Journal of Industrial and Engineering Chemistry

Transcription

1 Journal of Industrial and Engineering Chemistry 34 (2016) Contents lists available at ScienceDirect Journal of Industrial and Engineering Chemistry jou r n al h o mep ag e: w ww.elsevier.co m /loc ate/jiec Gas phase selective conversion of glycerol to acrolein over supported silicotungstic acid catalyst Amin Talebian-Kiakalaieh, Nor Aishah Saidina Amin *, Zaki Yamani Zakaria Chemical Reaction Engineering Group (CREG), Faculty of Chemical Engineering, Universiti Teknologi Malaysia (UTM), Skudai, Johor, Malaysia A R T I C L E I N F O Article history: Received 1 October 2015 Received in revised form 14 November 2015 Accepted 30 November 2015 Available online 10 December 2015 Keywords: Glycerol Acrolein Dehydration Supported silicotungstic acid Keggin anion A B S T R A C T Gas phase dehydration of glycerol to acrolein over a series of supported HSiW on ZrO 2 and nano-sized g- Al 2 O 3 catalyst has been investigated. The characterization results revealed that impregnation of g-al 2 O 3 nanoparticles increased the specific surface area, pore diameter, and thermal stability of the supported catalysts. The highest acrolein selectivity of 88.5% at 97.0% glycerol conversion was achieved over 0.5 g 30HZ-20A catalyst in 3 h at glycerol feed concentration of 10 wt%, temperature = 300?C and TOF = 136 h 1. The coke deposition has no significant effect on the activity of 30HZ-20A catalyst. Indeed, the catalyst was stable even after 40 h. ß 2015 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights reserved. Introduction The catalytic conversion of renewable feedstock for synthesis of value-added chemicals has attracted much attention in the last two decades to compensate various environmental concerns and excessive demand for fossil fuel consumption [1]. The production of biodiesel worldwide has caused surplus of glycerol, the main by product. In accordance to the global surge, over 1.54 million tones of glycerol are anticipated in 2015 and beyond. Hence, enormous availability of glycerol has drastically decreased the price of glycerol from 1.2 USD/kg in 2001 to 0.3 USD/kg in 2015 [2]. Industrial application of glycerol as a bio-renewable feedstock has been increasing not only because of the rapid surge in glycerol capacity, but also because glycerol is bio-sustainable, non-toxic, and bio-degradable. Furthermore, the multi-functional structure and physico-chemical characteristics of glycerol make it as one of the top 12 most important bio-based chemicals in the world [3]. Value-added chemicals such as acrolein, ethylene glycol, propanediol, and lactic acid are produced by catalytic dehydration, oxidation, reforming, hydrogenolysis, esterification, and oligomerization of glycerol [4,5]. * Corresponding author. Tel.: ; fax: addresses: amin.talebian63@gmail.com, noraishah@cheme.utm.my (N.A.S. Amin). Acrolein is also known as 2-propenal or acrylic aldehyde, the simplest unsaturated aldehyde. Moreover, acrolein is one of the most significant intermediates in the chemical industry for production of acrylic acid and methionine. For instance, superabsorbent polymers from acrylic acid are being used in the hygienic products due to the extreme absorbance of liquids (more than 500 times their weight). The conventional method for production of acrolein is selective oxidation of propylene in presence of complex BiMoO x based catalyst with approximately 85% acrolein selectivity at 95% propylene conversion. However, petrochemical exhaustion is foreseen in the near future. Therefore, the notion to produce significant industrial materials like acrolein from sustainable and renewable resources is more prevalent recently. The renewable route for acrolein production is through glycerol dehydration over various acid catalysts such as mineral acids, zeolites, heteropoly acids, and metal oxides in gas [6 9] or liquid [10] phases. Among several types of catalysts for glycerol conversion, heteropoly acids (HPAs) have attracted much attention because of its strong and easily tunable acidity as well as uniform acidic sites. Acidic active sites are the most important factor for higher catalytic activity and acrolein yield. The main advantage of HPAs is the unique Keggin structures favorable for acrolein production due to high Bronsted acid strength [11]. The disadvantages are its low thermal stability and surface area. In order to overcome these problems, HPAs are often supported on acidic or natural carriers such as zirconia (ZrO 2 ), alumina (Al 2 O 3 ), and silica (SiO 2 ) X/ß 2015 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights reserved.

2 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) Experimental Materials Fig. 1. Electronic interaction between HSiW and ZrO 2 support. Mixing ZrO 2 with other oxides can easily adjust its acid-base characteristics. In addition, the supported HPAs species on ZrO 2 could lead to full disappearance of the surface anions, and emergence of very strong Bronsted acidic sites [12]. ZrO 2 support increases the long-term [13] and thermal stability of the HPA supported catalysts, even better than silica supported catalysts [11]. The main reason for higher thermal stability and activity of supported HSiW on ZrO 2 and Al 2 O 3 catalysts is the strong interaction between HSiW and the supports. In fact, the hydroxyl groups on the surface of support are protonated in contact with water during the impregnation process to produce Zr OH 2 + species. The support and the negatively charged heteropoly anions have strong electronic interaction [14,15] as described in Fig. 1. Cs-HPW catalyst displayed the highest ever reported acrolein selectivity of 98% at 100% glycerol conversion, but the catalyst was stable only for a few hours and deactivated quickly [16]. The main reason for fast deactivation is coke deposition on the catalyst surface. Therefore, the large scale production of acrolein by glycerol dehydration depends on circumventing the fast deactivation problem. Small pore diameters, low surface acidity, and even very strong acidic sites on the catalyst surface are known to aggravate coke formation [17]. Co-feeding of hydrogen or oxygen is one of the common methods intended for eliminating or slowing down coke deposition on catalyst surface. However, this method may increase the possibility of explosive conditions or oxidation of products, finally decreasing the acrolein selectivity. Otherwise, large pore structure and high surface acidity are the two key catalyst properties for higher acrolein production. The effects of these two characteristics have been studied elsewhere [18]. As reported by Stosic et al. [12] the catalyst activity is dependent on the total acidity of catalyst as well as the nature of acid sites on the catalyst surface. Thus, the current study proposes the catalyst long-term stability can be enhanced without gas co-feeding or any noble metals application and it would be possible only by improving the textural characteristics of the catalyst such as acidity, thermal stability, and particularly pore size during the catalyst preparation step. Among the HPAs (HSiW, HPW, and HPMo), HSiW was chosen due to several reasons including strong acidic sites (Bronsted) and higher water tolerance abilities. According to several literatures, ZrO 2 is one of the best supports for HPAs because ZrO 2 increases the long-life and thermal stability of catalyst. Since both ZrO 2 and HSiW have low surface areas, g-al 2 O 3 nanoparticle was added as the third component to the catalyst in order to increase the surface area and pore diameter. The main objective of this study is to prepare a series of thermally stable, high surface acidity, large surface area, and large pore diameter catalysts for gas phase glycerol conversion to acrolein. The catalysts were characterized to elucidate the textural characteristics. After determining the best catalyst with the highest selectivity to acrolein, the effects of various reaction temperatures ( ?C), prolonged time, and Keggin anion density on the catalyst activity were also evaluated. Glycerol (purity > 99%), silicotungstic acid (H 4 SiW 12 O 40 14H 2 O (HSiW)), aluminum oxide (Al 2 O 3 ) nanoparticle, zirconium oxide (ZrO 2 ), acetic acid, allyl alcohol, hydroxyacetone, acetone, propanal, ethanal at reagent grade were supplied by Sigma Aldrich (Malaysia). Acrolein at reagent grade was purchased from Scientific Trends (M) Sdn. Bhd. Catalyst preparation In the initial step, a series of catalysts with various HSiW loadings (10, 20, 30, and 40 wt%) on ZrO 2 were prepared by the incipient-wetness impregnation method. In this method, aqueous HSiW solutions were added drop-wise to the ZrO 2 support. The slurries were rigorously stirred for 12 h followed by drying at 110?C for 18 h. The HSiW-ZrO 2 supported catalysts were denoted as 10HZ, 20HZ, 30HZ, and 40HZ for 10, 20, 30 and 40 wt% of HSiW loading, respectively. Next, the second series of catalysts were prepared by impregnation of Al 2 O 3 nanoparticles with different weight percentages (10 30 wt%) on the 30HZ sample. Initially, 30HZ suspension was prepared while the nanoparticle alumina was mixed with water to form slurry. Finally, the 30HZ suspension was added drop-wise to the alumina slurry under continuous stirring for at least 20 h followed by drying at 120?C for 18 h. The synthesized catalysts were denoted to as 30HZ-10A, 30HZ-20A, and 30HZ-30A. The 70Z-30A sample was similarly prepared, but without HSiW to evaluate the effect of ZrO 2 and Al 2 O 3 combination on the catalyst activity. Catalyst characterization Nitrogen adsorption/desorption at 196?C was employed to determine the BET surface area, pore volume, and average pore diameter (Thermo Scientific, SURFAR). Prior to analysis, the sample was degassed under vacuum at 200?C for 4 h to ensure complete removal of adsorbed moisture. The BJH method was used for average pore diameter calculation. X-ray diffraction (XRD) was used to determine the structure of the prepared catalysts and to evaluate the degree of crystallinity. The XRD patterns were obtained by D5000 Siemens instrument using CuKa radiation (40 kev, 40 ma). The patterns were scanned in the 2u range of 10 80? (step width 0.05?, 1 s per step). The crystalline size of Al 2 O 3 in supported 30HZ catalyst with wt% Al 2 O 3 samples is calculated using Scherrer s equation [19,20], D P ¼ ð 0:94l=bcosuÞ, where D p, l, b, and u are average crystalline size, X-ray wavelength, line broadening (Peak halfwidth) in radius, and diffraction angle, respectively. The total acidity of the prepared catalysts was determined by temperature programmed desorption of ammonia (NH 3 TPD) using the Micrometrics Auto Chem II instrument. A specific amount of catalyst was dried at 200?C for 1 h by saturation of NH 3 at 60?C for 0.5 h before the physically adsorbed NH 3 was removed by helium purging at 60?C for 0.5 h. The temperature was ramped at a heating rate of 20?C/min to 700?C to desorb the chemically absorbed NH 3. Thermogravimetric analysis (TG) of prepared catalyst was performed using a THERMO TGA instrument in a temperature range of 30 to 800?C with a ramp rate of 10 K/min under nitrogen flow. The Fourier Transform Infrared Spectroscopy (FTIR) spectra of the freshly prepared catalysts were performed by Perkin Spectrum One FTIR spectrometer with a spectral resolution of 2 cm 1 and scan time of 1 s using KBr disc method.

3 302 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) The FTIR spectra after pyridine adsorption (Py-IR) on the prepared catalysts were performed using 170-SX FTIR spectrophotometer to evaluate the nature of the acid sites (Bronsted and Lewis). Prior to pyridine adsorption test, the powder samples (50 mg) were pressed to form a thin pellet (5 t/cm 2 ) with approximately 20 mm diameter. The prepared pellets were then pretreated in high vacuum system at 300?C and Pa overnight. After the samples were cooled to room temperature under vacuum, 2 ml of liquid pyridine was injected to the system. Finally, the spectra were recorded at 150?C and 250?C after 60 min evacuation. The morphology of the samples was observed by Field Emission Scanning Electron Microscopy (FESEM, HITACHI, SU-8000) with the STEM detector. All samples were coated with gold under vacuum condition in an argon atmosphere ionization chamber. Energy dispersive X-ray (EDX) analysis was carried out with the same equipment. In addition, the particle size, dispersion, and distribution were evaluated by FEI-Tecni G2 Transmission Electron Microscope (TEM) operating at 300 kv. The elemental characterization of the used catalysts substances in terms of coke content was determined by Thermo Scientific Flash 2000 instrument. First, each sample was weighed in a tin capsule and then placed in the combustion reactor. The reaction was performed in a furnace at ?C temperature. Small volume of oxygen (O 2 ) was added to the system to burn the inorganic or organic materials. As a result, the sample was converted into simple elemental gases. Finally, TCD detector with separation column was used to determine the coke content. Catalytic reaction The gas phase dehydration of glycerol was conducted at atmospheric pressure and temperature range of ?C in a vertical fixed- bed quartz reactor (30 cm length, 11 mm i.d.) using 0.5 g catalyst sandwiched between plugs of glass wool. Prior to reaction, the catalyst was pretreated at reaction temperature (300?C) under nitrogen (N 2 ) flow (1200 ml/h) for 1 h. Liquid aqueous glycerol (10 wt%) was fed by a syringe pump with 2 ml/h flow-rate. The liquid was vaporized in a pre-heater, mixed with inert and run through the reactor. Gas hourly speed velocity (GHSV) of the inert carrier gas was 1200 h 1. After 3 h, the products and unconverted glycerol were condensed in a water-ice salt bath ( 5?C) and collected hourly for analysis. n-butanol was added to the condensed products as internal standard. The final solution was analyzed by a gas chromatograph (GC) equipped with capillary column (DB Wax; 30 m 0.53 mm 0.25 mm) and FID detector. To achieve effective product separation, the column was held at 40?C for 4 min before the temperature was ramped up to 200?C with a rate of 12?C/min and remained for 23 min. The glycerol conversion, product selectivities and yields are defined by Eqs. (1) (3) [15], and carbon balance is calculated by equation (4) [21]. All the experimental runs were repeated at least 3 times and the average values with error bars have been reported as the final results. M Gl; in feed M Gl; in outlet X Gl ð%þ ¼ 100% (1) M Gl; in feed M C in product Selectivity N ð%þ ¼ M C in Gl; feed M 100% C in Gl; outlet (2) Yield N ð%þ ¼ X Gl ð%þselectivity N ð%þ (3) Carbon Balance ¼ Sum of the selectives of identified products (4) where M Gl, M C, and X Gl are moles of glycerol, moles of carbon for each product (N), and glycerol conversion, respectively. Results and discussion Surface area and pore structure analysis The textural characteristics, surface area (S BET ), pore volume (V p ), average pore diameter (D p ), and pore area (A p ) of supported HSiW catalysts derived from nitrogen physisorption isotherms are presented in Table 1. The surface areas of pure HSiW, ZrO 2, Al 2 O 3 and 70Z-30A are 8, 14, 143, and 56 m 2 /g, respectively. By increasing the amount of HSiW from 10 to 40 wt% on the support (ZrO 2 ), the specific surface areas of the samples increased significantly from 18 to 22 m 2 /g. Chai et al. [22] reported similar results with the addition of 5 to 15 wt% phosphotungstic acid (HPW) on ZrO 2. Theoretically, there should be some plugging after HSiW impregnation to the support. The surface areas of porous ZrO 2 enhanced after combination with HSiW [13]. Large size metal cations increased the interstitial volume among Keggin anions to form new pores. Consequently, the effective surface impeded by plugging was renewed [23]. Conversely, impregnation of HSiW on ZrO 2 reduced the pore volume and diameter. Indeed, the pore volume dropped from 0.12 to 0.06 cm 3 /g by increasing the HSiW loading from 10 to 40 wt%. Similarly, the pore diameter decreased from 39.5 nm to 18.7 nm for 10HZ and 40HZ samples, respectively. The final series of 30HZ catalysts, impregnated with 10 to 30 wt% nano-sized g-al 2 O 3, exhibited remarkable effects on the surface area, pore volume and pore diameter. Indeed, the specific surface area increased from 20 m 2 /g up to 91 m 2 /g; approximately 5 times larger than the surface area of 30HZ. Similarly, the pore volume of 30HZ-30A was about 5 times more than the 30HZ. Initially, the pore diameter surged to 41.4 nm with 10 wt% Al 2 O 3 loading, but decreased to 28.6 nm and 27.5 nm for the 30HZ-20A and 30HZ-30A catalysts, respectively. The reduction was due to plugging of the pores by agglomeration [23] or the HSiW Keggin unit size (D = 1.2 nm) was small enough to plug the support pores and decreased the pore diameter [17]. According to the size of pore diameters, the mesoporous (2 50 nm) network remained fully accessible to the reactant [15]. Catalyst acidity NH 3 TPD was carried out to determine the amount and strength of acidity for different prepared catalysts (Table 2). The strength of acid sites was divided into three groups: weak, medium, and strong based on the temperature of the TPD position [24]. The total acidity of bare ZrO 2, Al 2 O 3, HSiW, and 70Z-30A Table 1 Pure and supported catalysts surface area (S BET ), pore volume (V p ), pore diameter (D p ), and pore area (A p ). Catalyst S BET (m 2 /g) V p (cm 3 /g) D p (nm) A p (m 2 /g) HSiW ZrO Al 2 O Z-30A HZ HZ HZ HZ HZ-10A HZ-20A 45 (36) a 0.20 (0.16) a 28.6 (27.8) a 55.0 (41.8) a 30HZ-30A a Used catalyst characteristics.

4 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) Table 2 NH 3 TPD results (acidity) for pure and supported silicotungstic acid catalysts. Catalyst Weak ( C) Medium ( C) Strong (450 C) Total acidity v(mmol/g cat) NH 3 peak position ( C) Acid amounts (NH 3 /Cat, mmol/g) NH 3 peak position ( C) Acid amounts (NH 3 /Cat, mmol/g) NH 3 peak position ( C) Acid amounts (NH 3 /Cat, mmol/g) HSiW 131 & ZrO & Al 2 O & Z-30A HZ 125 & HZ 122 & HZ 133 & HZ 129 & HZ-10A HZ-20A HZ-30A catalysts were 0.2, 1.2, 1.6, and 0.9 mmol/g.cat, respectively. The huge difference between the total acidity of HSiW and ZrO 2 is attributed to Bronsted acidic sites and Lewis acidic sites, respectively. The acidity of the supported HSiW catalysts increased from 0.4 to 1.2 mmol/g.cat with increasing HSiW loading (10 to 40%) on ZrO 2. All samples displayed three peaks: the first and second peaks displayed weak acidic sites ( ?C) and the third was in the strong acidic sites (>450?C). Finally, the impregnation of g-al 2 O 3 nanoparticles on the supported 30HZ catalyst increased the acidity from 2.4 mmol/g.cat for 30HZ-10A to 2.6 mmol/g.cat for 30HZ-30A sample. This group of catalysts has two (weak and medium) acidic strengths. However, the medium acid densities were significantly higher than the weak ones. In fact, bare Al 2 O 3 with only weak acid sites (1.2 mmol/g.cat) enhanced the weak and medium acidic sites on the final group of catalysts. All the catalysts exhibited slight shifts in their TPD peak position to higher temperatures (133 to 135?C for weak acid strength and 300 to 316?C for medium acid strength). The main reason for the shift of the TPD peak position is attributed to the protonated hydroxyl groups on the supported surface during the impregnation process. As a consequence, an electronic interaction occurred between the HSiW and support, strengthening the bonding between the two [15]. Indeed, the addition of g-al 2 O 3 nanoparticles enhanced the interaction between support and active compounds. Nature of acidity (Bronsted and Lewis) Fig. 2a c illustrate the pyridine adsorption spectra for the impregnated catalysts (30HZ-10A, 30HZ-20A, 30HZ-30A) at two different temperatures (150 and 250?C). The spectra exhibited typical bands at around cm 1 and cm 1 for Lewis and Bronsted acid sites, respectively [25]. The prepared samples were observed to have significant amount of Bronsted acid sites due to the large peaks at 1540 cm 1. The peaks at 1490 cm 1 are related to the combination of Bronsted and Lewis acid sites. All the samples represented very small amount of Lewis acid sites as evident by the peaks at cm 1. In fact, these results supported the notion HPAs are typical Bronsted acid catalysts [1,15,16]. The 30HZ-20A sample was also evaluated at reaction temperature (300?C) and surprisingly the Bronsted and Lewis acid sites did not show any reduction compared with the other two samples. It confirms the 30HZ-20A catalyst would be stable at the reaction temperature without any decomposition or leaching of active compounds. Besides, Fig. 2b clearly exhibits adding 20 wt% Al 2 O 3 increased the peak at 1490 cm 1. This is attributed, to the Bronsted and Lewis combination due to the increasing number of Lewis acid sites by deposition of Al 2 O 3 on the catalyst surface. However, for Al 2 O 3 loading above 20 wt% the peak at 1540 cm 1 was slightly reduced due to large amount of Al 2 O 3 deposition on the surface blocking some Lewis acid sites [11]. The huge difference between the surface areas is the other reason for the reduced Bronsted and Lewis acid site peaks at 1540 and 1490 cm 1 in 30HZ-30A sample compared with 30HZ-20A [24]. Finally, Fig. 2d compares the standard pyridine adsorption spectra of the three prepared catalysts. The results demonstrate all the samples have typical Bronsted acid sites with low amount of Lewis acid sites. X-ray diffraction (XRD) Fig. 3 illustrates the XRD patterns of pure HSiW, ZrO 2, Al 2 O 3, 30HZ and supported 30HZ by (10 30 wt%) nano g-al 2 O 3 catalysts. The XRD spectrum of pure HSiW illustrates characteristics peaks at 10.9?, 25.5?, and 34.7? related to the Keggin anions. Bulk ZrO 2 exhibits characteristics peaks attributed to the monoclinic phase at 24.05?, 28.2?, 31.5?, 34.25?, and 50.15?. In addition, diffraction peaks related to the pure Al 2 O 3 appeared at 18.8?, 36.9?, 44.2?, and 67.3?. The supported 30HZ catalyst impregnated with10 to 30 wt% nano g-al 2 O 3 particles indicates the presence of diffraction peaks attributed to zirconia at 28.2? and 31.5?. Furthermore, the main alumina peak at 67.3? appeared in the new prepared catalysts. Meanwhile, increasing the Al 2 O 3 loading from 10 to 30% increased the diffraction peaks of Al 2 O 3. Similarly, the XRD patterns, attributed to the 30HZ and supported 30HZ with Al 2 O 3, did not correspond to the peaks of HSiW Keggin anions inferring high dispersion of 30%HSiW on ZrO 2 support [26]. The main reason that prevented HSiW crystallization on ZrO 2 support is the surface coverage of ZrO 2 by HSiW, attributed to the small HSiW Keggin anion size (D = 1.2 nm) and low surface area of ZrO 2 (14.0 m 2 /g) [11]. The most important advantage of surface HSiW coverage is making a strong interaction between the HSiW and ZrO 2 to enhance the catalyst thermal stability and to avoid leaching of HSiW during reaction [11,17]. The surface coverage of HSiW on the prepared catalyst was calculated by an equation mentioned elsewhere [27]. The theoretical 30 wt% HSiW surface coverage on ZrO 2 was 9.2 mmol/g; however, the HSiW coverage decreased to 5.6, 4.1, and 2.5 mmol/g by loading 10 to 30 wt% nano Al 2 O 3 on the catalyst, respectively. As reported, less surface coverage showed better dispersion of the active compounds (HSiW) on the support surface [17]. The inverse relation between molecular weight and surface coverage indicated HSiW has a higher dispersion among HPAs due to its higher molecular weight compared with HPW and HPMo catalysts [17].

5 304 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) Fig. 2. FTIR spectra of pyridine adsorbed on prepared catalysts at 150 and 250?C, (a) 30HZ-10A, (b) 30HZ-20A, (c) 30HZ-30A, and (d) overall comparison of pyridine adsorption for the three prepared samples at 150?C. The crystallite size of Al 2 O 3, calculated by Scherrer s equation, [19,20] using the peak at 2u = 67.7? for 30HZ-10A, 30HZ-20A, and 30HZ-30A samples were 6.3, 7.7, and 10 nm, respectively. The larger Al 2 O 3 crystalline size with increasing amount of Al 2 O 3 loading could be explained by the following: large number of Al 2 O 3 caused higher number of Al 2 O 3 molecules capable of aggregating to form large crystals. Katryniok et al. [13] also reported higher loading (40 wt%) of zirconia caused larger crystal size. FTIR spectroscopy Fig. 4 displays the FTIR spectra for the HSiW, 30HZ, fresh 30HZ- 20A, and used 30HZ-20A samples. The pure HSiW spectra rendered four major absorption peaks at 874, 931, 981, and 1004 cm 1. These four peaks are assigned to Si O (1004 cm 1 ), W5O (981 cm 1 ), W O W (931 cm 1 ) inter-octahedral, and W O W (874 cm 1 ) intra-octahedral silicotungstic anions, respectively [28]. All of the supported catalysts exhibit absorption peaks of 931, 981, and 1004 cm 1 to consolidate the presence of the Keggin structure. Katryniok et al. [15] have only reported 2 absorption bands of STA Keggin units at 981 and 928 cm 1 to confirm the structural stability of active compounds (HSiW) for fresh and used catalysts. Thermal stability Fig. 3. XRD patterns of the pure and supported samples. TG-DTA analysis was performed to evaluate the thermal stability of the prepared 30HZ and 30HZ-20A catalysts under nitrogen flow (Fig. 5). Pure HSiW exhibited three weight- loss

6 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) steps. The first one with 8% and the second one with 3% weight loss were observed at 85 and 200?C, respectively. These two weight losses referred to the loss of the physically absorbed water and removal of water from HSiW hydrated heteropolyacid, respectively [29,30]. The third weight loss (1%) at 480?C indicated the decomposition of the HSiW Keggin anions [13,31]. The DTA curves clearly exhibited two weight losses of pure Al 2 O 3 at 75 and 250?C, related to the loss of physically adsorbed water and dehydroxylation of hydroxyl groups, respectively. Moreover, pure ZrO 2 registered minor weight loss at 70?C (TG curve) due to the loss of physically adsorbed water. The TG-DTA curve for 30HZ catalyst clearly indicated two endothermic peaks at 75 and 190?C is attributed to the removal of crystalline water. Finally, the TG-DTA curves of 30HZ-20A clearly showed only one small weight loss at 75?C due to the removal of the physically absorbed water. However, no important structural changes were observed above 75?C, confirming high thermal stability of the 30HZ-20A sample. The main reason for such high catalyst stability was the electronic interaction between HSiW and supports [13]. Indeed, a strong interaction between HSiW and support is essential for long-term catalyst activity [32] to prevent the catalyst from losing constitutional water [15,26]. Morphological analysis Fig. 4. FTIR spectra of bulk HSiW, 30HZ, fresh 30HZ-20A, and spent 30HZ-20A supported catalysts. The morphological analysis of the fresh and used 30HZ-20A samples observed by FESEM and TEM are displayed in Fig. 6. The HSiW with plate like morphology seems to be dispersed on the external surface of the ZrO 2 [33]. In fact, the HSiW particles are regular in shape and dispersed uniformly increasing the specific surface area and porosity in accordance to the BET results (the Surface area and pore structure analysis section). Moreover, the addition of g-al 2 O 3 nanoparticles doubled the surface area of the catalyst from 20 to 45 m 2 /g for 30HZ and 30HZ-20A, respectively. Fig. 6a and b illustrate the fresh 30HZ-20A sample. Furthermore, Fig. 6c and d exhibit no dramatic change on the used catalyst surface morphology. The FESEM study reveals that the catalyst morphological characteristics remain unaffected during the reaction due to high hydrothermal stability of Al 2 O 3 and ZrO 2 (confirmed by TG-DTA results). Similar results have been reported elsewhere [25]. Indeed, high stability of 30HZ-20A catalyst depends on the strong electronic interaction between active phase compounds (HSiW) and supports (ZrO 2 and Al 2 O 3 ) as well as high thermal stability of the HSiW Keggin anion structure. The EDX analysis results related to the fresh 30HZ-20A catalyst are Fig. 5. TG-DTA plots for bulk (ZrO 2, Al 2 O 3, HSiW) and supported HSiW samples.

7 306 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) Fig. 6. FESEM images for (a and b) fresh 30HZ-20A, (c and d) used 30HZ-20A, (e) EDX mapping for fresh 30HZ-20A catalyst, (f) TEM micrographs related to active components particle size and distribution in the synthesized 30HZ-20A sample and (g) nano Al 2 O 3 size estimation by TEM. illustrated in Fig. 6e. The EDX mapping demonstrates an extremely uniform distribution of tungsten (W), zirconia (Zr), and alumina (Al) over the catalyst. In addition, the EDX analysis (Fig. 6e) indicates 46.5, 36.4, and 17.1 wt% of Zr, W, and Al in the catalyst sample conforming to the theoretical values of 50, 30, and 20 wt%, respectively. The TEM micrographs demonstrate the active components (HSiW) are well-dispersed with regular spherical shape (Fig. 6f and g). In fact, high dispersion of HSiW on the support as revealed by TEM corroborated with the XRD results (the X-ray diffraction (XRD) section). The dark micelles represented the HSiW particles. Indeed, the HPAs are heavy metals and strong electronic scattering are demonstrated in the TEM characterization as similarly observed before [34]. The particle sizes of the HSiW are in the range of nm (Fig. 6f). The large surface area (45 m 2 /g), pore volume (0.2 cm 3 /g), and uniform mesoporous structure of the support could trap the HSiW particles inside the pores. Similar results have been reported elsewhere [35,36]. Furthermore, the nano g-al 2 O 3 crystal size was estimated around 7 10 nm by TEM image (Fig. 7g) based on the lattice fringes with the interplanar spacing of 0.48 nm. The Al 2 O 3 crystal size (7 10 nm) in this section is in good agreement with the Scherrer s equation results (the X- ray diffraction (XRD) section) for 30HZ-20A sample with 7 nm crystal size. Catalyst performance Application of all bulk samples (HSiW, ZrO 2, and Al 2 O 3 ) and even combination of two catalyst supports (70Z-30A) in gas phase dehydration of glycerol could not obtain more than 38.1% selectivity of acrolein. These results demonstrate the strong Lewis acid sites samples (ZrO 2, Al 2 O 3, and 70Z-30A) and even the strong

8 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) Fig. 7. (a) Glycerol conversion versus time and (b) Acrolein selectivity versus time for 30HZ-10A, 30HZ-20A, 30HZ-30A samples at 300?C, 12 h reaction time, 2 ml/h glycerol feed, and 20 ml/min carrier gas flow (c) Acrolein selectivity versus glycerol conversion only for the most stable and active sample (30HZ-20A), and (d) overall selectivity versus conversion related to the 30HZ-20A sample (e) Long-term stability investigation of 30HZ-20A sample in 40 h. Bronsted acid site catalyst (HSiW) were not suitable for high acrolein selectivity due to various reasons. The byproducts prevailed over strong Lewis acid sites while strong Bronsted acids enhanced coke formation on catalyst surface. Therefore, impregnation of the active compounds (HSiW) on supports (ZrO 2 and Al 2 O 3 ) is required to increase the glycerol conversion and acrolein selectivity. Initially, the gas phase dehydration of glycerol to acrolein was investigated by the ZrO 2 supported HSiW (10 40 wt%) catalysts at different temperatures: 280, 300, and 320?C in 3 h reaction time (Table 3). The results indicated increasing the HSiW loading from 10 to 30 wt% led to higher acrolein selectivity. In fact, impregnation of HSiW on ZrO 2 significantly enhanced the catalytic activity compared to the pure HSiW or ZrO 2. Well- dispersed HSiW on the support (ZrO 2 ) surface increased accessibility of glycerol molecules to the catalytically active protons on the supported catalyst. As a result, the glycerol conversion is increased [17]. It was also observed HSiW loading above 30 wt% led to the deposition of a huge amount of active phase on ZrO 2 and again increased the acid strength. Consequently, the coke formation on the catalyst surface increased while catalyst deactivated faster [1,17,19]. The 30HZ catalyst displayed the best performance with the highest acrolein selectivity of 69.3% at 92.0% glycerol conversion at 300?C. Increasing the reaction temperature from 280 to 320?C steadily increased the glycerol conversion from 84.1% to 93.6% over the 30HZ catalysts. Table 4 lists the glycerol conversion, acrolein yield, and selectivities of all the products. The results demonstrated 30HZ- 20A catalyst registered the highest acrolein selectivity of 88.5% with 97.0% glycerol conversion at 300?C. Furthermore, increasing the temperature enhanced the conversion from 94.2% to 98.4%. The main by-products were acetone, ethanal, and acetic acid with 1.9%, 1.3%, and 1.3% selectivity, respectively. The carbon balance was determined based on identified products in gas and liquid phases for each experimental run. The carbon balance did not reach to Table 3 Catalyst performance results over blanks and wt% HSiW supported on zirconia catalysts a Catalyst T (?C) X Gl (%) Y (%) b Selectivity (%) Acrolein Ethanal Propanal Acetone Hydroxyacetone Allyl alcohol Acetic acid Carbon balance (%) Blank Al 2 O ZrO Z-30A HSiW HZ HZ HZ HZ a Reaction condition: 300?C reaction temperature, 10 wt% glycerol concentration, 0.5 g catalyst weight, 2 ml/h glycerol feed flow, 20 ml/min carrier gas flow. b Acrolein yield.

9 308 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) Table 4 Catalyst performance results in glycerol dehydration to acrolein over wt% nano sized Al 2 O 3 supported on 30HZ catalyst a. Catalyst T (?C) X Gl (%) Y (%) b Selectivity (%) TOF e (h 1 ) Acrolein Ethanal Propanal Acetone Hydroxyacetone Allyl alcohol Acetic acid CB d 30HZ-10A [86.0] c HZ-20A [88.8] c HZ-30A [86.7] c a Reaction condition: 300?C reaction temperature, 10 wt% glycerol concentration, 0.5 g catalyst weight, 2 ml/h glycerol feed flow, 20 ml/min carrier gas flow. b Acrolein yield c Acrolein selectivity after 1 h. d Carbon Balance e Turn over frequency after 3 h of reaction time. 100% in Tables 3 and 4 due to some unidentified minor peaks (products) in GC and deposits of heavy compounds on the catalyst surface. Various reasons have been reported for production of heavy compounds such as: glycerol dimerization and oligomerization, formation of 2-ethylenyl due to reaction between glycerol and acrolein, and consecutive acid-catalyzed (self- or cross) aldol condensation reaction of primary oxygenates, such as acrolein and acetaldehyde [17,37,38]. The nature of acid sites present on the catalyst surface attributed to the differences in glycerol conversions and acrolein selectivities. In fact, the HSiW loading on ZrO 2 formed significant Bronsted acidic sites, which are superior to Lewis acid sites for dehydration of glycerol to acrolein [39]. The high activity of HSiW supported catalysts is attributed to the acidity and hydrolytic stability of these samples [32]. Correspondingly, HSiW was reported to be more active than the other HPAs due to the presence of four Keggin protons of HSiW compared to the three Keggin protons of HPW and HPMo catalysts in different reactions [40]. Moreover, according to Tesukuda et al. [19] the HSiW displayed higher activity in dehydration of glycerol to acrolein compared to other HPAs due to higher Bronsted acid strength and water resistance. Proper balance between the strengths of Bronsted (HSiW) and Lewis acidic sites (ZrO 2 ) attributed to higher catalyst activity and acrolein selectivity. Impregnation of HSiW on ZrO 2 decreased the Bronsted acidity of HSiW. The inherent Bronsted acid-site is good for long-term catalyst stability due to less coke deposition on the catalyst surface. In addition, the Lewis acid sites on the ZrO 2 support also decreased. As a result, by-products reduced significantly and acrolein yield prevailed [16]. Moreover, loading g-al 2 O 3 nanoparticles (10 to 30 wt%) increased the total acidity from 2.4 to 2.6 mmol/g.cat for the final series of catalysts, more than doubled the 30HZ acidity (1 mmol/g.cat) (Table 2). The most important difference between the two groups of catalysts is acid strength. Indeed, the 10-40HZ samples possessed weak and strong acid strengths, but the final series of samples (30HZ-(10-30A)) have weak and medium acidity strengths. In fact, the results of this study confirmed very high or low acidity is not suitable for acrolein selectivity. Therefore, the types of acidic sites and proper balance between Bronsted and Lewis strength acidity have significant impact on the acrolein selectivity. The strong interaction between HSiW and supports (ZrO 2 and Al 2 O 3 ) is required for prolonged catalyst activity [32]. The strong interaction led to higher stability during reaction at higher temperatures. The profile of TG-DTA confirmed the strong binding between HSiW and ZrO 2. In addition, all the prepared catalysts have large pore diameters (>27 nm), an important attribute for high catalyst stability. Indeed, pore size larger than 4 nm formed a more effective catalyst by providing the required space for Keggin anions (D = 1.2 nm) at the surface and enhanced the interaction between adsorbed glycerol molecules and Keggin anions [17]. The FESEM results (Fig. 6) of used 30HZ-20A catalyst also indicated no significant change on the spent catalyst surface morphology compared to the fresh one inferring the catalyst still remained highly active even after reaction due to the large pore diameter (>27 nm). Fig. 7a illustrates the glycerol conversion profiles versus time for all the prepared catalysts. The glycerol conversion decreased steadily during 12 h reaction time for all the samples. However, 30HZ-20A recorded the least reduction (97.1% to 82.3%) amongst all the tested samples. The acrolein selectivity for the three samples (30HZ-10A, 30HZ-20A, and 30HZ-30A) are exhibited in Fig. 7b. All samples exhibited rapid reduction in acrolein selectivity, but the most active and stable catalyst, 30HZ-20A registered a difference of only 10% from 88.5 to 78.0% during the first 6 h and remained constant beyond that period. The 30HZ-20A sample has higher total acidity and stronger acidic sites compared to 30HZ-10A due to a shift of TPD peaks from 300 to 305?C. However, the number of acid sites ð Total acidity ð NH 3 TPDÞ=Surface area ð BETÞÞ mmol=m 2 decreased from 0.1 to 0.03 mmol/m 2 by increasing the Al 2 O 3 loading from 10 to 30 wt% due to significant surge in the catalyst surface area. The surface area of 30HZ-30A sample doubled that of 30HZ- 20A catalyst. The drastic reduction in number of acid sites is in good agreement with reduction of catalyst activity. Indeed, 30HZ- 20A sample was more active than 30HZ-10A and 30HZ-30A catalysts due to stronger acidity and higher number of acidic sites, respectively. Fig. 7c exhibits glycerol conversion versus acrolein selectivity for 30HZ-20A sample during 12 h reaction time. As the glycerol conversion decreased from 97.1 to 82.3% the acrolein selectivity also dropped steadily from 88.5 to 78.0%. Fig. 7d presents the catalytic performance of 30HZ-20A laid in the fourth quadrant confirming the remarkable performance of 30HZ-20A catalyst with above 78% and 88% acrolein selectivity and glycerol conversion, respectively. Fig. 7e exhibits the long-term stability of 30HZ-20A catalyst during 40 h reaction time. The stability of 30HZ-20A sample with 75% acrolein selectivity at 80% glycerol conversion prevailed up to 40 h. Turn over frequency (TOF) The mechanism of heteropoly acid (HPA) catalysts is divided into two groups: bulk-type and surface-type according to Misono s

10 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) Fig. 8. (a) Effect of different glycerol feed concentrations on acrolein selectivity and glycerol conversion over 30HZ-20A catalyst and (b) effect of catalyst loading (30HZ-20A) on acrolein selectivity and glycerol conversion at 300?C, 2 ml/h glycerol feed, and 20 ml/min carrier gas flow. classification [41]. Glycerol dehydration to acrolein reaction over the water soluble and hydrophilic supported HSiW catalysts (30HZ-10A, 30HZ-20A, and 30HZ-30A) in this study can be classified as bulk-type mechanism [41]. Thus, the turn over frequency (TOF) is determined by the following equation [42,43]: TOF ¼ M AC M HSiW t where M AC mole of produced acrolein, M HSiW is mole of HSiW in catalyst, and t is time (h). The values of TOF are listed in the last column of Table 4 for 30HZ-10A, 30HZ-20A, and 30HZ-30A samples at different reaction temperatures. The 30HZ-20A catalyst is selected as the best catalyst in this study due to the highest acrolein selectivity corresponds to the highest TOF value of 136 h 1. The TOF values are approximately similar ( h 1 ) for 30 wt% HSiW in all samples. The TOF values of other similar catalysts are as follows: 20%HPW/SiO 2 with 88 h 1 and 30%HSiW/SiO 2 with 14.3 h 1 [16], 15%HSiW-Zr with 25.2 h 1 [44], 5%Ru/CsPW with 21 h 1 [45], and 15%HPW/ZrO 2 with 111 h 1 [46]. The main factor that changed the TOF values is due to the number of moles of acrolein produced. In addition, increasing the reaction temperature from 280 to 300?C enhanced the TOF values for all the catalysts. However, increasing the reaction temperature above 300?C has a negative effect. Effect of glycerol feed concentration (4) the coke deposition. Indeed, more than one glycerol monolayer on the catalyst surface caused the formation of monoaromatics that dramatically surged the coke deposition. Thus, low glycerol concentration in the feed significantly decreased the catalyst deactivation process. Effect of catalyst loading Fig. 8b depicts the effect of different amounts of catalyst loading on glycerol conversion and acrolein selectivity. The results revealed with increasing catalyst loading up to 0.5 g the acrolein selectivity increase from 73.2% to 88.5%, but beyond that the selectivity remained constant. Yadav et al. [49] reported similar results for the catalyst loading. Besides, they found higher amount of catalyst increased the lifespan of catalyst due to the increase in catalyst active sites. The glycerol conversion steadily rose with increasing catalyst loading. Effect of the Keggin anion density on catalyst performance High acidity of HSiW originated from its Keggin structure properties and decomposition or destruction of Keggin anions led to a significant drop in the acidity and finally decreased the catalyst activity [50]. The Keggin anion densities of the nano-alumina doped 30HZ catalysts were calculated by the following equation [22]: Fig. 8a illustrates the effect of different feed concentrations on glycerol conversion and acrolein selectivity. Various glycerol concentrations (0.5, 5, 10, 15, and 20 wt%) were utilized to examine the activity of the 30HZ-20A catalyst. Low glycerol feed concentration (0.5, 5 and 10 wt%) gave similar results in conversion, but the acrolein selectivity increased 10% from 77.8% for 0.5 wt% to 88.5% for 10 wt% glycerol feed concentration. Both the glycerol conversion (68.2%) and acrolein selectivity (58.9%) reduced dramatically for 20 wt% glycerol concentration. Indeed, increasing the glycerol concentration significantly decrease the catalyst activity. Haider et al. [47] discussed glycerol condensation (b.p. 290?C) on the catalyst surface was the main reason for reduction of reaction conversion. Furthermore, Foo et al. [48] reported high concentration of glycerol in the feed increased The Keggin anion density decreased from 2.3 to 0.7 HSiW nm 2 by increasing the Al 2 O 3 loading from 10 to 30 wt% on the 30HZ sample. In fact, according to equation (5), larger specific surface area decreased the Keggin anion densities. Fig. 9 exhibits the relationship between glycerol conversion and acrolein selectivity with the Keggin anion density of HSiW atoms. The acrolein selectivity increased from 85.8% to 88.5% with reducing Keggin anion density from 2.3 HSiW nm 2 for 30HZ-10A to 1.4 HSiW nm 2 for 30HZ-20A catalyst. The selectivity reduced to 83.9% over 30HZ- 30A catalyst as the Keggin anion density was further reduced to 0.7 HSiW nm 2. These results are in good agreement with the reduction of acid sites from 0.1 to 0.03 mmol/m 2 reported in the Catalyst performance section. The highest acrolein selectivity (88.5%) was obtained over the catalyst with Keggin density of

11 310 A. Talebian-Kiakalaieh et al. / Journal of Industrial and Engineering Chemistry 34 (2016) Fig. 9. Effect of Keggin-anion density on glycerol conversion and acrolein selectivity at 300?C, 2 ml/h glycerol feed, and 20 ml/min carrier gas flow. 1.4 HSiW nm 2 since the HSiW Keggin structure at this coverage was a consequence of the strong interaction between HSiW and its support. For instance, the 30HZ-30A catalyst possessed the highest surface area (91.3 m 2 /g) compared to the other two supported catalysts (27 and 45 m 2 /g). Since the surface area of 30HZ-30A is the largest, some parts of the support may not be covered by the HSiW leading to less acrolein selectivity. Although the previous study attributed the decomposition of the HSiW Keggin structures to reduced acrolein selectivity [22], the HSiW decomposition was not observed in this study. Used catalyst evaluation Coke deposition Table 5 summarizes the coke content of used catalysts determined by CHNS-elemental analyzer. Increasing the HSiW loading from 10 to 40 wt% surged the coke content from 1.0 to 2.0 wt% in 3 h reaction time. However, the final series of catalysts were registered reduction of coke content from 3.7 to 0.7 wt% by increasing the Al 2 O 3 loading from 10 to 30 wt%. The reduction is attributed to the enlarged surface area from 27.1 to 91.3 m 2 /g and pore volume from 0.1 to 0.26 cm 3 /g compared to the 10 40HZ used catalysts with considerably lower surface areas and pore volumes. The total coke content on the catalyst surface for the 30HZ-20A sample was measured to be 2.9 wt% after 12 h. Generally, catalysts can tolerate about 6 10 wt% coke content with limited amount of activity loss particularly when the coke is deposited on inactive sites [51]. The 2.9 wt% coke deposition on the 30HZ-20A catalyst surface after 12 h reaction time in this study was much lower than 6 10 wt%. Thus, elemental analyzer confirmed only limited amount of coke was deposited on the used 30HZ-20A catalyst surface. As a consequence, the catalyst could still retain its high activity and stability. Table 5 Coke content over used catalysts surface a. Catalyst Coke (wt%) 10HZ HZ HZ HZ HZ-10A HZ-20A 2.2 [2.9] b 30HZ-30A 0.7 a Reaction condition: 300?C reaction temperature, ml/h glycerol feed, and 2 ml/min carrier gas flow, h reaction time. b Coke content after 1 h Thermal stability analysis Thermogravimetric analysis (TGA) of the spent catalyst (30HA- 20A) was also performed and the results compared with the fresh sample in Fig. 10. The TG curves showed 1.5% and 4% weight losses for fresh and used catalysts, respectively in agreement with the acrolein selectivity results. In fact, the fresh catalyst was more active than the used one. The small weight loss difference between the fresh and used catalysts revealed the used catalyst still was highly active even after 12 h reaction time. The DTA curves, attributed to the fresh and used catalysts, exhibited similar peaks at 80?C, and ?C, which referred to the loss of physically adsorbed water and removal of the structural water molecules, respectively. The used sample depicted an endothermic peak at ?C as a result of the HSiW Keggin anions decomposition compared to the fresh catalyst [15,18,28]. Babad-Zakhryapin and Gorunov [31] reported decomposition of supported HSiW catalyst at 540?C was related to the crystallization of WO 3 (Si) considered as a solid solution of silica in tungsten trioxide. In addition, decomposition of the deposited carbonaceous species, formed during the dehydration reaction, is the main reason for sample weight loss [52]. Thus, the results confirmed spent catalyst was still stable at reaction temperatures ranging from 280 to 320?C even after 12 h reaction time and only coke deposition on the catalyst surface caused decomposition at 600?C, way above the reaction temperature range. Surface area and pore structure analysis The surface area, pore volume, average pore diameter, and pore area related to the used 30HZ-20A catalyst are summarized in Table 1. Results confirmed the used catalyst still has high activity even after 12 h reaction time although the surface area was reduced by 20% with negligible pore diameter reduction. Also, the BET analysis revealed the used catalyst still has substantially large pore diameter (27.8 nm) and pore volume (0.16 cm 3 /g) which are suitable for the long-term catalytic activity since coke deposition could only partially filled the pores. Yun et al. [53] reported large pore diameter and Bronsted acidic sites were the main reasons for significant reduction of coke deposition on the catalyst activity and selectivity for acrolein production. Reaction mechanism Glycerol dehydration to acrolein has been widely studied in gas and liquid phases. The major products are acrolein, ethanal (acetaldehyde), acetone, propanal, hydroxyacetone (acetol), allyl alcohol, and acetic acid. There are various proposed mechanisms for formation of these products [1,14,19,54,55], but the mechanisms, reaction rates, and activation energies of most glycerol reactions are still unknown. Thus, identification of the intermediate steps and by-product formation is necessary [2]. Scheme 1 depicts a proposed reaction pathway for dehydration of glycerol to acrolein over supported HSiW catalyst in the current study. One of the major reasons for utilizing the HSiW supported catalysts in this study was the presence of strong Bronsted acidic sites. Katryniok et al. [13] reported Bronsted acidic sites directly protonated the secondary hydroxyl group of glycerol to form acrolein. On the contrary, the Lewis acidic sites favor hydroxyacetone formation due to activation of the glycerol terminal OH groups [1,56]. Therefore, the low amount of hydroxyacetone detected in our research inferred the reaction moved to the desired acrolein formation path. In addition, hydroxyacetone is highly reactive and it is hydrogenated to acetone with acid catalyst, but the intermediate product, 1,2-propanediol, is not detected in the current study. Similar results have been reported elsewhere [56,57]. The hydrogen origins which is involve in the hydrogena-