Top Catal (2013) 56:714–724 DOI 10.1007/s11244-013-0030-5
ORIGINAL PAPER
Deactivation in Continuous Deoxygenation of C18-Fatty Feedstock over Pd/Sibunit Anders Theilgaard Madsen • Bartosz Rozmysłowicz • Pa¨ivi Ma¨ki-Arvela • Irina L. Simakova • Kari Era¨nen • Dmitry Yu. Murzin • Rasmus Fehrmann
Published online: 18 April 2013 Springer Science+Business Media New York 2013
Abstract Catalytic continuous deoxygenation of stearic acid, ethyl stearate and tristearin without any solvents was investigated using Pd/Sibunit as a catalyst in a trickle bed reactor at 300 C. The main emphasis was to investigate the effect of gas atmosphere and catalyst deactivation. In addition to liquid-phase analysis made offline by GC, also online gas-phase analysis with IR were performed. The main liquid-phase product coming from all reactants was n-heptadecane. In addition to deoxygenation, which was observed for all substrates, also C18 and C16 alkanes were formed from tristearin. The relative ratios between stearic acid, ethyl stearate and tristearin conversions to alkanes after 3 days time-on-stream were 2.8/2.3/1.0, respectively using 5 % H2/Ar as a gas atmosphere, whereas rapid catalyst deactivation occurred with all substrates under H2-lacking atmosphere. The spent catalyst’s specific surface area profile along the downward reactor was maximum in the middle of the catalyst beds with the highest pore shrinking in the beginning and at the end of the reactor catalyst segments in the case of stearic acid and tristearin deoxygenation whereas that decreased consecutively as ethyl stearate passed through the reactor.
A. T. Madsen R. Fehrmann Department of Chemistry, Technical University of Denmark, Kemitorvet, Bygning 207, 2800 Kgs. Lyngby, Denmark A. T. Madsen B. Rozmysłowicz P. Ma¨ki-Arvela K. Era¨nen D. Yu. Murzin (&) ˚ bo Akademi University, Process Chemistry Centre, A Biskopsgatan 8, 20500 Turku, Finland e-mail:
[email protected] I. L. Simakova Boreskov Institute of Catalysis, Novosibirsk, Russia
123
Keywords Renewable diesel Deoxygenation Decarboxylation Fatty acids Fats Pd catalyst
1 Introduction Biofuels can supply a part of the oil products needed in the transition to a green and sustainable transport economy. For instance, so-called ‘‘renewable diesel’’ can be produced from fats and oils by removal of oxygen functionalities. This can be done for instance by hydrotreating with Ni based catalysts [1–4] or by noble-metal-catalysed deoxygenation of fatty feedstock at elevated temperatures [5–9]. Optimally this process yields premium quality bioderived diesel in the form of long-chain alkanes with a low consumption of hydrogen and very few by-products. In reality, however, the products formed in both gas and liquid-phase reflect the complexity of reactions taking place which should be properly understood to reveal the origin of catalyst deactivation. Sna˚re et al. [5] found Pd and Pt to be the most active catalysts as well as the most selective towards decarboxylation of stearic acid to n-heptadecane, yielding gas mixtures of CO and CO2 in varying ratios. In connected works Pd/C was used as a catalyst to convert tristearin, ethyl stearate, and stearic acid to alkanes [6, 7], as well as other fatty acids [8, 9]. Conversion of the carboxylic acid itself took place almost exclusively via decarboxylation, but ester functionalities underwent more complicated mechanisms in the conversion to n-alkanes [6, 7]. Do et al. [10] reported formation primarily of CO over Pt/Al2O3 during deoxygenation of methyl esters, observing suppression of heavier condensation products and almost no methane formation in H2 atmosphere. Immer et al. [11] claimed that solely CO2 was formed from decarboxylation
Top Catal (2013) 56:714–724
of stearic acid and suspected that H2 was in fact inhibiting the reaction. Berenblyum et al. [12] recently proposed that the deoxygenation could take place via a formic acid intermediate on the adsorbed surface of Pd. The formic acid would decompose once formed over Pd catalysts at 300–350 C [13], yielding either CO2 ? H2 via dehydrogenation or CO ? H2O via dehydration. Naturally occurring oils and fats represent quite chemically varied fatty acid profiles, which change between plants, animals, and it varies considerably within a species. Normally approximate fatty acid distributions are known [14, 15]. Simakova et al. [8] showed, however, that the reaction rate was independent of the chain length of saturated C17–C22 fatty acids during deoxygenation of dilute dissolved fatty acids over Pd/Sibunit at 300 C. For shorter carbon chains, however, Ford et al. [16] found that the deoxygenation rate decreases with decreasing chain length. Palladium and platinum as catalysts have been investigated for deoxygenation of fatty feedstock on a number of nanoporous and microporous supports, for instance reactions of stearic acid over Pd on SBA-15 [1] or Pd on a mesocellular SiO2-foam [2], transformations of sunflower oil were studied over Pd/SAPO-31 [3] or conversion of saturated vegetable oil over Pt/H-ZSM-22/Al2O3 [4]. These studies indicated that acidic functionalities of the support lead to modest isomerisation of the formed n-alkanes, but cracking may occur at too high temperatures. At 350 C conversion of triglycerides of varying saturation over carbon-supported Ni, Pd and Pt without added H2 was studied by Crocker and co-authors. Both CO2, CO, CH4 and short-chain hydrocarbons were observed in the gas-phase while free fatty acids as intermediates and a range of alkanes and heavier paraffins as products were obtained in the liquid-phase [17]. In semibatch reactionmode at 300 C Rozmysłowicz et al. studied conversion of C18 (unsaturated) tall oil fatty acids over Pd on mesoporous carbon Sibunit via saturation-desaturation reactions and deoxygenation. Considerable differences in the heptadecane yield were found as a function of H2 pressure, and extensive amounts of coke and catalyst deactivation resulted from H2-sparse conditions, especially with more concentrated solutions of the unsaturated fatty acids [18]. Deactivation is suspected to be connected with the presence of unsaturated fatty acids, which undergo cyclisation, dehydrogenations and Diels–Alder reactions [9, 15, 18]. Figures 1, 2, 3, 4 demonstrate reactions occuring during deoxygenation of various C18 fatty feedstock.
Fig. 1 Water–gas shift reaction and methanation of COx
715
Fig. 2 Decarboxylation and decarbonylation of stearic acid
Deoxygenation of unsaturated fatty acids and their methyl esters was investigated by Sna˚re et al. [19] and confirmed the rapid isomerisation and hydrogen-transfer reactions taking place on the Pd/C active surface at 300 C. Both Immer et al. [11] and Rozmysłowicz et al. [18] confirmed the hydrogenation transfer between feedstock and solvent and also found that saturation was necessary for obtaining high reaction rates. Fatty acids, fatty acid alkyl esters and triglycerides were all applied as feedstock for the noble-metal catalysed deoxygenation [6, 9, 10, 19]. It is therefore highly interesting to relate these substrates in terms of deoxygenation reactivity and reaction pathways. Most of the studies in literature have focused on batch or semi-batch reactors, i.e. autoclaves, where the semibatch-type system allows for renewal of the gas atmosphere while maintaining the liquid-phase in the reactor. Industrially, however, concentrated reactant streams and continuous reactors are used. It is thus important as a part of scaling up to utilise fixed bed reactors using longer reaction times-on-stream to assess catalyst properties. The aim of the present study is thus to compare the deoxygenation activity and the liquid/gas product distribution in both H2-containing and inert gas flow as well as to monitor concurrently the catalyst deactivation profile through the reactor for various functional groups in the C18 fatty feedstock. Stearic acid, ethyl stearate and tristearin (glyceryltristearate) have been deoxygenated over a 2 wt% palladium supported on beads of a synthetic mesoporous carbon-type ‘‘Sibunit’’ [20] in a continuous fixed bed reactor under comparable conditions. Both liquid and gas samples have been taken out and analysed by off-line analysis for reactants and products using gas chromatography, while the gas content of COx has been determined by online IR analysis.
2 Experimental 2.1 Synthesis of Pd on Sibunit Carbon Catalyst The 2 wt% palladium on Sibunit (Pd/C) catalyst was prepared by the following method [21]: Sibunit beads of about 1.6 mm in diameter were treated with 5 wt% HNO3 overnight at 25 C, washed by distilled water and dried for several hours at 110 C. An aqueous solution of H2PdCl4 ([99.9 %) was regulated to pH 8–9 with Na2CO3, and the dried Sibunit beads were added to this solution for
123
716
Top Catal (2013) 56:714–724
Fig. 3 Decarboxylation and decarbonylation of ethyl stearate
Fig. 4 Decarboxylation, decarbonylation and hydrodeoxygenation of tristearin
vessel and solvent vessel, as well as gas flow controllers and gas lines for Ar and H2. The liquid vessels were continuously flushed by bubbling N2 at 20 ml/min through the liquid and out of the system to evacuate dissolved oxygen in the feed. On the outlet side of the reactor the setup was equipped with heated lines, a liquid sampling valve and a collector for the remaining liquid effluent, as well as gas flow controllers for regulating pressure and blending the effluent with the carrier gas before online analysis of CO and CO2. During experimentation, the tubes, fittings, the pump and the vessel exposed to the liquid fatty feed were heated to 100 C as the saturated fatty compounds used as reactants have melting points between 45 and 90 C. The configuration of the continuous downward reactor tube is shown in Fig. 5. The reactor itself was a tube of 18 cm height and 1.58 cm inner diameter for a total volume of 35.1 ml. Before each run, the reactor was loaded in three steps: •
• Fig. 5 A schematic picture of the trickle bed reactor loaded with catalyst in five separated segments
•
In the bottom a layer of quartz wool, then 6 ml of quartz sand 0.2–0.8 mm in diameter, and then a layer of quartz wool; In the middle the catalyst bed of totally 10 g of 2 wt% Pd supported on Sibunit, divided into five equally big layers of 2 g each, separated by a layer of quartz wool (for easier unloading the catalysts after the experimentation); In the top a layer of quartz wool, then 4 ml of quartz sand with the diameter of 0.2–0.8 mm, and then a layer of quartz wool.
deposition of the resulted Pd(II) polyhydroxy-complexes. The deposition was continued for 6 h at 25 C; then, the beads were separated and washed with distilled water until no chloride ions could be detected in the filtrate. Then the catalyst was dried at 110 C overnight and cooled until RT.
A schematic picture of the reactor system is shown in Fig. 6. A thermocouple was inserted through the reactor outlet to measure temperature from the interior of the reactor.
2.2 Reactor Configuration and Loading
2.3 Catalyst Activation and Catalytic Test
On the inlet side of the reactor the setup consisted of a specially designed heated liquid piston pump, heated feed
After flushing with Ar ([99.99 %, AGA), the catalysts were reduced prior to reaction in 5 % H2/Ar ([99.99 % for
123
Top Catal (2013) 56:714–724
717
Fig. 6 A schematic picture of the reactor system with control equipment
H2, AGA) as follows: ca. 10 C/min to 150 C (1 h)— 10 C/min h–300 C (0.5 h) before the flow of liquid feed was started. The liquid feed consisted of either stearic acid ([99 %), ethyl stearate ([99 %), or technical tristearin (ca. 65 % purity, the remaining 35 % of the fatty acids were palmitic acid) in each experiment. The substrates were pumped at 0.075 ml/min for 150 h each, corresponding to 0.22, 0.20, and 0.21 mmol/min on a fatty acid basis of respectively stearic acid, ethyl stearate and technical tristearin. The gas flow was supplied at 20 bar, first using 5 % H2/Ar at 42 ml/ min and after 75 h (stearic acid: 96 h) time-on-stream (TOS) the gas flow was switched to pure Ar and the reaction was continued for another 75 h (stearic acid: 54 h). 2.4 Liquid-Phase Analysis Liquid samples were taken out from the outlet of the reactor by two valves. Two minutes prior to sampling, the content of the sampling nozzle was purged. Samples were collected in glass vials and weighed, followed by addition of eicosane ([99 %, Sigma-Aldrich) as external standard. This was diluted in 4 ml of dodecane ([99 %, Fluka) and heated in an oven at 70 C until the dodecane had dissolved the sample. Then, 10 ll of the dodecane-diluted sample was added to a GC vial, along with 100 ll pyridine ([99 %, Fluka), 100 ll BSTFA (N,O-bis(trimethylsilyl)trifluoroacetamide: silylation reagent, [99 % purity, Acros Organics) and 40 ll 0.013 M myristic ([99 %, SigmaAldrich) acid in dodecane for control of the silylation. The vial was capped and allowed to silylate in an oven at 70 C
for at least 30 min before being analysed by a gas chromatograph (GC). The liquid samples taken out during the deoxygenation of ethyl stearate and stearic acid were analysed on a 5890 N HP Agilent split/splitless-injection (S/SL) GC with a flame-ionisation detector (FID). The liquid samples taken out during technical tristearin deoxygenation were analysed on a 7890A Network HP Agilent Cool-on-column-injection high-temperature GC equipped with FID. The analysis of the triglycerides can be somewhat troublesome, and in this case the GC areas obtained from the triglyceride analysis were too low to make the mass balance check to unity. The products pentadecane (C15) and heptadecane (C17) can be expected from decarbonylation and decarboxylation while hexadecane (C16) and octadecane (C18) are expected from complete reduction with H2, thereafter these products were used to assess the extent of reaction in the liquid-phase. Reference standard solutions with known concentrations of each relevant analyte and internal standard were prepared to calculate response factors relative to the internal standard for each compound. Conversion and yields from the catalytic test samples were then calculated for each compound from the obtained GC areas relative to the internal standard added to the samples, divided by the response factor from the reference standards. 2.5 Gas-Phase Analysis The gas-phase content of CO and CO2 in all experiments was analysed on-line with a Siemens Ultramat 6 IR-analyser.
123
718
Top Catal (2013) 56:714–724
Fig. 7 Molar fractions of the main compounds in the deoxygenation of stearic acid. Symbols: (green filled triangle) n-heptadecane, (red filled square) 1-heptadecene, (x) C17heavy/undecylbenzene, (orange filled circles) stearic acid, (red line) CO conversion online and (blue line) CO2 conversion online
Fig. 8 Molar flows of gasphase products during deoxygenation of stearic acid. Symbols: (x) H2, (circle) methane, (filled diamond) ethene and (spade) propane
The gas was diluted in Ar before measurements with the Ultramat. The CO/CO2 online analysis was switched off during part of the experimentation due to calibration and servicing. Gas samples for off-line analysis were taken out by bubbling the gas stream through an aqueous solution of 0.01 M HCl and 0.1 M KCl into 100 ml glass flasks, which were capped after filling. The gas samples were analysed by injection with a gas syringe on a 5890 N HP Agilent S/SL GC equipped with both thermoconductivity detector (TCD) and FID.
tested. The fresh and spent catalysts from the deoxygenation were analysed with a Micromeritics ASAP 2020 physisorption apparatus using liquid N2 at 77 K. In both cases, specific surface areas were calculated using the Brunauer– Emmett–Teller (BET) equation from the N2 adsorption– desorption isotherms. The pore size distributions were obtained from the Dollimore–Heal correlation. The Pd layer was found to be 64 lm thick confirmed by ICP-MS in the outer layer of the egg-shell catalyst bead as reported in [22].
2.6 Catalyst Characterisation
3 Results and Discussion
After the reaction, the catalyst bed was taken out of the reactor sequentially and sorted into five batches depending on their position in the catalyst bed. Specific surface area and pore size distributions were obtained for all the catalysts
3.1 Deoxygenation of Stearic Acid
123
Pumping of stearic acid feed to the reactor was started TOS = 0 min. While CO and CO2-levels at 5 and 30 %
Top Catal (2013) 56:714–724
719
Fig. 9 Molar fractions of the main compounds in the deoxygenation of ethyl stearate. Symbols: (red filled square) 1-heptadecene, (green filled triangle) heptadecane, (purple x) C17-heavy, (orange filled circles) stearic acid, (black filled circles) ethyl stearate, (red line) CO conversion and (blue line) and CO2 conversion
Fig. 10 Molar flows of gasphase products during deoxygenation of ethyl stearate. Symbols: (circle) methane, (filled diamond) ethene, (open diamond) ethane and (x) H2
increased from 60 min after start, the liquid samples only responded to the reactant flow after 180 min as visualised in Fig. 7—there is a delay in sampling time with this reactant system, which is described in [23]. Initially almost 100 % heptadecane was formed and only traces of stearic acid were observed. However, the yield of heptadecane had decreased to 80 % at TOS = 1,560 min and to 72 % at TOS = 5,760 min, thus showing a slow and gradual deactivation during the deoxygenation in 5 % H2/Ar. Correspondingly, the amount of unconverted stearic acid of 17–23 % were observed between TOS of 1,560 and 5,760 min. The online analyser was off-line during the period of TOS = 360–2,280 min, but during the remaining experimentation period the same tendency of slight deactivation was observed from the COx yield. At TOS = 5,760 min the gas atmosphere was changed to pure Ar. This led to a drop in the CO-level to 2 % conversion and a peak increase in the CO2-level in the gas as well as in an increase from 75 to 85 % of heptadecane yield
for a few hours; however, the CO2-level dropped again to a level of 0 % conversion at TOS = 6,040 min, while the CO-concentration rose to 5 % and decreased thereafter. The heptadecane concentration dropped steeply to 8 % at TOS = 6,040 min and levelled off at ca. 2 %, following the decreasing concentration of CO. The steep drop in heptadecane concentration is followed by an equally steep rise in stearic acid concentration a few hours after the pure Ar-flow was started. The maximum of CO-peak at TOS = 6,040 min is accompanied by a peak of 2 % of 1-heptadecene and 1 % of undecylbenzene (‘‘C17-heavy’’). Thus, the deactivation in hydrogen-deficient gas flow was accompanied by formation of unsaturated C17-products and aromatics. After this, 3 % decline to 2 % heptadecane concentration was obtained until TOS = 9,000 min, when the experiment was stopped. The conversion peak indicated that H2 is also partially a reaction inhibitor during decarboxylation, however, it is unclear if it is due to the formic acid mechanism suggested by Berenblyum or the inhibition reaction suggested by Immer [24].
123
720
Top Catal (2013) 56:714–724
Fig. 11 Molar fractions of the main compounds in the deoxygenation of tristearin. Symbols: (black filled diamond) C15/pentadecane, (black x) C16/ hexadecane, (green filled diamond) C17/heptadecane, (green x) C18/octadecane, (red line) CO conversion and (blue line) CO2 conversion
Fig. 12 Molar flows of gasphase products during deoxygenation of tristearin. Symbols: (circle) methane, (filled diamond) ethene, (open diamond) ethane, (filled triangle) propene, (spade) propane and (x) H2
During deoxygenation in hydrogen-containing atmosphere, the yield of liquid heptadecane was several times higher than the CO- and CO2-levels obtained from the online gas analyser indicated. It can be speculated if this is due to methanation of the formed COx species by consumption of H2. However, only trace amounts of light hydrocarbons, let alone methane, were observed from the offline GC analysis of the effluent gas-phase—thus methanation did not take place during deoxygenation of stearic acid as also observed by Immer et al. [24]. The cause for the differences in conversion and yield in the liquid-phase and COx levels in the gas-phase is still largely unclear. Interestingly, the gas atmosphere during the stearic acid experiment showed a gradual rise of CO and decline of CO2 formation in H2-containing gas flow at TOS = 140–5,760 min. Such a shift in the gas production
123
with time-of-reaction during stearic acid deoxygenation over Pd has previously been observed and suggested to be caused by a growing selectivity towards decarbonylation [24, 25]. As evident from Fig. 8, the H2-level in the gasphase samples obtained showed a decrease over the same time interval. This is therefore an evidence that the Pd catalyst was active for the water–gas shift reaction during deoxygenation conditions as no other H2-consuming reactions were observed. 3.2 Deoxygenation of Ethyl Stearate The ethyl stearate liquid pumping was started at TOS = 0 and a response in terms of the initial heptadecane yield was 90 % after 3 h, decreasing quickly to 75 % in the next few hours, as is shown in Fig. 9. In the beginning of the run
Top Catal (2013) 56:714–724
721
Fig. 13 The specific surface areas of the fresh and spent catalysts for each of the deoxygenation substrates as a function of segment no. from the top in the catalyst bed
Table 1 Overview of conversion to deoxygenation products at different times-on-stream for fatty feedstock TOS
2.5 h
6h
75 h
Peak at switch to inert gas
Steady-state in inert gas
Spent cat BET, m2/g
Gas atm.
5 % H2/Ar
5 % H2/Ar
5 % H2/Ar
Ar
Ar
(Fresh: 362)
Stearic acid (%)
100
80
72
85
2–3
193
Ethyl stearate (%)
90
75
62
62
\1
229
Techn. Tristearin (%)
94
56
22
26
\1
200
Conditions: 300 C, 20 bar gas supplied by at a flow of 42 Nml/min, 0.075 ml/min liquid reactant flow
about 2–3 % unsaturated compounds in the form of 1-heptadecene and undecylbenzene were formed from the ethyl stearate, although their content decreased to 1 % during the reaction in 5 % H2/Ar. The heptadecane formation stabilised at ca. 62 % yield from TOS = 800 min while the analysis of the ethyl stearate showed between 25 and 40 % unconverted ester—this varied considerably, as did the amount of stearic acid observed in 2–12 % yield. It can be suspected that stearic acid is formed as an intermediate in the deoxygenation of ethyl stearate [6]. Initially, the online gas analyser was shut off during the first 2,500 min TOS, thus no information on CO/CO2-formation was found for the first time interval. However, the gas analysis from TOS = 2,500 min showed a remarkably constant level of CO corresponding to 1.5 % conversion, while the CO2-concentration in the gas-phase levelled off at around 18–19 % conversion before TOS = 4,500 min. According to Fig. 10 ethane is formed almost as the sole light hydrocarbon co-product during ethyl stearate deoxygenation in 5 % H2/Ar, albeit trace amounts of methane and ethene appeared. The gas atmosphere was changed from 5 % H2/Ar to pure Ar at TOS = 4,500 min and effect started to be seen after about 20 min. A steep drop in CO2 concentration took place for about an hour to a CO2-level corresponding to practically 0 % conversion. The CO-level fell more moderately over a number of hours. One hour after the change to Ar gas flow the heptadecane yield started to drop steeply as well, after another 60 min it reached 7 % and after
120 min only 2 %. After 6 h from the gas change the yield of heptadecane was below 1 % and remained for the rest of the experimentation period. The ethyl stearate responded to the change by increasing to a level over 100 % within a few hours and during the remainder of the experiment. This is clearly not possible at steady-state, but may be due to inaccuracies with concentration response factors during GC-FID liquid analysis. The concentration of ethane and hydrogen in the gasphase diminished over a few hours after switch to pure Ar gas feed, consistent with the deactivation observed by the liquid-phase analysis. Stearic acid was observed throughout the entire experimentation between 12 and 2 % yield, possibly as an intermediate in the conversion of ethyl stearate. Despite the catalyst being completely inactive in the Ar atmosphere, about 4–5 % stearic acid were still observed in the liquid samples from the reactor. Interestingly, no ethene was observed either after switch to Ar gas feed despite of stearic acid formation at TOS = 4,800–9,000 min—an otherwise possible product of ethyl stearate decomposition. Other reaction routes may thus play a role. 3.3 Deoxygenation of Tristearin Gas-phase products appeared after about 60 min TOS after start-up of tristearin pumping followed by a rapid increase in CO and CO2 concentration. After 150 min the first liquid products appeared, corresponding to ca. 93 % total
123
722
conversion to alkanes (Fig. 11). The main liquid-phase product is n-heptadecane as in [6]. The CO2-concentration peaked corresponding to ca. 21 % yield and CO peaked at ca. 2.5 %. The CO-level was almost constant for the remaining reaction in 5 % H2/Ar, while the CO2-level dropped to 7 % conversion in the same time interval. The yields of the primary products heptadecane and pentadecane declined towards respectively 21 and 7 % when nearing TOS = 4,500 min. Thus, productivity of the catalyst bed had been reduced to ca. 30 % of the initial activity. Interestingly, the gas-phase products from offline GC analysis revealed formation of propane (from the triglyceride C3-backbone) along with minor amounts of ethane and traces of methane. Hydrogen was found in increasing amounts after TOS = 360 min until 4,500 min, in accordance with the decreasing conversion of triglycerides observed from liquid-phase analysis. It is possible that propane or some other C3-species during deoxygenation underwent C–C-scission in some form (Fig. 12). As the gas flow was changed to pure Ar at TOS = 4,500 min, the CO- and CO2-concentrations dropped to zero immediately, then increased for a short period of time followed by a gradual decline over the next hours to a level below 1 % conversion. The production of heptadecane and pentadecane showed a small increase and declined after 120 min, levelling off and nearing zero in the following hours. After the switch to inert atmosphere, the offline gasphase samples did not show any traces of unsaturated compounds, but solely the gradual disappearance of hydrogen and light alkanes. Ethyl stearate and tristearin yielded identical production levels of CO (despite different conversions of the reactants), and it may be speculated that the same gas-producing reactions took place. Interestingly, varying minor amounts of stearic acid were obtained during the entire experimentation with ethyl stearate as a possible part of the deoxygenation reaction mechanism, but during the run in technical tristearin no palmitic or stearic acid were observed. It could be suspected that part of the mechanism requires removal of H2 to the sweeping gas—such as the formic acid formation/decomposition mechanism proposed by Berenblyum et al. [12], where the formic acid must decompose to H2 and CO2. Thus, H2 is both needed as a reactant to avoid formation of and saturate unsaturated compounds and prevent at least partially deactivation, but also inhibiting part of the mechanism or leading to formation of CO, as observed by Immer et al. [11, 24].
123
Top Catal (2013) 56:714–724
Deactivation of technical tristearin resulted in about 5 % selectivity towards even-carbon-numbered alkanes (hexadecane and octadecane) during the reaction in hydrogencontaining atmosphere, despite the fact that neither stearic acid nor ethyl stearate yielded any trace of even-carbonnumbered product (octadecane). It may be speculated that the hexadecane and octadecane in the technical tristearin were derived from impurities of C19 and C17 fatty acids. They were not found, however, during fatty acid analysis by transesterification, thus octadecane and hexadecane must be formed from stearic and palmitic acid. This is somewhat contrary to the assumption that ethyl stearate and tristearin should have reacted via identical reaction routes, thus it can be concluded that the reaction network was more complex for triglycerides than for simple esters and fatty acids. No unsaturated compounds were observed during the entire treatment of tristearin, but minor amounts of 1-heptadecene and monoaromatic C17compounds were observed for both stearic acid and ethyl stearate, especially during start-up and gas atmosphere change. Ethyl stearate and tristearin yielded primarily CO2 and a constant level of CO during reaction in H2/Ar atmosphere, which is in contrast to the results reported by Resasco et al. [10] who reported about methyl stearate deoxygenation over Pt/c-Al2O3 in a semibatch reactor and primarily detected CO formation in the gas-phase during heptadecane formation in both H2 and He atmosphere at 325 C. The reason could be the differences in the metal, catalyst support or reactor. 3.4 Catalyst Characterisation and Deactivation An overview of the deoxygenation results with different substrates is given in Table 1. It can be observed that the catalyst deactivated faster during tristearin conversion than during ethyl stearate deoxygenation, which again deactivated faster than during stearic acid deoxygenation. The results follow a ‘‘the simpler, the better’’-trend regarding the feedstock during the deoxygenation—the more complicated and hindered is the molecule, the faster is the catalyst deactivation. It is also evident that the shift to pure Ar atmosphere swiftly led to deactivation in all of the liquid feeds applied. It appeared that the reactor still produced a few percent of n-heptadecane in the pure Ar gas flow with stearic acid as a reactant. This behaviour is not seen for the ester compounds. An assessment of the degree of deactivation was made by BET measurements of different segments of the spent catalyst beads from the three different feeds by comparison
Top Catal (2013) 56:714–724
with the fresh catalyst as seen in Fig. 13. Here it is evident that the catalyst beads used for ethyl stearate deoxygenation have endured the least deactivation by loss of surface area compared to the fresh catalyst (36 % surface area lost on average), tristearin somewhat more (44 % surface area lost) and stearic acid the most (46 % surface area lost). This order is different from the order of the degree of deactivation of the catalysts mentioned above, although overall the decline in surface area is somewhat similar. Coking or build-up of reactants or products is usually the mechanism behind the shrinking of the internal pore surface area associated with deactivation [23, 24, 26]. Differences in deactivation are possibly due to different accessibility of the active Pd sites for the two esters and the fatty acid, especially on the partly deactivated catalysts. The degree of leaching of the Pd nanoparticles has been assessed in previous studies and been found low or insignificant using free fatty acids as feedstock [23, 27]. No data has been obtained on leaching in this study, but it is not expected that ester compounds should deactivate the catalyst more due to leaching. Sintering of Pd particles on Sibunit had previously been found not to take place at 300 C to a significant degree even after ca. 300 h TOS [23]. The specific surface area distribution of the spent catalyst beads in Fig. 13 was also surprising and somewhat different compared to a clear deactivation profile described earlier in [23]. In the present study, the surface areas in the case of stearic acid and tristearin deoxygenation were the largest in the middle of the catalysts beds, therefore deactivation by pore clogging was more abundant in the top and the bottom of the catalyst bed. The ethyl stearate transformation resulted in the deactivation profile where the surface areas decreased downwards in the bed, contrary to the deactivation observed previously in [23]. Such unexpected behaviour might be due to small temperature differences depending on axial bed position, which have led to the observed surface area distributions of the catalyst segments for the different feedstock. 4 Conclusion The supported noble-metal catalysed deoxygenation of fatty compounds has been investigated in a continuous reactor. The reactivities of stearic acid, ethyl stearate and tristearin over 2 wt% Pd on mesoporous carbon Sibunit in continuous deoxygenation were assessed under 5 % H2/Ar. After 3 days TOS the conversion to alkanes was 75 % with stearic acid, 62 % with ethyl stearate, and 27 % with technical tristearin as the liquid feed molecules at a feed of 0.075 ml/min.
723
The catalyst deactivation was fast with all of the feed molecules, when the gas flow was switched to Ar. The stearic acid sustained the slowest deactivation compared to ethyl stearate and tristearin. Technical tristearin yielded minor amounts of C18- and C16-alkane through full reduction of the fatty acid chains, which was otherwise not observed with ethyl stearate or stearic acid. Activity for water–gas shift equilibrium was confirmed for the Pd/Sibunit catalyst during stearic acid deoxygenation, while the methanation reaction of COx did not necessarily accompany it.
References 1. Lestari S, Ma¨ki-Arvela P, Era¨nen K, Beltramini JN, Max Lu GQ, Murzin DYu (2009) Catal Lett 134:250–257 2. Ping EW, Wallace R, Pierson J, Fuller TF, Jones CW (2010) Microporous Mesoporous Mater 132:174–180 3. Kikhtyanin OV, Rubanov AE, Ayupov AB, Echevsky GV (2010) Fuel 89:3085–3092 4. Hancso´k J, Kra´r M, Magyar S, Boda L, Hollo´ A, Kallo´ D (2007) Microporous Mesoporous Mater 101:148–152 5. Sna˚re M, Kubicˇkova´ I, Ma¨ki-Arvela P, Era¨nen K, Murzin DYu (2006) Ind Eng Chem Res 45:5708–5715 6. Kubicˇkova´ I, Sna˚re M, Era¨nen K, Ma¨ki-Arvela P, Murzin DYu (2005) Catal Today 106:197–200 7. Sna˚re M, Kubicˇkova´ I, Ma¨ki-Arvela P, Era¨nen K, Wa¨rna˚ J, Murzin DYu (2007) Chem Eng J 134:29–34 8. Simakova I, Simakova O, Ma¨ki-Arvela P, Murzin DYu (2010) Catal Today 150:28–31 9. Ma¨ki-Arvela P, Holmbom B, Salmi T, Murzin DYu (2007) Catal Rev 49:197–340 10. Do PT, Chiappero M, Lobban LL, Resasco DE (2009) Catal Lett 130:9–18 11. Immer JG, Kelly MJ, Lamb HH (2010) Appl Catal A 375: 134–139 12. Berenblyum AS, Danyushevsky VY, Katsman EA, Podoplelova TA, Flid VR (2010) Pet Chem 50:305–311 13. Bulushev DA, Beloshapkin S, Ross JRH (2010) Catal Today 154:7–12 14. Knothe G, Van Gerpen J, Krahl J (2005) The biodiesel handbook, 1st edn. AOCS Press, Urbana 15. Lestari S, Ma¨ki-Arvela P, Beltramini JN, Lu GQM, Murzin DYu (2009) ChemSusChem 2:1109–1119 16. Ford JP, Immer JG, Lamb HH (2012) Top Catal 55:175–184 17. Morgan T, Grubb D, Santillan-Jimenez E, Crocker M (2010) Top Catal 53:820–829 18. Rozmysłowicz B, Ma¨ki-Arvela P, Lestari S, Simakova OA, Era¨nen K, Simakova I, Murzin DYu, Salmi T (2010) Top Catal 53:1274–1277 19. Sna˚re M, Kubicˇkova´ I, Ma¨ki-Arvela P, Chichova D, Era¨nen K, Murzin DYu (2008) Fuel 87:933–945 20. Surovikin VF, Plaxin GV, Semikolenov VA, Likholobov VA, Tiunova IJ (1990) US Patent US 49786491990 21. Simonov PA, Troitskii SY, Likholobov VA (2000) Kinet Catal 41:255–269 22. Lestari S, Ma¨ki-Arvela P, Bernas H, Simakova O, Sjo¨holm R, Beltramini J, Max Lu GQ, Myllyoja J, Simakova I, Murzin DYu (2009) Energy Fuels 23:3842–3845
123
724 23. Madsen AT, Rozmysłowicz B, Simakova IL, Kilpio¨ T, Leino AR, Korda´s K, Era¨nen K, Ma¨ki-Arvela P, Murzin DYu (2011) Ind Eng Chem Res 50:11049–11058 24. Immer JG, Lamb HH (2010) Energy Fuels 24:5291–5299 25. Wang W, Thapaliya N, Campos A, Stikeleather LF, Roberts WL (2012) Fuel 95:622–629
123
Top Catal (2013) 56:714–724 26. Ma¨ki-Arvela P, Rozmysłowicz B, Lestari S, Simakova O, Era¨nen K, Salmi T, Murzin DYu (2011) Energy Fuels 25:2815–2825 27. Bernas H, Era¨nen K, Simakova I, Leino A-R, Korda´s K, Myllyoja J, Ma¨ki-Arvela P, Salmi T, Murzin DYu (2010) Fuel 89: 2033–2039
