Sulfur-free Ni catalyst for production of green diesel by hydrodeoxygenation

Journal of Catalysis 347 (2017) 205–221

Contents lists available at ScienceDirect

Journal of Catalysis journal homepage: www.elsevier.com/locate/jcat

Sulfur-free Ni catalyst for production of green diesel by hydrodeoxygenation Imane Hachemi a, Narendra Kumar a, Päivi Mäki-Arvela a, Jorma Roine b, Markus Peurla b, Jarl Hemming a, Jarno Salonen b, Dmitry Yu. Murzin a,⇑ a b

Faculty of Science and Engineering, Åbo Akademi University, Turku, Finland Department of Physics and Astronomy, University of Turku, Turku, Finland

a r t i c l e

i n f o

Article history: Received 9 June 2016 Revised 17 September 2016 Accepted 18 December 2016

Keywords: Hydrodeoxygenation FAME TOFA Triglycerides Green diesel

a b s t r a c t Sulfur-free Ni supported on H-Y zeolites, c-Al2O3 and SiO2 synthesized by the wet impregnation method, and Pd/C were tested in hydrodeoxygenation (HDO) of stearic acid. The catalysts were forming n-heptadecane except Ni/H-Y-80, which was producing n-heptadecane and n-octadecane. Ni/H-Y 80 and Pd/C were tested on HDO of fatty acid methyl esters from chlorella, tall oil fatty acids, and animal fat. The reactions converting the substrates to the final products followed the path from unsaturated esters to acids, with hydrogenation of the latter into alcohols (i.e., stearyl alcohol), and finally formation of hydrocarbons. Ni/H-Y-80 permitted rapid and complete conversion into hydrocarbons, while Pd/C displayed 5–20 times lower turnover frequency, producing saturated intermediates along with hydrocarbons. The catalyst reusability of Ni supported on Y zeolites was studied by recycling and regenerating the spent catalyst from fatty acid HDO. The catalysts demonstrated the possibility of restoring the rates per unit of surface area after regeneration. Catalysts used in HDO of different feedstocks were investigated by thermogravimetric analysis, inductively coupled plasma-optical emission spectroscopy, transmission electron microscopy, surface area measurements, and pore size analysis. Ó 2016 Elsevier Inc. All rights reserved.

1. Introduction Fossil fuel resources are being depleted and the world is gradually adapting to this new paradigm [1]. Biodiesel is a notable alternative to petroleum-derived diesel; however, substituting the latter would take decades, since the production of biodiesel is very small compared to that of petrodiesel. The cost of biodiesel is one disadvantage leading to limited commercial application, even though it has several advantages; the price of raw material constitutes 70–95% of the total biodiesel cost [2–4]. Different possibilities of biodiesel production from biomass resources have been explored in recent years. Biofuel feedstock includes agricultural corps, plant material, animal byproducts, and recycled waste. It is crucial that feedstock for biodiesel production, besides being low-cost, should not compete with food resources. In order not to compete with edible vegetable oils, low-cost profitable biodiesel should be produced from low-cost feedstocks such as nonedible oils, animal fat, soap stocks, and greases. However, the availability of these feed sources is not sufficient to ⇑ Corresponding author. E-mail address: [email protected] (D.Yu. Murzin). http://dx.doi.org/10.1016/j.jcat.2016.12.009 0021-9517/Ó 2016 Elsevier Inc. All rights reserved.

match present-day demands for biodiesel, which are experiencing fast growth. For example, biomass energy consumption in the United States grew more than 60% from 2002 to 2013 [5]. In general, the feedstock for biodiesel production can be divided into four groups: vegetable oils (edible and nonedible), animal fats, used cooking oils, and algae [6]. Algae are considered a potential feedstock for secondgeneration biodiesel. Microalgae do not compete with food resources, as they can be cultivated in open ponds on nonarable lands using sidestream water and CO2. They have a higher growth rate than land crops, requiring much less area than other biodiesel feedstocks of agricultural origin. In 2006, worldwide aquacultural macroalgae production was estimated to be approximately 15.1 million tons (worth about €5.4 billion) [7]. In 2011, Norsker et al. [8] estimated algae production cost to be 400 €/1000 kg, while the value of algae biomass is 1650 €/1000 kg. Algae are firmly recognized as a potential source for direct production of bioenergy, food, pharmaceuticals, and nutraceuticals [9,10]. Nowadays, microalgae are seen as an alternative feedstock for biodiesel production, being the target of a large number of consortia, private and public organizations that are investing in R&D aiming at the most effective and cheapest technology for producing large

206

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

amounts of algal oil [11]. Chlorella algae contain fatty acid methyl esters (FAME), typically a mixture of saturated and unsaturated fatty acids [12]. Tall oil fatty acid (TOFA) is a byproduct of the pulp and paper industry. The use of TOFA as a feedstock for the production of biodiesel is of great interest, since it does not compete with food production and has the potential to be used in hydrodeoxygenation (HDO) processes. In Europe, especially Finland and Sweden, utilization of tall oil for fuel has been popular, given the vast availability of the feedstock. The two countries provide 90% of total EU production and 80% of EU consumption of tall oil, according to industry sources. One big producer in Sweden is Sunpine, which started producing 830 ktons/yr of tall oil-based biodiesel in mid-2010 in Pitea [13]. Tall oil is used in various applications, such as paints and coatings, biolubricants, and performance polymers, and can also be used in pharmaceuticals and health-enhancing food additives. In the biofuels industry, tall oil fatty acids mainly with 18-carbon chains are used as a feedstock for green diesel production. These acids undergo oxygen removal and saturation to be converted at the end into n-heptadecane and/or n-octadecane. In Finland, there is production of green diesel from TOFA on an industrial scale by UPM [13,14]. Waste animal fat is a promising cheap alternative feedstock for the production of biodiesel that does not compete with food resources. In addition, the use of animal fat is considered profitable because these fats frequently offer an economic advantage over plant crops, often being favorably priced [15]. Animal fat is a completely renewable source of energy. It is biodegradable, nontoxic, and safe to use in diesel engines without modification. Multiple animal fat feedstocks for biodiesel fuels are making animal fatbased biodiesel of high quality. In the United States, the meat production system produces nearly 10,000 tons of a byproduct in the form of various types of fats and proteins [16]. Animal fats consist essentially of fatty acid triglycerides. A number of catalytic methods have been established for upgrading biomass feedstock into biofuels. These routes include HDO over sulfided catalysts such as alumina-supported NiMo and CoMo [17–22], or over noble metals namely palladium and platinum catalysts [20,21,23–27], as well as cracking and catalytic pyrolysis [28–30]. HDO is an attractive method that has been widely used for conversion of fatty acids into green diesel. In previous studies [31–33], palladium supported on carbon was demonstrated to be the best catalyst for deoxygenation of fatty acids. Noble metals are advantageous since they have high activity without a need for sulfidation. However, they are less economically attractive than nickel. In this work, we have synthesized Ni catalysts supported on H-Y, c-Al2O3, and SiO2 by wetness impregnation and compared their catalytic performance in HDO of stearic acid (model compound of fatty acids) to produce diesel-range hydrocarbons with that of Pd supported on active carbon. HDO of four feedstocks was thereafter investigated over Ni/H-Y and Pd/C in order to determine activity with different substrates. Reuse of these catalysts has been also studied. The properties of the catalysts were characterized by various physicochemical methods.

2. Experimental 2.1. Catalyst preparation A wetness impregnation evaporation method was used for the preparation of 5 wt.% Ni (Fluka, 98%) on SiO2 (Silica Gel, Merck), sieved H-Y-80 zeolites (Zeolyst International, SiO2/Al2O3 = 80), and c-Al2O3 (Versal Alumina VGL-25, UOP) with a size below 63 mm. H-Y support was calcined in air at 400 °C for 4 h, followed

by sieving. Thereafter, nickel(II) nitrate hexahydrate was dissolved in 100 ml deionized water, giving a pH equal to 6.1, followed by addition of 10 g of H-Y to the solution, resulting in a decrease of pH to 3.2. The mixture remained under continuous stirring for 24 h in a 60 °C oil bath before water was evaporated and impregnation of Ni on the zeolites was permitted. Thereafter, the catalyst was dried overnight at 100 °C and then calcined at 400 °C for 3 h. Similar methods were used to synthesize 5 wt.% Ni/c-Al2O3 and 5 wt.% Ni/SiO2. A quantity of 5 wt.% Pd/C (20,568-0) was purchased from Sigma-Aldrich.

2.2. Catalyst characterization methods The specific surface area of the catalysts was measured by nitrogen physisorption using a Sorptometer 1900 apparatus (Carlo-Erba Instruments). Prior to measurements, the fresh samples and those spent in HDO of FAME and TOFA were outgassed at 150 °C for 3 h, and the catalysts spent in HDO of animal fat were outgassed at 300 °C for 6 h. The specific surface area was determined using the BET equation, while the pore diameter was determined using the Barrett-Joyner-Halenda method [34]. The spent catalysts were collected, washed with acetone, and dried at 70 °C for 24 h prior to nitrogen adsorption analysis. Reduction temperature was determined by temperatureprogrammed reduction (TPR) performed with an Autochem Micrometrics 2910. The catalysts were kept overnight in an oven at 100 °C; thereafter, around 100 mg of the sample was loaded into a U-shaped quartz tube and heated to 650 °C at a constant rate of 10 °C/min in the hydrogen reductive medium. Palladium dispersion was measured by pulse CO chemisorption with an Autochem 2910 apparatus with 10% CO in 90% He. Prior to measurements, the sample was kept overnight at 100 °C, and thereafter reduced in hydrogen at 250 °C for 2 h at 5 °C/min prior to the analysis. The sample was then flushed with inert gas for 30 min. The pulse chemisorption was performed in a water bath at a monitored temperature. The stoichiometry CO:Pd was assumed to be 2:1. Acid and basic properties of nickel supported on H-Y and the parent H-Y were determined by ammonia and carbon dioxide temperature-programmed desorption (TPD), respectively, performed with an Autochem Micrometrics 2910. The catalyst and the bare support were kept overnight in an oven at 100 °C prior to the analysis. The sample was heated in He to 300 °C at a rate of 30 °C/min and kept at this temperature for 10 min, after which it was cooled down and flushed with helium, followed by heating to 100 °C. The measurements began with the increase of temperature to 900 °C with a heating ramp of 20 °C/min, which remained constant for 30 min. The morphology and crystal size distribution of fresh and spent Ni supported on H-Y samples were studied by a scanning electron microscope (Zeiss Leo 1530 Gemini) equipped with a ThermoNORAN vantage X-ray detector. An average of 200 particles were counted for the crystal size distribution. Energy-dispersive X-ray analysis (EDXA) was carried out with the same instrument. Nickel and palladium particle size distributions were measured from images obtained by a JEM-1400 plus TEM (transmission electron microscope, voltage 120 kV). The samples were dissolved in ethanol in an ultrasonic bath, and 100 ml of the suspension was fixed on a carbon film-grid. An average of 200 particles were counted for the particle size distribution. Thermogravimetric analysis (TGA) was carried out to measure coke formed on the spent catalysts. A CHAN D-200 instrument was used to monitor real-time weight loss as a function of time. A quantity of 45 mg was loaded into a quartz holder and burnt at up to 600 °C in air with the heating rate of 5 °C/min to 450 °C,

207

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

holding at this temperature from 15 min, and a heating rate of 20 °C/min to 600 °C. The content of metal in the fresh and spent catalysts was determined by inductively coupled plasma optical emission spectroscopy (ICP-OES), using an Optima 4300 DV optical atomic emission spectrometer. For the analysis, the catalysts were dissolved in a microwave oven by a mixture of acids (4 ml hydrofuran (40%), 1 ml HNO3 (65%), and 1 ml hydrochloric acid (30%)). Fresh and spent catalysts were characterized by X-ray diffraction (XRD) using a Philips X’Pert Pro MPD X-ray powder diffractometer that was operated in the Bragg–Brentano diffraction mode. Monochromatized CuKa radiation was generated with a voltage of 40 kV and a current of 45 mA. The primary X-ray beam was collimated with a fixed 0.25° divergence slit and a fixed 10 mm mask. A fixed 7.5 mm antiscatter slit was used for the diffracted beam. The measured 2h angle range was 5.0°–70.0°, with a step size of 0.026°and a measurement time of 10.0 s per step. The samples were measured on Al sample holders. 2.3. Catalyst regeneration Regeneration was applied to spent Ni catalysts obtained from fatty acid HDO. The spent catalyst was filtered, followed by washing with methanol and acetone in order to remove the organics, which are in liquid form at room temperature. Drying at 100 °C for 12 h was performed prior to oxidative regeneration of the sample, which was carried out in air at 400 °C for 2 h; the heating ramp was kept constant at 2 °C/min. 2.4. Raw material Stearic acid (Merck, 97%) was used as a model compound to study the activity of different catalysts. Fatty acid methyl esters (FAME) were obtained from Chlorella microalgae (Fuqing King, Drarmsa Spirulina Co, Ltd, China) by in situ transesterification, which was performed at 60 °C for 4 h in 20 wt.% H2SO4, followed by purification with a clay (Argila verde, www.forcadaterra.com, Brazil). FAME is composed mainly of saturated and unsaturated esters with the respective fractions 62% and 34% [35]. Commercial distillate tall oil fatty acids (TOFA) obtained from the pulp and paper industry contained mainly C18 with different degrees of unsaturation (1.5 wt.% 18:0, 25.1 wt.% 18:1, 53.4 wt.% 18:2, and 18 wt.% 18:3). Animal fat containing mainly C18 (17% C18:0, 43.4% C18:1) and C16 (24.1% C16:0, 3.1% C16:1) and longer fatty acids was provided by a refinery. SEC analysis shows that the animal fat contains 93.57% triglycerides, 5.21% diglycerides, and 1.43% monoglycerides [36]. 2.5. Reaction and analysis HDO of fatty acids was carried in a semi-batch reactor at 300 °C under 30 bar H2. The reaction was run in a 300 ml stirred reactor equipped with a heating jacket. In all experiments, the catalyst

was charged into the reactor with 40 ml of dodecane (Sigma Aldrich,
90%). The reactor was flushed with argon to remove oxygen. The reactant was injected into a prereactor and bubbled with argon before being put into the reactor. The HDO started when the substrate was introduced into the reactor at the desired temperature and pressure. The same weights of catalyst and reactant were used for all runs. The stirring rate corresponded to 1200 rpm. Prior to the run, the catalyst underwent a reduction step at a reduction temperature determined by TPR of nickel and palladium supported catalysts. The reduction was carried on for 2 h. Liquid samples were taken from the reactor using a two-valve system. At 30 s prior to the sample taking, the sampling nozzle was purged. Samples were collected in glass vials and silylated thereafter, prior to analysis by gas chromatography (GC). The silylated samples were prepared by the following procedure: 100 ml of a sample was diluted in 1 ml of pyridine (
99%, Sigma–Aldrich); 100 ml of eicosane (99%, Acros-Organics) as an internal standard, 100 ml of BSTFA (N, O-bis(trimethylsilyl)-trifluoroacetamide, >98%, Acros-Organics) as a silylation agent, and 50 ml of TMCS (chlorotrimethylsilane,
98%, Sigma–Aldrich) were added. The sample was kept in an oven at 70 °C for 1 h before being injected into the GC equipped with a split column (film thickness 0.25 mm) under He pressure 1 bar flowing at 75 ml/min. The column was heated to 300 °C and the signal was recorded with a FID detector. The triglyceride fraction was analyzed with the following procedure: 2 ml of MTBE containing the internal standards (0.04 mg each) was added to the samples. The internal standards used were (21:0) heneicosanoic acid (Sigma 99%), betulinol (purified in our lab), cholesteryl heptadecanoate (Sigma > 95%), and 1,3-dipalmitoyl-2-oleoylglycerol (Sigma 99%). MTBE was evaporated under a nitrogen stream at 40 °C until only docosanol remained. The evaporation was repeated after 1 ml of acetone was added. A quantity of 150 ml of the silylating reagent mixture (pyridine-BSTFA-TMCS 1:4:1) was added and then the final mixture was put into a 70 °C oven for 45 min. The silylated sample was injected into a PerkinElmer Clarus 680 gas chromatograph equipped with a FID detector and an Agilent J&W HP-1/SIMDIST column (film thickness 0.15 mm) using H2 (7 ml/min) as carrier gas. The column was heated to 340 °C. 3. Results and discussion 3.1. Catalyst characterization Ni content in fresh samples analyzed by ICP-OES was found to be 4.9%. The physical properties of the selected catalysts for HDO of stearic acid are summarized in Table 1. The specific surface area of the catalyst varied from the moderately high surface area of Ni/H-Y 80 (660 m2/g) to the medium values of Ni/SiO2 (588 m2/g) and Pd/C (551 m2/g) and to the low surface area of Ni/c-Al2O3 (258 m2/g). The large specific surface area of Ni/H-Y is due to the contribution of the internal pores. The impregnation of Ni on H-Y 80 (994 m2/g) resulted in a decrease of the specific surface area by 33% and of the pore volume from 0.32 to 0.22 cm3/g. The increase in the pore diameter (from 0.97 to 1.81 nm) is caused by

Table 1 Physicochemical properties of fresh materials [40].

H-Y 80 5 wt.% Ni/H-Y 80 5 wt.% Ni/c-Al2O3 5 wt.% Ni/SiO2 5 wt.% Pd/C

Support particle size (mm)

Surface area (m2/g)

Pore volume (cm3/g)

Pore average diameter (nm)

Total NH3 adsorption (mmol/gcat)

Total CO2 adsorption (mmol/gcat)

<63 <63 <63 40–63 –

995 660 260 590 550

0.32 0.22 0.09 0.12 0.10

0.97 1.81 2.06 1.44 0.99

1.77 1.47 – – –

134.23 99.38 – – –

208

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

Fig. 1. Transmission electron microscope micrographs and metal particles distribution for (a) 5 wt.% Ni/H-Y and (b) 5 wt.% Pd/C.

the synthesis conditions, which, besides possible pore blocking, can have a large impact on the support structure, as reported in [37,38]. These results are in agreement with those reported in previous studies, where Zuo et al. [39] reported 7 wt.% Ni/H-Y having a specific surface area of 616 m2/g. Ni/SiO2 (average particle pore diameter of the silica gel is 6 nm) exhibited a similar value of the specific surface area, but a smaller pore volume. The catalyst acidity and basicity were determined by NH3 TPD and CO2 TPD, respectively, for Ni supported on H-Y and the parent H-Y. The total acidity and basicity demonstrated a small relative decrease after impregnation with 5 wt.% Ni (Table 1). Basicity of microporous materials was discussed recently as being influenced by the amount of metal, interactions between the metal and the support, and synthesis conditions, such as calcination temperature [41]. The nickel and palladium size distributions were measured by transmission electron microscopy (TEM). Fig. 1 displays average Ni and Pd particle sizes, showing average Ni particle diameter of 4 nm, while average Pd particle diameter was found to be 2.5 nm. The validity of average Pd size was successfully confirmed by CO chemisorption, showing a metal dispersion of 60%.

The Weisz–Prater criterion (Eq. (1)) depends strongly on the reaction order. Values of / below 6, 1 and 0.3 respectively for zero-, first-, and second-order reactions can be considered sufficient conditions for avoiding significant pore diffusion limitations. The effective diffusivity was evaluated using the Wilke–Chang equation [43],

D0AB ¼

7:4  108 ð/M B Þ1=2 T

gB  V 0:6 bðAÞ

½cm2 =s

ð2Þ

The dimensionless association factor / was taken as unity for dodecane, MB is the molecular weight of the solvent (170 g/mol), gB is the solvent viscosity at reaction temperature (0.2 cP), and Vb(A) is the liquid molar volume at the solute’s normal boiling point (cm3/mol). The diffusion coefficients for stearic acid and hydrogen, respectively, are calculated to be DSA=C 12 ¼ 1:9  107 m2 =s. DH2 =C 12 ¼ 0:15  107 m2 =s. 73 K, 30 bar), assuming that n/s = 1/10. For the maximal reaction rate (5.8  107 mol/L s) obtained from stearic acid HDO over Ni/H-Y, the values of the Weisz–Prater modulus for stearic acid and hydrogen are several orders below the criteria. This indicates that the substrate diffusion inside the catalyst pores does not affect the reaction rate.

3.2. Hydrodeoxygenation results 3.2.1. Influence of reaction parameters on mass transfer A parametric study was carried out to optimize and determine the influence of reaction conditions on mass transfer. For this purpose, the stirring speed was varied in the range 1000–1200 rpm, and the amount of catalyst used in the HDO reaction was varied from 0.1 to 0.4 g. The results demonstrated 1200 rpm to be a sufficient stirring speed for fatty acid HDO, overcoming external mass transfer limitations. Increasing the catalyst amount leads to an increase in reactor productivity. An amount of 0.25 g is sufficient to ensure the absence of gas/liquid mass transfer limitation. The influence of pore diffusion on the reaction rate was determined using the Weisz–Prater criterion [42],

/WP ¼

r obs  R2 cDeff

ð1Þ

In Eq. (1), /WP is the dimensionless Weisz–Prater parameter, R is the mean catalyst particle radius (63 mm), robs is the observed rate per catalyst volume, c is the substrate concentration, and Deff is the effective diffusion coefficient, defined as Deff = D(n/s), where D is the substrate diffusion coefficient in the solvent and n,s are catalyst porosity and tortuosity, respectively, and have typical values in the ranges 0.3–0.6 and 2–5.

3.3. Catalytic screening for sulfur-free hydrodeoxygenation of stearic acid To study the catalytic performance, HDO of stearic acid was investigated over a variety of catalysts under the same reaction conditions. An initial catalyst screening was performed to determine the influence of the support on catalytic activity and to compare the sulfur-free catalysts with a noble metal catalyst. The concentration curves for different catalysts are presented in Fig. 2. Pd/C and Ni supported on H-Y, c-Al2O3, and SiO2 catalysts were reduced prior to the HDO reaction at the respective reduction temperatures of 350, 450, 350, and 200 °C, determined by TPR [40]. Stearic acid conversion was tested in HDO over metal-free H-Y under the same reaction conditions (300 °C, 30 bar). The absence of any activity highlights the necessity for a metal-containing catalyst. The choice of H-Y-80 was based on a preliminary screening study, where H-BEA and H-Y zeolites with several different SiO2/ Al2O3 ratios were used in the synthesis of 5 wt.% Ni by a wet imp regnation–evaporation method. Hydrodeoxygenation of stearic acid showed complete conversion of stearic acid for Ni supported catalysts. Although many previous studies of fatty acid deoxygenation over NiMo and CoMo supported catalysts have been carried out where a complete conversion was attained, giving hydrocarbons with the same carbon number as the starting substrate,

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

209

Fig. 2. Hydrodeoxygenation of stearic acid over (a) H-Y-80, (b) 5 wt.% Ni/H-Y-80, (c) 5 wt.% Ni/c-Al2O3, (d) 5 wt.% Ni/SiO2, and (e) Pd/C in a semibatch at 300 °C and 30 bar.

addition of sulfur was required. For example, in [18], HDO of a mixture of C7 fatty acids in a batch reactor at 250 °C under 7.5 MPa H2 was performed over 0.5 g presulfided NiMo/c-Al2O3 in 5 vol.% H2S/H2. Sulfur-free Ni catalysts with high metal loading [44] (>50 wt.%) were used in deoxygenation of stearic acid at 300 °C and 6 bar, giving conversion lower than 20% after 6 h. In another work, sulfur-free Ni/H-Beta prepared by impregnation (3.3% dispersion, Ni particle size 15 nm) gave 33% conversion of stearic acid after 2 h of HDO at 260 °C and 40 bar H2, while complete conversion was attained when the preparation method allowed smaller Ni to be formed (2.2–3.9 nm) [45]. Turnover frequencies for hydrogenation of stearic acid were reported to be in the range 100–200 h1 for Ni-BEA catalysts [45]. The authors commented on the essential structure insensitivity of the reaction.

As will be shown below, the normalized rates reported in the literature are similar to the values of normalized rates obtained in the current work. Transformation of stearic acid was more rapid over Ni/H-Y than over Ni/c-Al2O3 and Ni/SiO2, giving respective conversions values after 90 min of 94%, 43%, and 46%. Conversion over Pd/C reached 54% after 6 h of reaction. Pd used in stearic acid HDO was reported to allow fast and complete conversion when supported on activated carbon [23,44]. The main products formed were aliphatic hydrocarbons (n-heptadecane and n-octadecane) obtained in different ratios over different catalysts. C17 was the main product from HDO of stearic acid over Pd/C, Ni/c-Al2O3, and Ni/SiO2, representing 50% of products obtained from HDO over Pd and more than 94% from HDO over

210

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

Ni supported on c-Al2O3 and SiO2. HDO of stearic acid over Ni/H-Y led to production of C18 and C17 hydrocarbons with the respective proportions 60% and 36%. Conversion of stearic acid over Ni/H-Y followed several pathways allowing formation of C18 and C17 products, as illustrated in Scheme 1. Selectivity values in the present work are in agreement with several studies [23,32,39,44,46,47]. As an intermediate product, stearyl alcohol appeared in small amounts in the first 2 h of HDO over Ni/c-Al2O3 and Ni/SiO2. This compound disappeared completely after the complete conversion of stearic acid. Ni/H-Y exhibited the highest rate among the tested catalysts for hydrodeoxygenation of stearic acid. Influence of the support and its acidic properties can be discussed from several angles. The support acidity might have an effect on the catalyst electronic structure, influencing activity and selectivity [48–50]. Alternatively, acid sites can be involved directly in the reactions. Similarly to the present work, where the least acidic catalysts promote decarboxylation and decarbonylation, Wu et al. also reported high selectivity to C17 products in

decarboxylation of fatty acids over nickel supported on nonacidic activated carbon [51]. A plausible explanation of high selectivity toward C18 products when acidic supports are used can be related to sequential hydrogenation of fatty acids to aldehydes and alcohols on nickel metal clusters followed by dehydration of fatty alcohols on Brønsted acid sites to olefins (i.e., octadecene in Scheme 1) and hydrogenation of the double bond on the metal. This explanation of selectivity toward C18 products has been proposed recently by Lercher and co-workers [27]. Selectivity to C17 depends on conversion of stearic acid, increasing from 30% to 96% over Ni/c-Al2O3, 20% to 90% over Ni/SiO2, and 68% to 94% over Pd/C as the reaction was progressing. However, for Ni/H-Y, selectivity showed a fast increase from 24% to 50%, corresponding to a conversion of 30% of stearic acid, followed by a decrease to 37%. The opposite behavior was observed for the selectivity toward C18, which first decreases from 75% to 50%, followed by an increase to 62% with the increase of conversion (Fig. 3). A decrease of selectivity to C18 alkanes was noticed for Ni/Al2O3, Ni/SiO2, and Pd/C, with selectivity at higher conversions

Scheme 1. Different pathways for triglyceride and fatty acid methyl ester HDO. (I) Hydrogenation, (II) decarbonylation, (III) decarboxylation, (IV) hydrogenolysis, and (V) dehydration/hydrogenation.

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

211

Fig. 3. Selectivity toward n-heptadecane and n-octadecane in HDO of stearic acid over (a) Ni/H-Y-80, (b) Pd/C, (c) Ni/Al2O3, and (d) Ni/SiO2 in HDO of stearic acid.

being close to 1%. Ni/Y zeolite is twofold more selective to the hydrogenation pathway (C18) than to decarboxylation/decarbony lation (C17), as was shown in an early study [45], while Pd is selective to C17 through decarboxylation/decarbonylation pathways as extensively reported in previous studies [23,31,33,44]. Selectivity to stearyl alcohol shows an initial minor increase for Ni/Al2O3 and Ni/SiO2 and then a decrease from 61% and 43% respectively to zero at complete conversion of stearic acid. The selectivity to stearyl alcohol on Pd/C shows an initial decrease of 13%, corresponding to 54% conversion of stearic acid after 6 h of HDO. It was clearly visible that Ni/c-Al2O3 and Ni/SiO2 demonstrated similar conversion and selectivity. Recently, Santillan-Jimenez et al. demonstrated that Ni/C also shows high selectivity to C17 [48]. 3.4. Comparative study of the effect of different feeds on the catalytic performance of Ni/H-Y and Pd/C HDO of fatty acid resources presents a complex network of multiple parallel and consecutive reactions. As illustrated in Scheme 1, hydrocarbons are produced by two main reaction pathways, hydrogenation and decarboxylation/decarbonylation, resulting respectively in alkanes with the same carbon number as the reactant or one carbon number less. The intermediates are produced by hydrogenation; moreover, hydrogenation of the different feedstocks (methyl esters in FAME, triglycerides in animal fat) led to formation of saturated fatty acids as common intermediates.

Hydrogenation of the acids gives aldehydes, which, depending on the catalyst, either preferentially undergo decarboxylation into heptadecane or can be hydrogenated to alcohols. Aldehydes were not observed in the reaction mixture, suggesting that this intermediate is strongly adsorbed onto the surface of nickel and is transformed further without leaving the surface. Subsequent transformations of alcohols occur most probably through dehydration on acid sites to octadecene and hydrogenation of the latter. In the case of methyl esters, initial saturation occurred, producing methyl stearate and methyl palmitate. Hydrogenation of the latter compounds resulted in formation of hydrocarbons. A more detailed investigation of the activity and selectivity of Ni/H-Y and Pd/C catalysts was conducted in HDO of FAME, TOFA, and animal fat. Ni catalysts displayed high activity, resulting in nearly complete conversion of the different feedstocks into hydrocarbons in approximately 2 h. Pd catalysts demonstrated lower rates in HDO of the studied feedstocks. Complete conversion into hydrocarbons was not attained after 6 h of the reaction, and the final yield of intermediate compounds was higher than that of the desired products. HDO of FAME, TOFA, and triglycerides over Ni catalysts (Fig. 4a– c) exhibited a yield of hydrocarbons higher than 90% in less than 150 min with respective turnover frequencies (TOFs) of 0.02, 0.02, and 0.005 s1. Lercher and co-workers [27] reported similar values ca. 0.04–0.06 s1 for HDO of stearic acid over Ni-HBEA catalysts. The initial triglyceride conversion was lower than that of

212

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

Fig. 4. Conversion of (a) FAME, (b) TODA, and (c) animal fat over 5 wt.% Ni/H-Y-80 and (d) FAME, (e) TOFA, and (f) animal fat over 5 wt.% P/C.

FAME and TOFA. Previously, 78% conversion was achieved when HDO of FAME from jatropha was performed over bimetallic 20 wt.% NiCo supported on silica and alumina in the temperature range 300–400 °C at 20 bar [49]. Other studies showed that total conversion of fatty acid could be reached with metal loadings higher than the metal loading used in this study (30 g methyl palmitate: 1 g 7 wt.% Ni/H-Y [39]:0.02 g 20 wt.% Ni/HZSM-5 [50]). The yield of acids as major intermediates in the case of FAME increased to a maximum level of 23% in 30 min, subsequently decreasing as a function of time. Similar behavior was noticed in triglyceride HDO; however, the maximum was observed at 15 min with a yield of 21%. Stearic and palmitic acids were the main intermediates in conversion of FAME from Chlorella microalgae. Only traces of stearyl alcohol were found, since, as stated before, formation and transformation of this intermediate are very rapid. In the case of TOFA, the reaction intermediates consisted of stearyl alcohol, which was found in trace

amounts. HDO over Ni catalysts followed hydrogenation and decar boxylation/decarbonylation pathways, yielding twofold alkanes from hydrogenation relative to decarboxylation/decarbonylation. The respective yields for FAME HDO products are 30% n-octadecane and 29% n-hexadecane from hydrogenation and 18% n-heptadecane and 14% n-pentadecane from decarboxylation/ decarbonylation. In the case of TOFA, the yield of hydrocarbons reached 93%, with the products consisting mainly of n-octadecane and n-heptadecane. Triglyceride HDO yielded 38% n-octadecane, 24% n-heptadecane, 22% n-hexadecane, and 10% n-pentadecane. Note that the majority of previous studies [39,46,51,52] reported that HDO over Ni supported catalysts leads to C–C bond cleavage, resulting in shortening of carbon chains in the products. FAME and TOFA HDO over Pd showed an increase of products and intermediates yield, however, with lower TOFs being 0.001, 0.002, and 0.016 s1, respectively. In the case of FAME, the yield of intermediates (mainly acids) was twice the yield of hydrocar-

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221 Table 2 TOF values for HDO of different feedstocks. TOF (s1) (molreactant/molsurf.

Stearic acid FAME TOFA Animal fat

Metal.

s)

Ni

Pd

0.01 0.02 0.02 0.005

0.002 0.001 0.002 0.016

Note: Catalyst dispersions were obtained from CO chemisorption for Pd (60%) and TEM for Ni (24%).

213

hydrocarbons compared with selectivity profiles obtained in FAME and TOFA HDO. Animal fat HDO over Pd demonstrated a different profile of selectivities, where a constant selectivity of 90% toward the acids as the intermediates in hydrogenation of triglycerides over Pd remained unchanged until 95% conversion of triglycerides. Thereafter, it decreased considerably, resulting in a major increase of hydrocarbon selectivity, as illustrated in Fig. 6f. The observed selectivity of Pd was prevailing to hydrocarbons generated from decarboxylation/decarbonylation. 3.5. Long-term reaction

bons, as shown in Fig. 4d. After 6 h of HDO, n-octadecane and npentadecane were the main products, with respective yields 8% and 6%, in addition to 3% n-octadecane as a minor product. TOFA HDO on Pd/C yielded 40% hydrocarbons, mainly n-octadecane (38%). Higher yields were obtained from HDO over Pd loaded on different supports, such as mesoporous carbon Sibunit [52] and HPS (hypercrosslinked polystyrene) [53]. In FAME and TOFA HDO over Pd/C, 80% and 95% of hydrocarbons formed, respectively, were from the decarboxylation/decarbonylation pathway. Complete transformation of triglycerides into acids was successfully achieved in 1 h. The intermediates yield showed a fast increase in 1 h, reaching 70% and slowly decreasing thereafter as a function of time (Fig. 4f). The Pd/C catalyst permitted faster conversion of triglycerides than the nickel catalyst. Transformations of triglycerides includes first hydrolysis to acids. In fact, conversion of acids into hydrocarbons over Pd/C was slow, yielding 47% hydrocarbons after 6 h of reaction. The acidity of Ni-HY-80 is apparently much higher than that of Pd/C; thus, differences in triglycerides hydrolysis cannot be attributed solely to acidic properties of the support and can probably be related also to accessibility of the acid sites. The decarboxylation/decarbonylation path prevailed in animal fat HDO, forming n-octadecane and n-pentadecane as the main products with respective yields 3% and 15%. Table 2 displays reaction rates normalized to the number of exposed surface sites. TOFs for HDO of different substrates over Ni appear to be 5–20 times higher than over Pd, with the exception of triglycerides (animal fat), where Pd exhibited TOF values threefold higher than Ni. The product selectivities as a function of the fatty acid conversion are presented in Figs. 5 and 6. Selectivity of Ni/H-Y catalysts in HDO of fatty acid feedstocks toward the desired aliphatic hydrocarbons varied with time, reaching 100% as the conversion of the fatty acid feedstocks approached 100% (Fig. 6). Selectivity values at low conversion are related to high selectivity toward the intermediates, which were subsequently converted into hydrocarbons. Depending on the substrate, Ni and Pd exhibited different selectivity profiles toward the products. For both catalysts, FAME HDO showed the highest selectivity toward saturated FAME, the first intermediate compounds in hydrogenation of FAME at low conversion levels. Concentration of saturated FAME tended to decrease gradually, permitting an increase of selectivity to hydrocarbons in the case of Ni and acids in the case of Pd (Fig. 6a and d). The selectivity to acids in FAME HDO over Ni remained approximately constant (23%) until 80%, as they were being produced continuously from saturated esters and transformed into hydrocarbons, decreasing rapidly thereafter. Selectivity of Pd to hydrocarbons exhibited a slow increase as the conversion of FAME increased, reaching 18% at 55% conversion. In the case of TOFA HDO at low conversion of TOFA, the catalyst showed a high selectivity toward stearic acid (Fig. 6b and e). At higher conversions, this selectivity gradually decreased, along with an increase of selectivity toward aliphatic hydrocarbons. Animal fat HDO over Ni (Fig. 6c) exhibited a faster decrease in selectivity toward acids as the conversion of triglycerides increased and a faster increase in selectivity toward

Pd manifested a lower specific rate and required a longer period of time to achieve complete conversion. A prolonged experiment carried out with FAME as the substrate is shown in Fig. 7a. The reaction conditions were kept stable throughout the reaction. FAME HDO over Pd/C catalyst required 30 h to attain complete conversion, producing 80% hydrocarbons (Fig. 7b). 3.6. Decarboxylation of FAME over 5 wt.% Pd/C Selectivity of Pd/C catalyst to products obtained from the decar boxylation/decarbonylation pathway in the previous experiment was observed to be higher than 90%. Decarboxylation of FAME performed in an argon flow was carried out in a semi-batch reactor at 300 °C and 15 bar argon. The conversion of FAME against reaction time and the selectivity as a function of conversion are plotted in Fig. 8a and b, respectively. Decarboxylation was relatively fast in the first 15 min, with an initial rate of 0.03 mmol/L g min. During this time the saturated methyl esters were converted into acids, since the fraction of saturated methyl esters in the feedstock was low. Conversion of FAME over Pd supported on carbon reached 16% after 6 h, yielding aliphatic hydrocarbons. In recent studies, Pd supported on carbon was demonstrated to have a rate in fatty acid decarboxylation; however, using another type of carbon support, 5 wt.% Pd was, for example [25], supported on a mesoporous carbon Sibunit (400 m2/g), allowing 20% decarboxylation of lauric acid after 6 h of reaction (20 bar Ar). A sample of 1 wt.% Pd supported on mesoporous carbon Sibunit was tested in HDO of linoleic [54], oleic, and stearic acids (1 vol.% H2 in Ar). Depending on the saturation of the substrate, different kinetic profiles were obtained. Snåre et al. [23] studied decarboxylation of oleic acid over Pd/C catalyst similar to the one used in this study at 300 °C under 15 bar Ar, concluding that mainly stearic acid was produced and 50% conversion of the reactant was attained. The values of TOF for stearic acid decarboxylation in the presence of hydrogen were in the range 0.02–0.04 s1 and were reported to be dependent on the metal cluster size [54]. The selectivity of the Pd/C catalyst to saturated methyl esters was high (80%) and remained nearly unchanged as a result of the low reaction rates (Fig. 8). The same pattern was observed for the selectivity toward the intermediates. 3.7. Catalyst recycling and regeneration For a catalyst to be commercially viable, it is essential that it should be stable and resistant to deactivation. In the context of the current work performed in a batch reactor, catalyst stability means that the catalyst can be reused and recycled in a number of repeated reactions without demonstrating considerable activity loss [55]. The nickel catalyst described above was revealed to be active under the previous reaction conditions, providing high yields. To determine the catalyst stability and reusability, further investiga-

214

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

Fig. 5. Selectivity to hydrogenation vs. decarboxylation/decarbonylation of 5 wt.% Ni/H-Y-80 in HDO of (a) FAME, (b) TOFA, and (c) animal fat and selectivity of 5 wt.% Pd/C in HDO of (d) FAME, (e) TOFA, and (f) animal fat.

tions were carried under the same conditions over recycled and regenerated catalysts. Reusability studies of Ni catalysts were performed in a semibatch reactor at 300 °C and 30 bar H2. The catalyst was prereduced at reduction temperatures obtained from TPR and thereafter

loaded into the reactor. The substrate/catalyst ratio was kept the same in all experiments. For comparison, Pd/C catalysts were also studied. Fig. 9 represents conversion of FAME over these two catalysts. The results show that the catalyst performance was substantially altered after

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

215

Fig. 6. Selectivity of 5 wt.% Ni/H-Y-80 in HDO of (a) FAME, (b) TOFA, and (c) animal fat and selectivity of 5 wt.% Pd/C in HDO of (d) FAME, (e) TOFA, and (f) animal fat.

the first run, since after the second run the specific rate was ca. fivefold lower in FAME HDO over Ni/HY.

The activity decrease was caused by the specific surface area decrease, as visible from Table 3. After the first run, the surface area decreased by 39% and 22% for Ni and Pd, respectively. Another

216

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

Fig. 9. FAME conversion over fresh (1) and reused (2) 5 wt.% Ni/H-Y-80 and 5 wt.% Pd/C.

Fig. 7. Conversion of FAME after 30 h of HDO at 300 °C and 30 bar in a semi-batch reactor over (a) product distribution over 5 wt.% Pd/C and (b) selectivity to C17 and C18 hydrocarbons.

factor is the decrease in the pore size, which renders some micropores inaccessible, thereby decreasing the number of accessible active sites. The metal particles exhibited a minor size increase for both catalysts, according to TEM. As a consequence, the catalyst reuse in the second run led to a considerable decrease in reaction rates.

The zeolite-containing catalysts tend to become deactivated and are generally regenerated at high temperature with oxygencontaining gas to restore their activity [55,56]. Therefore, to recover the catalyst activity, the Ni catalyst was regenerated after the first run and reused in TOFA HDO. To study the effect of Ni recycling on different reactants alongside its regeneration, TOFA HDO was performed under similar conditions. Fatty acid conversion and hydrocarbon yields are illustrated in Fig. 10. Ni/HY reusability in TOFA HDO demonstrated a smaller activity decrease than for FAME HDO over a reused Ni catalyst, in line with catalyst characterization data. The average pore size remained fairly constant (6.5 nm) in TOFA HDO, whereas it decreased to 4.9 in the case of FAME HDO. The change in total pore volume is more noticeable in FAME HDO, where it decreased from 0.85 nm in the fresh catalyst to 0.66 nm in spent TOFA HDO catalyst to 0.51 nm in FAME HDO spent catalyst. An inferior catalyst activity decrease in TOFA HDO is related to the difference in the chemical structure of TOFA and FAME. Although regeneration seems to restore the catalytic activity as well as some of the physical properties, such as average pore size and microporous pore volume, the relative decrease in surface area remained 18%. The fresh, reused, and regenerated catalysts successfully converted fatty acids into hydrocarbons; however, the C18 to C17 ratio varied in the three runs (Fig. 10b), being 1.7, 0.4, and

Fig. 8. (a) Catalytic decarboxylation of FAME over Pd/C at 30 °C and 15 bar argon. (b) Selectivity to hydrocarbons (HCs), acids, and saturated FAME.

– – – – – –

4.9

– 3 3 0.3 19 0 45

4.6 – 3.4 – – 3.8 –

4.6 1

2.5 – 3.5 – – 3 –

4.9 – 4.8 – – 4.8 4.9 – 4.5 – – 4.1 – 2 3 2 2 1

4 – 5 – – 3.2 3.6,5.4 –

Metal/spent material

Coke (mg)

0.63 0.35 0.36 0.45 0.31 0.33 0.18 4.3–8.6 4.2–7.8 4.2–8.3 4.2–7.8 4.2–8.2 4.2–8.2 4.3–8.7 6.5 4.4 4.9 4.5 6.8 4.5 4.6

0.1 0.03 0.1 0.05 0.09 0.1 0.09

0.85 0.55 0.41 0.66 0.45 0.5 0.66 0.41 4.3–8.5 4.1–7.8 4.2–8.4 4.2–8.7 4.1–8 4.1–8.1 5.6–8.1 4.2–8 6.6 4.4 4.9 6.5 6.3 4.6 6.8 6.2

0.22 0.13 0.15 0.12 0.19 0.15 0.22 0.13

Pore size range (nm) Pore average size (nm)

1.1, respectively, for fresh, reused, and regenerated catalysts. These variations are caused by the changes in the catalyst morphology and physical properties. Selectivities of reused and regenerated catalysts to C17 and C18 are shown in Fig. 11. In contrast to the total selectivity to the desired products, octadecane and heptadecane, their respective selectivities displayed different behavior in the three experiments. An inversion of selectivities was noticed after the catalyst was reused, while initially the catalyst showed a higher selectivity to C18. The switch in selectivity should be tentatively ascribed to the influence of coking, which retards the hydrogenation pathway at the expense of decarboxylation/decarbonylation.

– Stearic acid HDO FAME HDO TOFA HDO Animal fat HDO Reused in FAME HDO FAME decarboxylation 5 wt.% Pd/C

– 26 22 18 15 45 51

– Stearic acid HDO FAME HDO TOFA HDO Animal fat HDO Reused in FAME HDO Regenerated 50 h FAME HDO 5 wt.% Ni/H-Y-80

– 14 39 35 36 33 18 44

3.8. XRD analysis

Reaction

Relative decrease in surface area (%)

217

Fig. 10. (a) Fatty acid conversion and (b) octadecane and heptadecane yield from fatty acid HDO at 300 °C and 30 bar in a semi-batch reactor over fresh, recycled, and regenerated 5 wt.% Ni/H-Y-80.

Catalyst

Table 3 Physical properties of fresh and spent catalysts and coke formation on spent catalysts.

Microporous pore volume (cm3/g)

Total pore volume (cm3/g)

TEM based particle average diameter (nm)

Metal fraction from ICP (wt.%)

Metal/(support)

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

The X-ray diffraction patterns of the fresh and spent samples of Ni/H-Y are shown in Fig. 12. The characteristic diffraction peaks in the parent H-Y are in line with the powder diffraction pattern of Y (FAU), indicating that all the observed peaks at positions lower than 35° are assigned to zeolites. XRD revealed the presence of NiO and Ni at 2h = 37.1°, 43.1°, 63.7°, and 65.1°, associated with Ni particles [57,58]. Changes in the crystallite sizes of the catalysts are related to changes in the peak intensity and width. Based on Fig. 12, it can be stated that the catalyst structure did not change significantly during the regeneration process, as also reported in [58]. The decrease in the intensity and disappearance of the peaks observed at 2h = 63.7° are due to a decrease in the particle size below the detection limit, confirmed by TEM. A faint peak was noticed at this position in the regenerated sample. Catalyst morphology, along with the size distribution of fresh, spent, reused, and regenerated Ni/HY catalysts, is presented in Fig. 13. Different treatments resulted in changes in the crystallite size. While the fresh catalyst exhibited bimodal distribution with

218

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

Fig. 11. Selectivity to 0ctadecane and heptadecane in HDO over (a) fresh, (b) reused, and (c) regenerated 5 wt.% Ni/H-Y-80.

Fig. 12. XRD patterns for the parent H-Y and metal-containing zeolites.

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

219

Fig. 13. Morphology and size distribution from SEM for (a) fresh, (b) spent, (c) recycled, and (d) regenerated Ni/H-Y-80.

two mean diameters of 0.17 and 0.37 mm, HDO over Ni/HY shifted the distribution to an average diameter of 0.34 mm. Narrowing of the distribution was seen for the recycled catalyst, further increasing the average diameter to 0.4 mm after catalyst regeneration. Particle size distribution exhibited changes for the spent catalysts (Fig. 14) compared with the fresh ones; namely, the spent catalysts showed an increase in the number of particles with smaller size. The reused catalysts demonstrated shifting of the mean particle diameter toward smaller diameters, while the regenerated ones displayed a bimodal dispersion, with the first maximum at 3.6 nm being clearly smaller than the average particle size in the fresh catalyst. The second maximum at 5.4 nm implies an increase in the particle size. The observed variation suggests that various mechanisms occurred during the reaction, leading to metal redispersion. Formation of small particles implies splitting of particles, which can be followed by a subsequent migration of atoms, as reported by Ruckenstein and Sushumna [59]. Metal particle redispersion is governed by two mechanisms, as described by Bartholomew and Farrauto [60]. The entire crystallite can migrate over the support surface, or metal atoms or molecule clusters can be detached from the crystallites. Atomic or small cluster migration is favored at low temperatures, since higher diffusivity would facilitate their migration, whereas thermal energy necessary to induce motion of larger crystallites would be available at higher temperatures (>500 °C) [61]. In addition, repeated use caused changes in the shape of the particles [59], as visible for nickel clusters, which tend to have a more spherical shape after several use.

3.9. Effect of HDO of different feedstocks on the catalyst Specific surface areas of fresh and spent Ni/H-Y catalysts are shown in Fig. 15. After the first run, the catalyst exhibited the greatest decrease in its specific surface area, 39%. After reuse, a moderate increase of 6% is partially due to a moderate increase in the pore size and the microporous pore volume, as illustrated in Table 3. The regenerated catalyst recovered a fraction of the lost specific surface area reaching 82% compared to the fresh sample area. The samples were analyzed to investigate metal leaching. The metal shows no decrease after the second run (Table 3). The spent catalysts were analyzed with thermogravimetric analysis in air in order to determine the effect of the feed on both catalytic performance and the formation of coke. TGA results of the spent catalysts are presented in Table 3. In the first run, 3% coke was formed on Ni/H-Y. The surface area decrease can be a result of micropore blocking, since the greatest surface area decrease was noticed for this sample. Coke is mainly formed inside the pores, since most of the reactions by zeolites occur inside the cavities and in the channel intersections where the acid sites are located. The sample of the reused catalyst contained 1% coke, in line with partial specific surface area recovery.

4. Conclusions Hydrodeoxygenation (HDO) of different feedstocks containing fatty acids and their derivatives was intensively investigated over several sulfur-free nickel-based catalysts, as well as

220

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

Fig. 14. TEM images and the particle size distribution for (a) spent and (b) recycled Pd/C in FAME HDO and for (c) spent, (d) recycled, and (e) regenerated Ni/H-Y-80 in FAME HDO.

Fig. 15. Specific surface area of fresh and spent Ni/H-Y-80 and Pd/C catalysts.

carbon-supported palladium. First, stearic acid was used as a model substrate to study activity and selectivity of several nickel-supported catalysts, showing that nickel supported on Y zeolite with a silica-to-alumina ratio of 80 (H-Y-80), provided the highest reaction rate, with the reactant being totally converted in a short time compared to other studied catalysts including Pd/C. HDO of FAME, TOFA, and triglycerides over Ni supported on HY-80 catalyst yields aliphatic hydrocarbons with high selectivity. HDO follows two main routes with formed hydrocarbons, either preserving the carbon number of the initial reactant or losing one carbon through decarboxylation/decarbonylation. The normalized reaction rate in the case of nickel catalysts was fivefold higher than for Pd. The final products in the latter case were mainly from the decarboxylation/decarbonylation route.

I. Hachemi et al. / Journal of Catalysis 347 (2017) 205–221

Reusability of Ni and Pd catalysts demonstrated non-negligible catalyst deactivation, associated with a decrease in the specific surface area and pore volume. TEM analysis showed a small increase in the metal particle size for both Ni and Pd catalysts. Analysis of the spent catalyst indicates that only 2% of coke was formed. To recover an apparent loss in Ni catalyst activity, regeneration was performed, successfully recovering catalyst activity. Catalyst regeneration led to metal redispersion, with formation of smaller particles resulting as a consequence of a shift of selectivity. Acknowledgments This work is part of the activities at Johan Gadolin Process Chemistry Centre (PCC) appointed by Åbo Akademi University. This work was supported by Academy of Finland. The authors acknowledge PCC colleagues Dr. Anton Tokarev and Dr. Pasi Virtanen (ÅAU) for their valuable assistance and Sten Linholm for performing ICPOES analysis. References [1] M.A. Sharara, E.C. Clausen, D.J. Carrier, An overview of biorefinery technology, in: C. Bergeron, D.J. Carrier, S. Ramawamy (Eds.), Biorefinery Co-Products: Phytochemicals, Primary Metabolites and Value-Added Biomass Processing, Wiley, 2012, pp. 1–18. [2] M. Balat, Energy Convers. Manage. 52 (2011) 1479–1492. [3] X. Fan, R. Burton, Open Fuels Energy Sci. J. 2 (2009) 100–109. [4] D.Y.C. Leung, X. Wu, M.K.H. Leung, Appl. Energy 87 (2010) 1083–1095. [5] M. Joyce, Biofuels Production Drives Growth in Overall Biomass Energy Use Over Past Decade - Today in Energy (accessed 7.8.2016). [6] I.B. Bankovic´-Ilic´, I.J. Stojkovic´, O.S. Stamenkovic´, V.B. Veljkovic, Y.T. Hung, Renew. Sustain. Energy Rev. 32 (2014) 238–254. [7] K.S. Bech, Ö. Nikolajsen, G. Trond, A.-B. Bjerre, Danish Technological Institute, Resultatkontrakt MP1.3, 2013, (accessed 9.06.2016). [8] N. Norsker, M.J. Barbosa, M.H. Vermue, R.H. Wijffels, Biotechnol. Adv. 29 (2011) 24–27. [9] R.H. Wijffels, M.J. Barbosa, M.H.M. Eppink, Biofuels Bioprod. Biorefin. 4 (2010) 287–295. [10] B. Demmig-Adams, W.W. Adams, Science 298 (2002) 2149–2153. [11] T.M. Mata, A.A. Martins, N.S. Caetano, Renew. Sustain. Energy Rev. 14 (2010) 217–232. [12] G.A. Dunstan, J.K. Volkman, S.W. Jeffrey, S.M. Barrett, J. Exp. Mar. Biol. Ecol. 161 (1992) 115–134. [13] D. De Guzman, Monitoring the Development of Green within the Chemical Industry . , (accessed 7.6.2016). [14] K. Mokkila, In World Biorefinery Conference, Jönköping, Sweden, 2012, http:// www.spci.se/shared/files/A_view_from_UPM-Kymmene._Kosti_Mokkila.pdf (accessed 8.06.2015). [15] W. Zhiyou, M.B. Johnson, Virginia Cooperative Extension (2009) 442–886. [16] P. Adewale, M.-J. Dumont, M. Ngadi, Renew. Sust. Energy Rev. 45 (2015) 574– 588. [17] I.V. Deliy, E.N. Vlasova, A.L. Nuzhdin, G.A. Bukhtiyarova, Recent Res. Eng. Automat. Control (2011) 24–29. [18] O.I. Sßenol, E.M. Ryymin, T.R. Viljava, A.O.I. Krause, J. Mol. Catal. A Chem. 268 (2007) 1–8. [19] T.R. Viljava, R.S. Komulainen, A.O.I. Krause, Catal. Today 60 (2000) 83–92. [20] B. Veriansyah, J. Young, S. Ki, S. Hong, Y. Jun, J. Sung, Y. Shu, S. Oh, J. Kim, Fuel 94 (2012) 578–585. [21] M. Krár, S. Kovács, D. Kalló, J. Hancsók, Bioresour. Technol. 101 (2010) 9287– 9293. [22] J. García-Dávila, E. Ocaranza-Sánchez, M. Rojas-López, J.A. Muñoz-Arroyo, J. Ramírez, Fuel 135 (2014) 380–396.

221

[23] M. Snåre, I. Kubicˇková, P. Mäki-Arvela, D. Chichova, K. Eränen, D.Yu. Murzin, Fuel 87 (2008) 933–945. [24] L. Yue, C. He, X. Zhang, P. Li, Z. Wang, H. Wang, Z. Hao, J. Hazard. Mater. 244– 245 (2013) 613–620. [25] B. Rozmysłowicz, P. Mäki-Arvela, A. Tokarev, A.R. Leino, K. Eränen, D.Yu. Murzin, Ind. Eng. Chem. Res. 51 (2012) 8922–8927. [26] S. Lestari, P. Mäki-Arvela, H. Bernas, O. Simakova, R. Sjöholm, J. Beltramini, G.Q. Max Lu, J. Myllyoja, I. Simakova, D.Yu. Murzin, Energy Fuels 23 (2009) 3842– 3845. [27] B. Peng, C. Zhao, S. Kasakov, S. Foraita, J.A. Lercher, Chem. Eur. J. 19 (2013) 4732–4741. [28] R. French, S. Czernik, Fuel Process. Technol. 91 (2010) 25–32. [29] J. Aguado, D.P. Serrano, J.M. Escola, L. Briones, Fuel 109 (2013) 679–686. [30] S. Thangalazhy-Gopakumar, S. Adhikari, S.A. Chattanathan, R.B. Gupta, Bioresour. Technol. 118 (2012) 150–157. [31] B. Rozmysłowicz, P. Mäki-Arvela, S. Lestari, O.A. Simakova, K. Eränen, I.L. Simakova, D.Yu. Murzin, T.O. Salmi, Top. Catal. 53 (2010) 1274–1277. [32] P. Mäki-Arvela, B. Rozmysłowicz, S. Lestari, O. Simakova, K. Eränen, T. Salmi, D. Yu. Murzin, Energy Fuels 25 (2011) 2815–2825. [33] A.T. Madsen, B. Rozmyszłowicz, I.L. Simakova, T. Kilpiö, A.R. Leino, K. Kordás, K. Eränen, P. Mäki-Arvela, D.Yu. Murzin, Ind. Eng. Chem. Res. 50 (2011) 11049– 11058. [34] E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373–380. [35] C.V. Viêgas, I. Hachemi, S.P. Freitas, P. Mäki-Arvela, A. Aho, J. Hemming, A. Smeds, I. Heinmaa, F.B. Fontes, D.C. da Silva Pereira, N. Kumar, D.A.G. Aranda, D.Yu. Murzin, Fuel 155 (2015) 144–154. [36] V. Kuuluvainen, P. Mäki-Arvela, K. Eränen, A. Holappa, J. Hemming, H. Österholm, B. Toukoniitty, D.Yu. Murzin, Chem. Eng. Technol. 38 (2015) 769–776. [37] N. Hanif, M.W. Anderson, V. Alfredsson, O. Terasaki, Phys. Chem. Chem. Phys. 2 (2000) 3349–3357. [38] M.A. Camblor, A. Mifsud, D. Valencia, Zeolites 11 (1991) 792–797. [39] H. Zuo, Q. Liu, T. Wang, L. Ma, Q. Zhang, Q. Zhang, Energy Fuels 26 (2012) 3747–3755. [40] I. Hachemi, K. Jeništová, P. Mäki-Arvela, N. Kumar, K. Eränen, J. Hemming, D. Yu. Murzin, Catal. Sci. Technol. 6 (2016) 1476–1487. [41] N. Kumar, E. Leino, P. Mäki-Arvela, A. Aho, M. Käldström, M. Tuominen, P. Laukkanen, K. Eränen, J.P. Mikkola, T. Salmi, D.Yu. Murzin, Micropor. Mesopor. Mater. 152 (2012) 71–77. [42] P.B. Weisz, C.D. Prater, Adv. Catal. 6 (1954) 144–195. [43] C.R. Wilke, P. Chang, AIChE J. 1 (1955) 264–270. [44] M. Snåre, I. Kubickova, P. Mäki-Arvela, K. Eränen, D.Yu. Murzin, Ind. Eng. Chem. Res. 45 (2006) 5708–5715. [45] W. Song, C. Zhao, J.A. Lercher, Chem. Eur. J. 19 (2013) 9833–9842. [46] H. Shi, J. Chen, Y. Yang, S. Tian, Fuel Process. Technol. 118 (2014) 161–170. [47] W.T. Pimbley, J. Phys. Chem. 68 (1965) 324. [48] E. Santillan-Jimenez, T. Morgan, J. Lacny, S. Mohapatra, M. Crocker, Fuel 103 (2013) 1010–1017. [49] T.K. Hari, Z. Yaakob, React. Kinet. Mech. Catal. 116 (2015) 131–145. [50] W. Song, Y. Liu, E. Baráth, C. Zhao, J.A. Lercher, Green Chem. 17 (2015) 1206– 1218. [51] Y. Yang, C. Ochoa-Hernández, V.A. De La Peña O’Shea, J.M. Coronado, D.P. Serrano, ACS Catal. 2 (2012) 592–598. [52] S. Lestari, P. Mäki-Arvela, I. Simakova, J. Beltramini, G.Q.M. Lu, D.Yu. Murzin, Catal. Lett. 130 (2009) 48–51. [53] A.A. Stepacheva, L.Z. Nikoshvili, E.M. Sulman, V.G. Matveera, Green Process. Synth. 3 (2014) 441–446. [54] I. Simakova, Catalytic Transformations of Fatty Acid Derivatives, Doctoral thesis, Åbo Akademi, Turku, Finland, 2010. [55] F.A. Suheil, Hydrocracking Process Utilizing a Catalyst Having a Reduced Zeolite Content US Patent 4857169 A, 1989. [56] L.A. Pine, G. Springs, US Patent 4148750, Redispersion of Noble Metals on Supported Catalysts, 1979. [57] V. Chandra, S. Palla, D. Shee, S.K. Maity, R. Soc, Chem. Adv. 4 (2014) 41612– 41621. [58] S.M. Hashemnejad, M. Parvari, Chin. J. Catal. 32 (2011) 273–279. [59] E. Ruckenstein, I. Sushumna, Hydrogen effects in catalysis: fundamentals and practical applications, in: Z. Paál, P.G. Menon (Eds.), Marcel Dekker, Inc., USA, 1988. [60] C.H. Bartholomew, R.J. Farrauto, Fundamentals of Industrial Catalytic Processes, second ed., Wiley-Interscience, Hoboken, USA, 2006. [61] M.D. Argyle, C.H. Bartholomew, Catalysts (2015) 145–269.