activated adsorption states, β1 and β2, give two desorption peaks at 346 and 384 K respectively on a defect scarce Ni(111) surface. Surface defects introduce a non-activated H2 adsorption/desorption channel to β1 and β2 states whose desorption positions are both shifted to lower temperatures at different degrees depending on the density of defects [25]. Our H2 TPD features from the 1.0 nm Ni/SiO2 surface resemble more the defect-rich, stepped Ni(445) surface [25] than the gently sputtered Ni(111) surface which possesses less density of defect sites. The surfaces of small Ni nanoparticles are mainly composed of low coordinated corner and edge atoms, which are more similar to high index surfaces to a certain extent. With the particle size increasing to 1.7 nm, a new H2 desorption peak developed at about 300 K, as shown in Fig. 5. We tentatively assign this newly developed desorption peak to H2 desorption from Ni(100) facets on a large particle [26, 27]. As the particle size increases, the Ni nanoparticles could develop into, for example, truncated octahedron shapes, consisting of (111) and (100) facets separated by edges [28]. From Fig. 5, we can see that, with increasing Ni particle sizes, in addition to the development of the ~300 K desorption feature, the other two H2 desorption features shift to higher temperatures to different degrees due to the well-defined (111) and (100) facets being developed on large particles.

Fig. 5
figure 5

H2 TPD from H2 saturated Ni/SiO2 surfaces (H2 dosed at 110 K). The average Ni particle sizes for the three surfaces are listed in the figure

Full size image
Fig. 6
figure 6

CO TPD from CO saturated Ni/SiO2 surfaces (CO dosed at room temperature). The average Ni particle sizes for three surfaces are listed in the figure

Full size image

Figure 6 shows CO TPD spectra from Ni/SiO2 surfaces with Ni particle sizes changing from 1.0 to 3.8 nm. On small Ni nanoparticles (1.0 and 1.7 nm), CO desorption reaches a maximum at 349 K with a shoulder at 387 K. On the surface of 3.8 nm Ni particles, the low temperature desorption peak maintains its position at 349 K while the high temperature peak shifts from 387 to 402 K. In addition, the intensity of the high temperature CO desorption peak increases more rapidly than that of the low temperature peak, and it becomes the dominant feature in the CO TPD spectra. The high temperature CO TPD feature for the 3.8 nm Ni particle desorbs at the similar temperature as that from a Ni(111) surface at saturated CO coverage [29]. We assign this high temperature peak to CO desorption from the terrace-like sites on supported Ni nanoparticles. It is generally believed that on metal surfaces CO binds more strongly to the low coordinated step sites than terrace sites. Therefore, CO usually desorbs at a higher temperature from step sites than from terrace sites. This has been demonstrated on vicinal Pt(221) and Pt(335) single crystal surfaces [30] and Pt particles supported on SiO2. However, on Ni particle surfaces, we assign this low temperature feature to CO desorption from low coordinated step-like sites, i.e. edge and corner sites on Ni particles. Benndorf and Meyer [31] showed that CO thermal desorption from step sites occurred at lower temperature than that from terrace sites on a Ni(331) surface. Stair [32] categorized a metal surface as an acid or a base with respect to an adsorbate according to whether the metal accepts or donates electrons during adsorption, which can be determined by work function changes. The fact that CO adsorption on Ni surfaces induces a working function increase [33] indicates that Ni surfaces act as a base when they interact with CO, i.e. Ni donates net electrons to adsorbed CO molecules. Furthermore, the step sites of a metal surface are positively charged and the terrace sites are negatively charged because of the Smoluchkowski [34] effect. Therefore, Ni step sites have less electron donation capability than terrace sites, resulting in a weaker CO bond at step sites. Again, the continue increase in the terrace-like to step-like CO desorption intensity ratio with increasing Ni particle size agrees with the model of terrace development on large particles.

Figure 7 shows CO PM-IRAS spectra on Ni/SiO2 model catalysts with different Ni particle sizes. To understand CO infrared spectra on Ni model catalysts, we refer to previous studies on single crystal surfaces with well-defined structures and CO surface coverage. CO adsorption geometries and CO over layer structures on various Ni single crystal surfaces are strongly coverage dependent. At low surface coverage, CO adsorbs on both threefold and twofold bridge sites on Ni(111) (θCO < 0.2 ML) [35–37] and on atop and bridge sites on Ni(001) (θCO < 0.5 ML) [38]. At high surface coverage (θCO > 0.5 ML), compressed CO over layers are discovered on both Ni(111) [33] and Ni(001) [39] surfaces, and both atop and bridge sites are occupied on Ni(111) [36, 37] and Ni(001) [38] surfaces. To mitigate the complexity caused by multiple-type CO adsorption sites and incommensurate CO over layer structures on the positions and line widths of CO stretch peaks, we study CO infrared spectra with the coverage around 0.5 ML, which can be obtained by saturating Ni particles with CO in UHV around 300 K.

Fig. 7
figure 7

PM-IRAS of saturated CO adsorption on Ni/SiO2 surfaces at room temperature. The average Ni particle sizes for three surfaces are listed in the figure

Full size image

On the surface of small Ni nanoparticles (1.0 nm), CO mainly adsorbs on atop Ni sites with a CO stretching frequency of 2,083 cm−1, accompanied by a broad bridge site CO adsorption band centered at 1,900 cm−1. With an increase in particle size, the frequency of atop adsorbed CO does not change, while the bridge site CO adsorption peak shifts to higher frequencies (blue shift). On the surface of 3.8 nm Ni particles, the bridge site CO adsorption peak shifts to 1,952 cm−1. The blue shift of bridge site CO vibrational frequency can be explained by the increasing CO molecule dipole–dipole interactions on the terrace-like sites of large particles.

As the Ni particle size increases, the intensity ratio of the bridge CO peak to that of atop CO peak increases. At saturated coverage, CO stretch frequencies on Ni(111) and Ni(100) have been well documented. On Ni(111), CO adsorbs exclusively onto twofold bridge site with a frequency of 1,908 cm−1, while forms an atop band at 2,048 cm−1 and a bridge band at 1,949 cm−1 on Ni(100) [9, 38]. With the increase of Ni particle size, the fractions of (111) and (100) facets, the two energetically most favored surfaces of an fcc single crystal metal, on a particle surface increase, resulting in enhanced intensities of the bridge site CO adsorption. The reasons for the observed relatively strong CO atop band on Ni particles are unclear at the present time. At saturation coverage, In addition to the fact that CO does not adsorb terminally on Ni(111), it adsorbs on atop and bridge sites on Ni(100) with the peak intensity ratio of atop to bridge sites being 2:1 [38]. We postulate that CO adsorbs terminally on the edge and corner sites on Ni particles, resulting the relatively intense atop adsorption band on Ni particles. The position of Ni atop band on Ni particle centers at 2,083 cm−1, about 35 cm−1 higher than that of the atop band on Ni(100) at saturation coverage. This blue shift of atop CO band on edge sites is consistent with our observed lower CO thermal desorption temperature on edge sites. The reduced Ni d-orbitals to CO 2π*-orbital electron back-donation at the Ni edge sites stiffens the CO stretch vibration.

Combining CO TPD and PM-IRAS results, it can be concluded that on small Ni particles, CO mainly adsorbs on low coordinated Ni atoms on atop sites, and desorbs at 349 K. On large Ni nanoparticles, CO adsorbs on both low coordinated Ni atoms and well developed Ni(111) and Ni(001) facets, desorbing at 349 and 402 K respectively.

The CO TPD results are also consistent with the H2 TPD results, which show that small Ni particle surfaces are mainly composed of low coordinated Ni atoms, whereas there are both low coordinated Ni atoms and well developed low index (111) and (100) facets on large Ni particles. The higher activity of small Ni nanoparticles could be attributed to their higher fraction of low coordinated Ni atoms. The H2 formed from the dehydrogenation of cyclohexane desorbs more easily from low coordinated Ni atoms on small nanoparticles, shifting the surface reaction toward the product formation direction.

4 Conclusions

We have studied cyclohexane dehydrogenation over Ni/SiO2 model catalysts under high pressure conditions. The surface reactivity (TOF of Benzene) of large Ni nanoparticles (>2.5 nm) was found to be independent of Ni particle size, consistent with the structure insensitivity reported in the literature. The specific reaction rate sharply increases with the decrease of Ni particle size below 2.5 nm. H2 and CO TPD show that the surfaces of small Ni nanoparticles are mainly composed of low coordinated Ni atoms. The easier desorption of H2 on low coordinated Ni atoms makes more sites available for cyclohexane dehydrogenation reactions. The higher activity of small Ni nanoparticles is attributed to the high fraction of low coordinated Ni atoms on small particles. The present results show that metal clusters supported on oxide thin films could be an ideal model system to simulate technical catalysis systems.