Introduction

Styrene is a very important precursor in the petrochemical industry and has been widely used for production of plastics and synthetic rubbers such as polystyrene, acrylonitrile–butadiene–styrene resins and styrene-butadiene rubber [1]. The global market value of styrene is expected to grow from $48.8 billion in 2020 to $69.9 billion in 2027 [2]. Nowadays, the main processes of styrene production are ethylbenzene dehydrogenation, styrene-propylene oxide co-production, and extraction from pyrolysis gasoline (pygas, a by-product of steam cracking of naphtha) [3,4,5]. However, a small amount of phenylacetylene is always present in the styrene feedstock, which will poison downstream polymerization catalysts [6, 7]. The similar properties of phenylacetylene and styrene make it difficult to separate them by distillation. At present, selective hydrogenation of phenylacetylene is the most efficient method to remove phenylacetylene from styrene, with the aim of achieving complete conversion of phenylacetylene while avoiding over-hydrogenation of styrene as much as possible. The main challenge for this process is the development of effective catalysts with high selectivity and stability.

To date, noble metal Pd-based catalysts have been the dominant materials in the selective hydrogenation of alkynes because of their excellent hydrogenation activity [8, 9]. However, due to the strong adsorption ability of Pd active site for the intermediate product–styrene, styrene cannot be desorbed in time but is further hydrogenated to ethylbenzene. Therefore, the selectivity to styrene on monometallic Pd catalysts typically decreases rapidly as the reaction approaches complete conversion of phenylacetylene, resulting in loss of styrene [10,11,12,13]. Various methods have been developed to enhance the selectivity of Pd-based catalysts, including alloying Pd with relatively inert metals such as Cu [14, 15], Zn [15, 16] and Pb [17, 18], tuning metal-support interaction [10, 19], and confining Pd nanoparticles in porous materials [20, 21]. In fact, the commercial Lindlar catalyst used for this reaction is 5 wt% Pd/CaCO3 modified with lead acetate or quinoline, which has environmental concerns and is also constrained by the scarcity and high cost of the precious metal Pd [22, 23].

Among non-precious metal alternatives, Ni-based catalysts are the most extensively studied due to their relatively high hydrogenation activity. In order to improve the selectivity to styrene over Ni-based catalysts, alloys/intermetallics are usually formed by combining Ni with another metal (Zn [24, 25], Cu [26], Ga [27, 28], and Sn [28]) or non-metal component (Si [29,30,31] and P [32, 33]) to optimize the geometric and electronic structure of the Ni active site. It is generally reported that the type and doping amount of the second metal have impacts on the particle size of metal nanoparticles, which, however, is not conducive to explore the effect of the second metal because the activity and selectivity of the catalyst in the selective hydrogenation of phenylacetylene is greatly influenced by the metal particle size [24, 27, 33]. Indeed, the conventional preparation of bimetallic alloy catalysts requires that the two metal precursors loaded on the support undergo a high-temperature reduction step in H2 atmosphere, which inevitably leads to the agglomeration of metal nanoparticles [34, 35]. Although the liquid-phase reduction method can be used to prepare alloys, the process is often more complex and the choice of solvent, capping agent, and reducing agent needs to be carefully considered [28, 35]. Therefore, it is necessary to develop a facile synthetic strategy for alloy catalysts with uniform particle distribution and controllable composition.

In 2013, Beller’s group [36] developed an efficient strategy for the preparation of catalysts by high-temperature pyrolysis of iron-phenanthroline complexes supported on carbon in an inert atmosphere. The resulting Fe-phen/C-800 catalyst achieved good to excellent yields in selective hydrogenation of more than 80 diverse nitroarenes. Then, this strategy was extended to the preparation of Co- and Ni-based catalysts and applied to the hydrogenation of various unsaturated compounds including alkynes [37,38,39,40]. Recently, the same group [41] prepared monodisperse nickel nanoparticle catalysts by simple pyrolysis of monosaccharide-nickel acetate templates loaded on silica, which showed high selectivity in the stereo- and chemo-selective hydrogenation of functionalized and structurally diverse alkynes. However, it remains unclear whether this strategy allows the preparation of highly dispersed Ni-based bimetallic alloy catalysts.

In this work, a series of bimetallic Ni–Zn/SiO2 and Ni–Co/SiO2 catalysts were prepared by pyrolysis of glucose and acetates supported on silica and applied to the selective hydrogenation of phenylacetylene to styrene. The effects of Zn/Ni and Co/Ni molar ratios on the structure and performance of the catalysts were investigated and the structure–activity/selectivity relationship was explored. Finally, the reusability of the best catalyst and the effect of reaction temperature and pressure on the catalytic performance were studied.

Experimental section

Materials

Nickel acetate tetrahydrate (Ni(OAc)2·4H2O, ≥ 99%), zinc acetate (Zn(OAc)2, ≥ 99%), cobalt acetate tetrahydrate (Co(OAc)2·4H2O, ≥ 99%), d-glucose (C6H12O6, ≥ 99%), n-octane (C8H18, ≥ 98%), and absolute ethanol (C2H5OH, ≥ 99.7%) were purchased from Adamas Reagent Co., Ltd., China. Phenylacetylene (C8H6, > 98%) was obtained from Alfa Aesar. SiO2 (Hydrophilic-380, particle size of 7–40 nm) was supplied by Makclin Biochemical Co., Ltd., China. All chemicals were used without treatment.

Catalyst preparation

Ni/SiO2 catalyst was prepared by a modified pyrolysis technique [41]. First, 133.8 mg of Ni(OAc)2·4H2O, 150 mg of d-glucose, 10 mL of deionized water, and 30 mL of ethanol were added into a 100 mL round-bottomed flask equipped with a reflux condenser. After stirring for 5 min at room temperature, the flask was immersed in a water bath at 60 °C and stirred for 2 h at 750 rpm. 0.6 g of SiO2 was then added and stirred for another 2 h at the same temperature. Next, solvent was evaporated from the solution using a rotary evaporator at 60 °C until a light-green powder was obtained. The powder was further dried in an oven at 110 °C for 10 h. Finally, the powder was placed in a quartz tube and inserted into an electronic furnace. The powder was heated to 800 °C at a heating rate of 25 °C min−1 and held at 800 °C for 2 h under flowing N2. The obtained black powder was Ni/SiO2 with a nominal Ni loading of 5 wt%.

The Ni–Zn/SiO2 and Ni–Co/SiO2 bimetallic catalysts were prepared by the same method except that Zn(OAc)2 or Co(OAc)2·4H2O was added together with Ni(OAc)2·4H2O. The theoretical molar ratios of Zn/Ni or Co/Ni in the bimetallic catalysts were 0.1, 0.5, 1, and 2. However, the evaporation of Zn during pyrolysis resulted in quite low actual Zn/Ni molar ratios of 0.004, 0.02, 0.05, and 0.07, which were determined by inductively coupled plasma-optical emission spectroscopy. For clarity, the resulting catalysts were denoted by NiZnx/SiO2 or NiCox/SiO2, where x represents the actual molar ratio of Zn/Ni or Co/Ni. Metal-free G/SiO2 (pyrolysis of glucose supported on SiO2) and monometallic Co/SiO2 (nominal Co loading of 5 wt%) were prepared by the same method. In addition, Ni/SiO2-I and ZnO/SiO2-I (both Ni and Zn at 5 wt% nominal metal loading) were prepared by conventional impregnation: nickel acetate or zinc acetate was first impregnated on SiO2, and then dried at 110 °C, followed by calcination at 500 °C. Ni/SiO2-I was further reduced in 30% H2/N2 at 500 °C.

Catalyst characterization

The Brunauer–Emmett–Teller (BET) surface area, pore volume (Vp), and average pore diameter (dp) of samples were acquired from N2 adsorption–desorption isotherms at − 196 °C using a Micrometrics ASAP 2010 instrument. The crystalline structure of catalyst was characterized by X-ray diffraction (XRD, Rigaku D/Max 2550 VB/PC) using Cu Kα radiation (λ = 0.15406 nm) in the range of 2θ = 10°–80° with a step size of 0.02°. The micro-morphology of catalyst was observed by high-resolution transmission electron microscopy (HRTEM, JEOL JEM-2100). The average metal particle size was obtained by measuring at least 100 particles. The surface composition of catalyst was analyzed by X-ray photoelectron spectroscopy (XPS, Thermo Scientific ESCALAB 250 Xi) using the Al Kα radiation (hν = 1486.6 eV) and a pass energy of 40 eV, with the binding energy referenced by C 1s (284.8 eV). The amount of active sites of catalyst was determined by CO-pulse chemisorption (Micromeritics AutoChem 2920), assuming the chemisorption stoichiometry of CO/Ni = 1. Before measurements, the catalyst was reduced at 300 °C for 1 h in 10% H2/Ar, followed by switching to He at 330 °C for 30 min to remove adsorbed H2. After the catalyst was cooled to room temperature in He, CO pulses were injected until the CO peak area remained constant, and the CO uptake was monitored by a thermal conductivity detector. The metal loading of catalyst was measured by inductively coupled plasma-optical emission spectroscopy (ICP-OES, Varian 710-ES). The weight loss of samples before and after pyrolysis was quantified by thermogravimetric analysis (TGA, PerkinElmer TGA-4000) heated from room temperature to 800 °C at 10 °C min−1 in N2 and air flow. The content and structure of carbon in the catalyst were examined by elemental analyzer (Elementar VARIO ELIII) and Raman spectroscopy (Renishaw inVia Reflex, 514 nm excitation wavelength). The structural properties of Ni active sites in the catalyst were probed by diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS, Bruker Tensor 27) after chemisorption of CO. The catalyst was first pretreated at 300 °C for 1 h in 10% H2/Ar, then cooled to room temperature under Ar flow and the background spectrum was recorded. Afterwards, 5% CO/Ar was passed through for 30 min to saturate the surface, followed by purging with Ar for 10 min to remove gaseous and weakly adsorbed CO, and the spectrum was collected by 32 scans at a resolution of 4 cm−1.

Catalyst test

The selective hydrogenation of phenylacetylene over various catalysts was performed in a semi-batch stirred tank reactor (300 mL, Parr 5100). Typically, 0.15 g of catalyst powder, 5 g of phenylacetylene, 5 g of n-octane (internal standard), and 90 g of ethanol (solvent) were first added into the reactor and heated to 60 °C in N2. Then, H2 was charged into the reactor to purge N2 for five times. Finally, the hydrogenation reaction started at 60 °C and 0.5 MPa in H2 with vigorously stirring at 1000 rpm. About 1 mL of liquid phase was discharged from the reactor in a certain interval and analyzed using a gas chromatography (Hewlett-Packard 6890) equipped with a capillary column (PEG-20M, 30 m × 0.32 mm × 0.50 μm) and a flame ionization detector.

The conversion of phenylacetylene and the selectivity to styrene are defined as follows:

$${\text{Conversion}} = \frac{{\text{Moles\, of\, phenylacetylene\, converted}}}{{\text{Moles\, of\, initial\, phenylacetylene}}} \times 100{\text{\% }}$$
(1)
$${\text{Selectivity}} = \frac{{\text{Moles\, of\, styrene\, produced}}}{{\text{Moles\, of\, phenylacetylene\, converted}}} \times 100{\text{\% }}$$
(2)

The catalyst activity is expressed in terms of the initial rate (Rini) that is defined as the number of moles of phenylacetylene consumed per second per gram of catalyst during the initial period. The Rini is calculated by:

$${\text{R}}_{{{\text{ini}}}} = \frac{{{\text{F}}\alpha }}{{{\text{tm}}}}$$
(3)

Here F is the number of moles of phenylacetylene in the feedstock, mol; α is the conversion of phenylacetylene; t is the reaction time, s; m is the mass of catalyst, g. The conversion data used to calculate Rini is below 30%.

Results and discussion

Physicochemical properties of catalysts

Fig. S1a shows the TGA curves of glucose, Ni(OAc)2·4H2O, and the precursors (without calcination treatment) of Ni/SiO2, NiZn0.05/SiO2 and NiCo1.02/SiO2 in N2 flow. For glucose, a weight loss of around 78% occurs between 180 and 420 °C, which is assigned to the pyrolysis of glucose [42]. For Ni(OAc)2·4H2O, two distinct weight loss steps are observed. The weight loss (28.7%) below 120 °C is caused by the removal of crystal water (the weight percentage of crystal water in Ni(OAc)2·4H2O is 28.9%), while the weight loss (33.2%) in the temperature range of 250 to 420 °C is due to the decomposition of Ni(OAc)2. The precursors of Ni/SiO2, NiZn0.05/SiO2, and NiCo1.02/SiO2 exhibit similar TGA curves with a weight loss of about 30% below 500 °C, which is mainly attributed to the decomposition of acetates and glucose. The pyrolysis of glucose in N2 atmosphere can yield reductive agents such as C and CO [43], which in turn can reduce nickel oxides (derived from decomposition of nickel acetate) immediately.

After pyrolysis treatment, Ni/SiO2, NiZn0.05/SiO2, and NiCo1.02/SiO2 are further characterized by TGA in air. As presented in Fig. S1b, the initial weight loss below 130 °C is due to desorption of physisorbed water. After then, a weight gain at 130–320 °C takes place for all the three catalysts, which is attributed to the oxidation of metals. The weight gain of Ni/SiO2 is quite similar to that of NiZn0.05/SiO2, but distinctly lower than that of NiCo1.02/SiO2. This is because: (1) the actual amount of Zn in NiZn0.05/SiO2 is very low (0.3 wt%, Table 1), and the possible contribution from Zn oxidation to the weight gain can be neglected; (2) the actual amount of Co in NiCo1.02/SiO2 is comparable to that of Ni (4.9 wt%, Table 1), and the metallic Co derived from the pyrolysis step is also oxidized, which contributes to the weight gain. For other NiZnx/SiO2 catalysts, the actual Zn content is also much lower than the nominal content (Table 1), which is ascribed to evaporation of zinc during the pyrolysis process at high temperatures [20, 44]. It is known that the boiling point of Ni, Co, and Zn is 2732, 2870, and 907 °C. Therefore, the metallic Ni and Co remain in the catalysts, which justifies the similarity between nominal and actual Co/Ni molar ratios in the catalysts (Table 1). After this weigh gain, a weight loss is observed for all samples, owing to the gasification of carbon. Indeed, the elemental analysis shows that the carbon content in Ni/SiO2 is 5.3 wt%, and the presence of layered carbon is also confirmed by Raman spectroscopy (Fig. S2). However, another weight gain appears on NiCo1.02/SiO2 at above 650 °C, which corresponds to the CoO → Co3O4 process [45].

Table 1 Physicochemical properties of support and catalysts
Full size table

Fig. S3 shows the XRD patterns of NiZnx/SiO2 and NiCox/SiO2 catalysts, and the corresponding local enlargements are given in Fig. 1. The broad peak located at 2θ = 15°–35° for all samples belongs to amorphous SiO2. Ni/SiO2 exhibits diffraction peaks at 2θ = 44.5° and 51.8° (Fig. S3a), corresponding to metallic Ni(111) and Ni(200) planes (JCPDS 04-0850). Co/SiO2 shows diffraction peaks at 2θ = 44.2° and 51.5° (Fig. S3b) that are indexed to metallic Co(111) and Co(200), respectively (JCPDS 15-0806). The presence of metallic Ni and Co peaks further demonstrates the reduction of nickel and cobalt oxides during pyrolysis of glucose in N2. For NiZnx/SiO2 (Fig. 1a), with increasing the Zn/Ni molar ratio, the Ni(111) peak first shifts to a lower angle and then recovers to 44.5° when the Zn/Ni ratio is 0.07, implying the formation of Ni–Zn alloy on some NiZnx/SiO2 [46, 47]. Similarly, with increasing the Co/Ni molar ratio, this peak shifts to a lower angle between 44.5° (Ni(111) plane) and 44.2° (Co(111) plane), suggesting the formation of Ni–Co alloy [48, 49].

Fig. 1
figure 1

Local enlarged XRD patterns of a NiZnx/SiO2 and b NiCox/SiO2

Full size image

Fig. S4 displays the N2 adsorption–desorption isotherms and the pore size distribution curves of SiO2, NiZnx/SiO2, and NiCox/SiO2. All catalysts exhibit type IV isotherms with H3-type hysteresis loops [50], indicating that the mesoporous structure of SiO2 is maintained after introduction of Ni and Zn or Co. In addition, all catalysts except NiCo1.97/SiO2 have a unimodal pore size distribution centered at around 20 nm, whereas NiCo1.97/SiO2 shows a bimodal pore size distribution (about 6 and 20 nm diameter pores). The specific surface area of all catalysts is reduced as compared to the support (Table 1) due to the metal deposition, while the pore volume and average pore size of the catalysts (excluding NiCo1.97/SiO2) have a small variation. For NiCo1.97/SiO2, the total metal loading (15.5 wt%) is much higher than that of other catalysts (5–9.7 wt%), so that more pores of NiCo1.97/SiO2 are blocked, leading to a bimodal pore size distribution and a decreased pore volume.

Fig. S5 shows the HRTEM images and the metal particle size distribution of NiZnx/SiO2 and NiCox/SiO2. The metal particles are uniformly distributed over all catalysts with a narrow particle size distribution. The average metal particle sizes of Ni/SiO2 and Co/SiO2 are 3.8 and 7.7 nm, respectively. By contrast, the average metal particle sizes of bimetallic catalysts increase slightly with increasing the Zn/Ni or Co/Ni ratio, varying from 4.2 to 5.5 nm for NiZnx/SiO2 or from 4.1 to 5.3 nm for NiCox/SiO2 (Table 1). Apparently, both Ni–Zn and Ni–Co bimetallic catalysts have similar metal particle sizes irrespective of variation in composition, so the effect of metal particle size on the catalytic performance can be ignored in the following discussion. Ni/SiO2 displays lattice spacings of 0.203 and 0.176 nm (determined by fast Fourier transform pattern (FFT)), corresponding to the (111) and (200) planes of fcc Ni. For NiZn0.02/SiO2 and NiZn0.05/SiO2, a larger lattice spacing of 0.206 nm is found (Fig. S5a), indicating the formation of Ni–Zn alloy by doping Zn atoms into the lattice of Ni [47, 51]. With further increasing the Zn/Ni ratio, a lattice spacing of 0.214 nm assigned to the ZnO(200) plane (JCPDS 65-0682) is observed on NiZn0.07/SiO2. As for NiCox/SiO2 (Fig. S5b), a lattice spacing of 0.204 nm (between 0.203 nm of the Ni(111) plane and 0.205 nm of Co(111)) occurs in NiCo0.52/SiO2, NiCo1.02/SiO2, and NiCo1.97/SiO2, which results from an expansion of Ni lattice and implies the formation of Ni–Co alloy [48, 52]. For NiCo1.97/SiO2 with a high Co/Ni ratio of 1.97, another lattice spacing of 0.205 nm is identified, implying that not all Co particles are alloyed with Ni. The above results illustrate that the Ni–Zn (or Ni–Co) alloy is formed in some bimetallic NiZnx/SiO2 (or NiCox/SiO2), in accordance with the XRD analysis.

Fig. 2 presents the XPS spectra of Ni 2p core level of NiZnx/SiO2 and NiCox/SiO2. For Ni/SiO2 (Fig. 2a), the Ni 2p3/2 spectrum can be deconvoluted into three peaks with binding energies at 852.7, 856.2, and 860.1 eV, which are assigned to Ni0, Ni2+, and satellite peak [53, 54]. With the incorporation of Zn, the peak of Ni0 shifts to 852.5 eV in NiZn0.02/SiO2 and further to 852.2 eV in NiZn0.05/SiO2 (Fig. 2a). According to the literature [47, 53], this implies electron transfer from Zn to Ni and the presence of electronic effect, although attempts to deconvolute the XPS spectrum of the Zn 2p core level fail because of the low Zn content (Fig. S6a). Note that the peak of Ni0 in NiZn0.004/SiO2 and NiZn0.07/SiO2 remains at 852.7 eV, indicative of no electron transfer between Ni and Zn, which may be due to the very low Zn content and the presence of ZnO. As for bimetallic NiCox/SiO2 (Fig. 2b), the Ni0 2p3/2 peak is located at 852.7 eV when x = 0.09, but with increasing x from 0.52 to 1.97, this peak gradually shifts positively from 852.9 to 853.7 eV; meanwhile, the Co0 2p3/2 peak varies between 778.0 and 778.8 eV (Fig. S6b), lower than that for Co/SiO2 (779.0 eV). This reveals the electron transfer from Ni to Co [52, 55]. The finding from the XPS analysis is consistent with the XRD and HRTEM results. In short, the Ni–Zn alloy is formed in NiZnx/SiO2 (x = 0.02 and 0.05) where Ni atoms are electron-rich, while the Ni–Co alloy is present in NiCox/SiO2 (x = 0.52, 1.02, and 1.97) where Ni atoms are electron-poor.

Fig. 2
figure 2

XPS spectra of Ni 2p for a NiZnx/SiO2 and b NiCox/SiO2 catalysts

Full size image

DRIFTS of CO chemisorption is used to further probe the structural information of Ni in the catalyst. Ni/SiO2 is compared with Ni/SiO2-I (prepared by conventional impregnation), and NiCo1.02/SiO2 is chosen as a representative bimetallic catalyst, considering the high content of the second metal. DRIFTS of CO adsorbed on Ni/SiO2-I shows bands in the range of 2100–2000, 2000–1900, and 1900–1800 cm−1 (Fig. S7), which correspond to linearly, bridged, and multibonded adsorbed CO on Ni [56, 57]. Obviously, the bridged and multibonded adsorption peaks on Ni/SiO2 almost disappear, while the linear adsorption peak at 2042 cm−1 with a shoulder at 2003 cm−1 is retained. This result can be attributed to the presence of carbon covering the surface of Ni nanoparticles, which hinders the formation of bridged and multibonded CO species with higher adsorption strength [56, 58]. For NiCo1.02/SiO2, only a peak of linearly adsorbed CO on Ni is observed located at 2052 cm−1, considering no peaks of adsorbed CO on Co at room temperature [59, 60]. However, this peak is blue-shifted compared to Ni/SiO2 because the electron-poor Ni decreases the electron back-donation from the d orbital of Ni to the 2π* antibonding orbital of CO. The result of DRIFTS further confirms the XPS analysis that the electron transfer from Ni to Co occurs on NiCo1.02/SiO2. It should be noted that the absence of peaks below 2000 cm−1 on NiCo1.02/SiO2 cannot be explained as a geometric effect caused by Co doping, due to the interference of carbon in the catalyst. Therefore, CO-pulse chemisorption is carried out to further investigate the change in the number of active sites induced by the second metal doping.

The CO uptake data of various catalysts are summarized in Table 1. Ni/SiO2 has the largest CO uptake (3.69 µmol g−1) among all catalysts. After introducing Zn or Co, the CO uptake of bimetallic catalysts decreases without exception, owing to the dilution of the active phase [61, 62]. In general, bimetallic NiZnx/SiO2 catalysts have higher CO uptake than NiCox/SiO2, which is associated with the much lower content of Zn that leads to a weak modification of Ni by Zn. For NiZnx/SiO2, the CO uptake gradually decreases with x from 0.004 to 0.05, but increases at x = 0.07. This is probably due to no Ni–Zn alloy formation on NiZn0.07/SiO2. With regard to bimetallic NiCox/SiO2, NiCo0.52/SiO2 and NiCo1.02/SiO2 show higher CO uptake than NiCo0.09/SiO2, despite the fact that the latter has the lowest Co/Ni ratio and contains no Ni–Co alloy. This is probably because metallic Co is able to chemisorb CO, as evidenced by a CO uptake of 0.43 µmol g−1 on Co/SiO2.

Activity and selectivity of catalysts

Fig. 3 compares the apparent activity and selectivity of different catalysts in the hydrogenation of phenylacetylene. Ni/SiO2 shows the highest apparent activity in terms of the reaction time that is needed for complete conversion of phenylacetylene (conversion > 99%), while Co/SiO2 has the lowest apparent activity. Bimetallic NiZnx/SiO2 (Fig. 3a) and NiCox/SiO2 (Fig. 3c) rank between Ni/SiO2 and Co/SiO2 with regard to the apparent activity. Besides, G/SiO2 (metal-free) and ZnO/SiO2-I have no obvious hydrogenation activity due to the absence of active metals (Fig. S8), reflecting that the SiO2 support and ZnO are almost inactive for phenylacetylene hydrogenation. The initial rates for each catalyst are calculated and listed in Table 1. It is found that the initial rates of either NiZnx/SiO2 or NiCox/SiO2 are correlated positively with the CO uptake data, which is reasonable as the CO uptake reflects the number of active sites. The CO uptake and reaction rates of NiCox/SiO2 are generally lower than those of NiZnx/SiO2, again confirming a weaker modification by Zn when compared to Co, which is owing to the low Zn content in NiZnx/SiO2 originating from the evaporation of Zn during pyrolysis.

Fig. 3
figure 3

a, c Conversion vs time and b, d selectivity vs conversion over NiZnx/SiO2 and NiCox/SiO2 bimetallic catalysts (reaction condition: 60 °C, 0.5 MPa, 5 wt% phenylacetylene in ethanol, 0.15 g of catalyst, and 100 g of solution in a semi-batch stirred tank reactor)

Full size image

The selectivity to styrene at complete conversion of phenylacetylene is of interest, which is an important indicator to evaluate the performance of the catalyst when considering the purpose of removing phenylacetylene from styrene. A high selectivity means that the loss of styrene caused by excessive hydrogenation in the removal of phenylacetylene can be avoided to a great extent. For NiZnx/SiO2, the selectivity to styrene at complete conversion of phenylacetylene follows the order NiZn0.05/SiO2 (selectivity: 86.3%) > NiZn0.02/SiO2 (79.9%) > NiZn0.004/SiO2 (76.0%) > Ni/SiO2 (64.4%) > NiZn0.07/SiO2 (54.1%) (Fig. 3b). Except for NiZn0.07/SiO2, the selectivity of bimetallic NiZnx/SiO2 is greatly increased as compared to Ni/SiO2. The improved selectivity of NiZn0.05/SiO2 and NiZn0.02/SiO2 can be attributed to both geometric and electronic effects induced by the formation of Ni–Zn alloy: on the one hand, the geometric effect reduces the number of active sites (as evidenced by the decreased CO uptake), whereby further hydrogenation of styrene to ethylbenzene can be inhibited to some extent [27, 28, 31]; on the other hand, the electron-rich Ni active sites can facilitate the desorption of electron-rich styrene [24, 27, 28], resulting in an increased selectivity. The relatively high selectivity of NiZn0.004/SiO2 is ascribed to the geometric effect, as no electronic effect is observed. An unexpected result is that the selectivity of NiZn0.07/SiO2 is lower than that of Ni/SiO2 although the number of active sites of the former is smaller than that of the latter. It seems that the presence of ZnO, with the opposite effect of Ni–Zn alloy, is responsible for the lower selectivity of NiZn0.07/SiO2.

For NiCox/SiO2, the selectivity to styrene is ranked in the order NiCo0.09/SiO2 (88.0%) > NiCo0.52/SiO2 (69.5%) > Ni/SiO2 (64.4%) > NiCo1.02/SiO2 (56.4%) > Co/SiO2 (55.4%) > NiCo1.97/SiO2 (43.1%) (Fig. 3d). Interestingly, NiCo0.09/SiO2 without Ni–Co alloy has the highest selectivity, while other NiCox/SiO2 (x = 0.52, 1.02, and 1.97) show a selectivity of below 70%. The low selectivity of NiCox/SiO2 (x = 0.52, 1.02, and 1.97) is a consequence of further hydrogenation of styrene induced by a strong adsorption of styrene, in view of the electron-poor nature of Ni active sites on these catalysts. By contrast, because there is only a geometric effect but no electronic effect between Ni and Co on NiCo0.09/SiO2, this catalyst shows a high selectivity to styrene. It is observed that the selectivity of NiCo0.52/SiO2 is slightly higher than that of Ni/SiO2, which is likely due to a trade-off between the electronic effect and the geometric effect.

It can be seen from the above results that among all the catalysts studied here, NiCo0.09/SiO2 exhibits the highest selectivity to styrene at complete conversion of phenylacetylene. When compared to non-precious metal catalysts in the literature, the selectivity of NiCo0.09/SiO2 remains at a high level. As presented in Table 2, under the premise of > 99% conversion of phenylacetylene, only four catalysts, i.e., NiZn3/AlSBA-15(40) [24], NiZn3/Al2O3 [25], Ni3Sn/MgAl2O4 [28], and 450-NiSix [61], show higher selectivity to styrene than NiCo0.09/SiO2. However, Ni3Sn/MgAl2O4 [28] and 450-NiSix [61] are tested under a phenylacetylene/Ni molar ratio (49 and 6) much lower than that for NiCo0.09/SiO2 (361), meaning that much more Ni is required. For NiZn3/AlSBA-15(40) [24], the complicated preparation method and expensive reagents limit its practical application. It should be noted that the above comparison is rather crude because the reaction conditions (e.g., temperature, pressure, solvent and time) vary with each study.

Table 2 Performance comparison of non-precious metal catalysts in selective hydrogenation of phenylacetylene
Full size table

In addition, NiCo0.09/SiO2 is comparable to the commercial Lindlar catalyst (Pd–Pb/CaCO3 with a Pd loading of 5 wt%, manufactured by Aladdin Reagent Int. (Shanghai, China)) in terms of selectivity. Under the same condition, the Lindlar catalyst can achieve a conversion of 99.5% and a selectivity of 88.7% at 640 min of reaction [25]. Although the apparent activity of NiCo0.09/SiO2 (910 min of reaction) is somewhat smaller than that of the Lindlar catalyst, NiCo0.09/SiO2 is substantially more cost effective from the economic point of view, and thus an attractive alternative to the Lindlar catalyst.

Reusability of NiCo0.09/SiO2

Six consecutive cycles were performed, with the reaction time fixed at 910 min for each cycle. The magnetic NiCo0.09/SiO2 can be easily recycled from the solution by magnetic separation. As shown in Fig. 4, the selectivity to styrene is maintained at 88% during 6 cycles, while the conversion of phenylacetylene decreases slightly from 99.5 to 96.0%, which may be related to the mass loss of the catalyst during the recycling process. The metal particles of the used NiCo0.09/SiO2 remain uniformly dispersed on SiO2 with an average particle size of 4.3 nm (Fig. S9), similar to that of the fresh catalyst (4.1 nm). Additionally, ICP-OES analysis shows that the loadings of Ni and Co in the used NiCo0.09/SiO2 are 5.1 and 0.5 wt%, respectively, which are almost the same as those of the fresh catalyst (5.3 and 0.5 wt% for Ni and Co), indicating no metal leaching during the cyclic operation. These results demonstrate that NiCo0.09/SiO2 has good stability in the selective hydrogenation of phenylacetylene.

Fig. 4
figure 4

Reusability of NiCo0.09/SiO2 in 6 consecutive cycles of reaction (reaction condition: 60 °C, 0.5 MPa, 5 wt% phenylacetylene in ethanol, 0.15 g of catalyst, and 100 g of solution in a semi-batch stirred tank reactor)

Full size image

Effect of reaction conditions

Finally, the effects of reaction temperature and pressure on the activity and selectivity of phenylacetylene hydrogenation over NiCo0.09/SiO2 are examined (Fig. S10). The reaction temperature and pressure range from 50 to 80 °C and from 0.3 to 0.9 MPa, which are determined based on industrial data [67]. When the pressure is constant at 0.5 MPa, the reaction time for complete conversion of phenylacetylene is shortened from 1140 to 450 min with increasing the reaction temperature from 50 to 80 °C, while the selectivity to styrene decreases from 89.1 to 82.7%. On the other hand, when the temperature is constant at 60 °C, the reaction time for complete conversion of phenylacetylene reduces from 1040 to 600 min as the pressure varies from 0.3 to 0.9 MPa, corresponding to a drop in the selectivity from 88.3 to 83.0%. It is clearly shown that an increase in temperature or pressure is beneficial for enhancing the apparent rate of phenylacetylene hydrogenation, but unfavorable for the selectivity to styrene. Therefore, a compromise is necessary in determining an acceptable reaction rate that would provide a good selectivity to styrene. However, this issue belongs to the optimization problem, which is beyond the scope of this paper.

Conclusions

A series of NiZnx/SiO2 and NiCox/SiO2 bimetallic catalysts were prepared by pyrolysis of glucose and applied to selective hydrogenation of phenylacetylene. CO chemisorption, XRD, HRTEM, and XPS analyses revealed that the Ni–Zn alloy was formed in NiZn0.02/SiO2 and NiZn0.05/SiO2, while the Ni–Co alloy was formed in NiCo0.52/SiO2, NiCo1.02/SiO2, and NiCo1.97/SiO2. In addition, XPS and DRIFTS measurements showed the electron transfer from Zn to Ni for Ni–Zn bimetallic catalysts and from Ni to Co for Ni–Co catalysts. The performance of bimetallic catalysts in the selective hydrogenation of phenylacetylene strongly depended on the Zn/Ni and Co/Ni molar ratios. Compared to monometallic Ni/SiO2, NiZn0.05/SiO2 and NiCo0.09/SiO2 exhibited reduced apparent activity but significantly enhanced styrene selectivity of 86.3% and 88.0%, at above 99% conversion of phenylacetylene. For NiZn0.05/SiO2, the enhancement in the selectivity was due to the geometric and electronic effects induced by Ni–Zn alloy, while for NiCo0.09/SiO2, it was mainly attributed to the geometric effect caused by Co doping. NiCo0.09/SiO2 was proven to be recyclable for 6 cycles almost without structural and performance changes. Therefore, non-precious NiCo0.09/SiO2 is a promising catalyst for the selective hydrogenation of phenylacetylene.

Supplementary information

The online version contains supplementary material available.