Bioresource Technology 126 (2012) 148–155

Contents lists available at SciVerse ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Temperature-dependent reaction pathways for the anomalous hydrocrackingof triglycerides in the presence of sulfided Co–Mo-catalyst

Mohit Anand, Anil K. Sinha ⇑CSIR-Indian Institute of Petroleum, Dehradun 248 005, India

h i g h l i g h t s

" Kinetic studies on anomaloushydrocracking of triglycerides overCo–Mo/Al2O3.

" Temperature has a major effect onthe reaction pathways fortriglyceride conversion.

" Higher activation energies forformation of light and middle thanfor heavy products.

0960-8524/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.biortech.2012.08.105

⇑ Corresponding author. Tel.: +91 135 2525842; faxE-mail address: asinha@iip.res.in (A.K. Sinha).

g r a p h i c a l a b s t r a c t

Triglyceride, CTg

Light (C5-C8), CL

Middle (C9-C14), CJ

Heavy (C15-C18), CH

k2, n2

Oligomerized (>C18), CP

k6, n6

k5, n5

Hydrocracking Co-Mo/Al2O3

360 oC, 80 bar,1500 H2/ Feed

Kinetic models for a shift in reaction pathways at different temperatures for hydrocracking of triglyceridemolecules, with high activation energies for the formation of lighter and middle distillates and lower forthe heavy and deoxygenated products, over Co–Mo/Al2O3 catalyst.

a r t i c l e i n f o

Article history:Received 12 December 2011Received in revised form 20 July 2012Accepted 23 August 2012Available online 10 September 2012

Keywords:HydrocrackingHydroprocessingTriglycerideBiofuelKinetic modeling

a b s t r a c t

Kinetic studies and product profiling was done to understand the anomalous cracking of jathropha oiltriglycerides in the presence of sulfided Co–Mo/Al2O3 catalyst. At temperatures between 320 and340 �C, only deoxygenation and oligomerization reactions took place whereas at temperatures above340 �C, internal conversions between the products and direct conversion to lighter and middle distillateswere favored High pressures (80 bar) and H2/feed ratios (>1500) were necessary to minimize oligomer-ization of the products and to increase the lifespan of the catalyst. Lumped kinetic models were validatedwith experimental results. Activation energies for the formation of lighter (83 kJ/mol) and middle frac-tions (126 kJ/mol) were higher than those for the heavy (47 kJ/mol) and deoxygenated (47 kJ/mol) prod-ucts. Jatropha oil triglycerides hydroconversion pathways were dependent on temperature and thetriglycerides could be hydrocracked to lower range hydrocarbons (C5–C14) by increasing the reactiontemperatures.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction triglyceride molecule (TG) is hydrogenated and cracked to liberate

Plant oils can be trans-esterified to produce biodiesel, or theymay be hydroprocessed to produce petroleum-like biofuels(Bezergianni et al., 2009a,b; Donnis et al., 2009; Huber et al.,2007; Kubicka et al., 2009; Kumar et al., 2010; Lappas et al.,2009; Melis et al., 2009; Šimácek et al., 2009).

The first step in plant-oil hydroprocessing is the saturation ofdouble bonds in the triglyceride molecule (TG) then the saturated

ll rights reserved.

: +91 135 2660202.

an Rx acid (RxCOOH) and a propane molecule. Rx may be either aC15 or C17 compound depending on the chain length of the triglyc-eride molecule. The Rx acid can follow three different reactionpathways for the removal of the oxygen atom to yield C15–C18compounds.

Hydrodeoxygenation gives a water molecule and a correspond-ing hydrocarbon (RxCH3). Decarboxylation yields a carbon dioxidemolecule and a C16 or C18 hydrocarbon (RxH). Decarbonylationyields a carbon monoxide molecule, a water molecule and a longchain hydrocarbon molecule (RxH) with 15- or 17-carbon chainlength. The compounds produced by oxygen removal reactions

M. Anand, A.K. Sinha / Bioresource Technology 126 (2012) 148–155 149

may crack and isomerize to yield lighter distillate products (Huberet al., 2007; Melis et al., 2009).

For hydroprocessing of various plant-derived oils, acidic sup-ports such as amorphous mixed oxides like silica–alumina (for cat-alysts such as sulfided Co–Mo, Ni–Mo, Ni–W) (Kumar et al., 2010;Liu et al.,2009; Tiwari et al., 2011; Verma et al., 2011), phosphate-modified silica-alumina (Verma et al., 2011), and crystalline acidicsupports like zeolites (Huber et al., 2007; Kumar et al., 2010;Murata et al., 2010; Verma et al., 2011) have been successfully em-ployed to produce cracked hydrocarbons ranging from gasoline tokerosene to diesel. The yield of cracked products increased withincreasing acidity of the catalysts. As the reactions are highly exo-thermic, catalyst stability and selection are important (Kumaret al., 2010; Michaelsen and Egeberg, 2009).

To realize a sustainable commercial future for biomass-oil fuelsby hydroprocessing, detailed kinetic studies, product profiling anda better understanding of the underlying mechanisms and reactionpathways are necessary.

With this aim, kinetic modeling was carried out in the presentstudy using a Co–Mo/Al2O3 catalyst. Krár et al. (2010) demonstratedthat a CoMo/Al2O3 catalyst was suitable for the conversion of sun-flower oil to paraffins and Kubicka et al. (2009) demonstrated thatCo–Mo supported over mesoporous alumina enhanced deoxygen-ation activity with rapeseed oil. Kumar et al. (2010) showed that aCo–Mo/Al2O3 hydrotreating catalyst had an unusual cracking abilityfor jatropha-oil with a high acid content. Co–Mo/Al2O3 despitebeing a low acidity catalyst gave an unexpectedly high yield (40%)of cracked hydrocarbon products. The present study was carriedout to identify the reaction pathways for plant-oil hydroconversionover a CoMo system using a low free fatty acid feed and to gain anunderstanding of the role of acidic intermediates in cracking.Lumped power law-based kinetic models (Kumar and Froment,2007; Zhang et al., 2009) were validated with experimental data.The model which best predicted the concentration profile of variouscomponents was chosen and kinetic parameters and activation en-ergy were evaluated. Jatropha oil which contains mainly triglycer-ides of C16 and C18 hydrocarbons was used as a feed stock toproduce deoxygenated hydrocarbons as the main reaction products.

Table 1Catalyst loading and processing conditions.

Catalyst 4% CoO, 16%MoO3, 1%P2O5 on c-Al2O3

Catalyst mass, g 2Catalyst volume, ml 2.4Catalyst shape PowderedBed length, cm 2.8Bed volume, ml 3.7165Reaction temperature, oC 320–360Reaction pressure, bar 20–90LHSV, hr�1 0.8–8.0H2/FEED, Nl/l 500–2000H2/FEED, molar ratio 21.1–84.4

LHSV, liquid hourly space velocity.

2. Methods

Jatropha oil (1.7% FFA (free fatty acids), 19.5% C16:0, 7.9% C18:0,45.4% C18:1, 27.3% C18:2, 77.01 wt% C, 13.6 wt.% H, 9.39 wt.% O,4.0 ppm S) was hydroprocessed in a fixed-bed reactor over a sulph-ided Co–Mo/Al2O3 catalyst. The catalyst was prepared by impreg-nation of mesoporous support extrudates of c-Al2O3 (BET surfacearea = 298 m2/g, pore volume of 1.1 ml/g, and BJH pore si-ze = 6.1 nm) (Kumar et al., 2010). The catalyst was powdered andmixed with �40 + 50 mesh SiC (1:1 volume/volume) and loadedinto a stainless steel tubular reactor with a 1.3-cm internal diame-ter and a heated zone length of 30 cm, with a-Al2O3 extrudate lay-ers above and below the catalyst bed. The reactions wereperformed in a commercial bench top micro-reactor (AutoclaveEngineers’ BTRS-Jr�) (Kumar et al., 2010). Liquid products werewithdrawn after stabilization of reaction conditions in two-hourintervals (at each temperature). The liquid products were analyzedthrice during the stabilization period by gas chromatography (GC)to determine if the activity of the catalyst was constant. The liquidproducts were analyzed offline after separation of the water phaseand diluting 100-fold in CCl4, using a gas chromatograph (Varian3800-GC) equipped with a flame ionization detector and a thermalconductivity detector. The liquid products were analyzed in a vf-5ms column (30 m � 0.25 mm, 0.25 lm) for hydrocarbons, freefatty acids and triglycerides. Internal standards (i.e., eicosane) wereused for quantification. The quantitative results for the triglycer-

ides were compared with those from HPLC analysis after derivati-zation. Injection temperature was set at 340 �C. A high injectionport temperature was used for reliable and direct quantificationof fatty acids and triglycerides without chemical derivatization(Fu et al., 2010; Peng et al., 2012). The temperature programwas: from 35 to 150 �C (rate: 3 �C/min and holding time for5 min), increase to 300 �C (rate: 12 �C/min and holding for 5 min)and increased to 320 �C (rate: 15 �C/min and holding time for15 min). The vapor phase was analyzed online by a gas chromato-graph (Agilent 7890A, RGA) with 2-TCD detectors, 1-FID and sixcolumns (5 packed columns and 1 capillary column from Agilent).The yield fractions were calculated on a relative basis consideringthe entire range of products formed as 100%. A complete mass bal-ance was calculated for the entire reactor and was always >99%accurate for all the components before and after the reaction.

Table 1 details the processing conditions for the kinetic analysisof the hydrotreatment of jatropha oil to understand the effects oftemperature, pressure, H2/feed ratio (Nl/l) and space velocity onproduct patterns. Deactivation of the catalyst during the runswas estimated by repeating the same experiment again after2 days of continuous operation to confirm, and if the change inactivity was more than 2%, fresh catalyst was loaded and used.

The feed for the reaction was of low TAN (Total Acidity Number)and mixed with 0.1% DMDS (dimethyl disulfide) for maintainingcatalytic activity and to minimize the cracking tendency observedin high TAN feed (Kumar et al., 2010). The feed was taken as onelump and the products were lumped in four lumps, i.e. heavy (fromC15–C18), middle (from C9–C14), light (<C9) and oligomerized(>C18). The Levenberg–Marquardt algorithm (LMA) was used forestimating the kinetic parameters (Alvarez and Ancheyta, 2008;Kumar and Froment, 2007). The values of the predicted kinetic rateconstants were analyzed using CHI Square test (v2) for goodness offit, with a level of significance of 95%. The lumped models whichdid not satisfy the v2 criteria were rejected. The rate constantswere used to calculate the activation energy and frequency factorusing the Arrhenius equation (Callejas and Martinez, 2000).

2.1. Model development

Lumped kinetic models were developed considering variousproduct patterns observed during the reactions (Kumar andFroment, 2007; Zhang et al., 2009). Various possibilities wereconsidered for the formation of different primary and secondaryreaction products (Scheme 1).

Differential rate equations were framed for the formation ofthese lumps and solved simultaneously for each scheme separatelyto obtain a system of non-linear equations, for various concentra-tions of lumps (detailed methodology in Supplemental Informa-tion). The equations were reduced to the formY ¼ Aieax þ BiebxþCiecx þ D where x is space time and Y is the concentration of indi-vidual lumps (Supplemental Table 1).

150 M. Anand, A.K. Sinha / Bioresource Technology 126 (2012) 148–155

The five lump models considered the triglycerides in feedjatropha oil as a lump and the liquid hydrocarbon products pro-duced after hydrotreating were divided into 4 lumps i.e. lighters(C5–C8), CL; middle distillates (C9–C14), CJ; heavy (C15–C18), Ch;and oligomerized (>C18), CP product. The four-lump model consid-ered all lumps similar to the five-lump models except that thelighter lump, CL, included the C5–C14 range. In the A1, A2, A6, A7and A8 model, the triglycerides in the jatropha molecules werenot only deoxygenated into C15–C18 but also directly convertedto other products. The A2, A6, A7 and A8 model specifically consid-ered internal conversions between the reaction products with dif-ferent routes for the formation of oligomerized products (>C18), A1instead did not consider any internal conversion between the pri-mary products. In the A3 and A5 models, the triglyceride moleculeswere only deoxygenated into heavy i.e. C15–C18 and oligomerized

Scheme 1. Four lump and five lump kinetic models for the hydroconversion of triglyceri(‘Cn’ hydrocarbons with carbon number ‘n’).

i.e. >C18 and then the deoxygenated products were further con-verted to lighter and middle distillates. The A5 model also consid-ered the internal conversion of oligomerized products to heavylump i.e. C15–C18, whereas the A4 model only considered deoxy-genation reaction and further secondary conversion of these deox-ygenated products to lighters (C5–C14) and oligomerization (>C18)products without internal conversion. The solution to these modelswas obtained after solving the differential rate equations, consider-ing a first order for all the lump profiles. (Supplemental Table 1).

2.2. Model evaluation

The system of equations obtained from the models was solvedsimultaneously using the Levenberg Marquardt algorithm (LMA).To cross-check the results, the rate parameters in each model were

des molecules (TG) to Light (CL), Middle (CJ), Heavy (CH), Oligomerized (CP) products

Table 2Conversion table for triglycerides as a function of process variables (a) temperature(80 bar, 1500 H2/FEED); (b) pressure (1500 H2/FEED, 4 h�1, 360 �C); (c) H2/feed ratio(80 bar, 360 �C); (d) space velocity (80 bar, 360 �C, 1500 vol/vol).

a Temp,�C

Conversion b Pressure, bar Conversion

4 h�1 6 h�1 8 h�1 80 98.5

320 92.2 60 94.1340 95.1 92.9 89.9 40 91.7360 98.5 97.9 98.0 20 63.0

c dH2/feed Conversion Space velocity Conversion

N1/1 4 h�1 6 h�1 8 h�1 h�1

2000 96.7 96.6 97.3 8.00 98.001500 98.5 97.9 98.0 6.00 98.001000 96.1 95.6 95.3 4.00 98.54500 96.0 94.5 89.8 2.00 98.97

1.00 99.90

Fig. 1. Mean yield fractions of various lumps (‘TG’ Triglycerides; ‘Cn’ hydrocarbonswith carbon number ‘n’) as a function of reaction temperature (80 bar, 1500Nl/l H2/FEED, 8 h�1).

Fig. 2. Mean yield fractions of various lumps (‘TG’ Triglycerides; ‘Cn’ hydrocarbonswith carbon number ‘n’) as a function of reaction temperature (80 bar, 1500Nl/l H2/FEED, 6 h�1).

Fig. 3. Mean yield fractions of various lumps (‘TG’ Triglycerides; ‘Cn’ hydrocarbonswith carbon number ‘n’) as a function of reaction temperature (80 bar, 1500Nl/l H2/FEED, 4 h�1).

M. Anand, A.K. Sinha / Bioresource Technology 126 (2012) 148–155 151

also estimated with the simplex method (Zhang et al., 2009). In boththe LMA and simplex methods, different initial points were taken toreach a global minimum. The predicted values of the kinetic rateparameters form the models were tested for goodness of fit usingCHI square (v2) analysis at a 95% significance level. The models weretested at 320, 340 and 360 �C. The number of observations or fre-quencies for the analysis was the concentration of lumps i.e. thereactions products at various space velocities. The results for all rateparameters for each model along with their CHI Square values arelisted in Supplemental Table 2. The limit for the 95% significance le-vel for 5� of freedom was 11.07, for 3�, 5.911 and for 2�, 3.8411; onlythe hypothesis of the models which resulted in a v2 values less thanthe acceptable limits were accepted. R2 and error percentage wereother criteria for acceptance or rejection of the models.

3. Result and discussion

The main reactions during hydroprocessing of jatropha oilwere the saturation of unsaturated hydrocarbon chains followedby hydrodeoxygenation/decarbonylation/decarboxylation for theremoval of the oxygen atom (Liu et al.,2009) and furtherhydrocracking.

3.1. Temperature

The conversion of triglycerides increased rapidly with the in-crease in reaction temperatures, with 98% conversion at 360 �Ccompared to 89% conversion at 340 �C (Table 2a). Also, the conver-sions at different space velocities (4–8 h�1) were higher at 360than at 340 �C. There was only a slight increase in the yield ofcracked products, light fraction (<C9) from 320 to 360 �C (Figs.1–3), but the increase was more obvious for the C9–C14 (middlefraction) of cracked products at higher temperatures (360 �C).The yield of oligomerized products (>C18) was high and increasedtill 340 �C at higher space velocities of 8 (Fig. 1) and 6 h�1 (Fig. 2)and decreased drastically at 360 �C, whereas at lower space veloc-ities, there was a linear decrease in formation of oligomerizedproducts from lower to higher temperature (Fig. 3). The deoxygen-ation products (i.e., C15–C18 hydrocarbons) decreased initially till340 �C at lower residence times (6, 8 h�1) and increased rapidly at360 �C (Figs. 1 and 2), but at higher residence time (Fig. 3), therewas a linear increase at higher temperature. This observationwas attributed to the fact that at lower temperatures and shorterresidence times, the triglycerides were prone to oligomerization,whereas at longer residence times, the triglycerides were directlyconverted to C15–C18 hydrocarbons. The results indicated theneed to operate at higher residence times and reactions tempera-tures to avoid undesirable oligomerization.

3.2. Pressure

Table 2b. details the conversion of triglycerides at differentpressures. Above 40 bar, there was a slight increase in the conver-sion of the triglycerides from 91% to 98% conversion at approxi-mately 80 bar. A similar observation was reported by Kubickaet al. (2009), who saw a drastic decrease in conversion when thereactor pressure decreased. A reduction in conversion of approxi-mately 31% was observed when the reactor pressure was reducedto 20 bars (Table 2b). This result was attributed to reduced partialpressures of hydrogen leading to hydrogen mass transfer limita-tions on the catalyst surface (Kubicka et al., 2009). At higher reac-tor pressures of around 40 bar, there was enough pressure for 91%conversion of triglyceride molecule, but not enough to convert the20% oligomerized (>C18) product (Fig. 4). Only above 60 bar of

Fig. 4. Mean yield fraction of various lumps (‘TG’ Triglycerides; ‘Cn’ hydrocarbonswith carbon number ‘n’) as a function of reactor pressure (1500 Nl/l H2/FEED, 4 h�1,360 �C).

Fig. 5. Mean yield fraction of various lumps (‘TG’ Triglycerides; ‘Cn’ hydrocarbonswith carbon number ‘n’) as a function of H2/FEED ratio (80 bar, 8 h�1, 360 �C).

152 M. Anand, A.K. Sinha / Bioresource Technology 126 (2012) 148–155

reactor pressure were the oligomerized products converted to hea-vy hydrocarbons, i.e. C15–C18. There was a slight effect of pressureon the yield of lighter and middle cracked products, as their yieldsincreased from 3–8% and 8–12%, respectively when the reactorpressure was reduced from 90 to 20 bar. A lower hydrogenation(saturation) rate of the unsaturated primary intermediates at lowerreaction pressures could be responsible for increased yields ofcracked products at lower pressure. The oligomerized (>C18) prod-ucts increased from 7% to 20% as the pressure was reduced to60 bar and remained constant on further reduction to 20 bar. Ini-tially the conversion of triglycerides increased rapidly from 20 to40 bar (Table 2b). leading to an increased rate of formation of deox-ygenated products. The rapid increase in heavy lump yield (i.e.C15–C18 hydrocarbons) above 60 bar was mainly attributed tothe sudden decrease in oligomerized (>C18) products (Fig. 4) inaddition to further conversion of triglycerides molecules. Thereforeit was necessary to work at higher pressure (80 bar) to obtain thebest catalytic performance.

Fig. 6. Mean yield fraction of various lumps (‘TG’ Triglycerides; ‘Cn’ hydrocarbonswith carbon number ‘n’) as a function of space velocity (80 bar, 360 �C, 1500 Nl/l H2/FEED).

3.3. Hydrogen to feed ratio

The average number of C–C double-bonds in the triglycerides,was 1–2 per molecule and olefins accounted for 76 mol%. The H2/triglyceride molar ratio in the present study was 30–60 times high-er than the average number of C–C double-bonds in the triglycer-ides. Table 2c. details the effect of the hydrogen to triglyceride ratio(H2/feed) on the conversion of triglyceride molecules. There was aslight improvement in the conversion of triglycerides at 8 h�1

space velocity as compared to velocities of 6 and 4 h�1, withincreasing H2/feed ratio, but at lower space velocities (4, 6 h�1)there was no considerable effect of H2/feed ratio on the conversion.The concentration of oligomerized (>C18) product decreased rap-idly from 27% at 500 to 7% at H2/feed ratio of 2000 (SupplementalFig. 1) (at 360 �C, 4 h�1 and 80 bar pressure). This sharp decrease inthe formation of oligomerized products at a higher H2/feed ratio(>1500 Nl/l) led to an increase in the yield of the heavy fractioni.e. C15–C18 from 50% to 80% (Fig. 5). At space velocity of 8 h�1

or higher and at a H2/feed ratio of 500 Nl/l or lower, the formationof undesirable oligomerized (>C18) products was increased but atspace velocities lower than 6 h�1, there was a drastic decrease inthe yield of these products. At H2/feed ratios higher than 1500Nl/l, the space velocity did not influence the yield of the oligomer-ized products. Hence 1500 Nl/l or more was an optimum ratio forminimizing the oligomerized product yield and to increase the life-span of the catalyst.

The yield of cracked products (light fractions) was suppressedwith increased H2/feed ratio (Fig. 5) due to a higher hydrogenation(saturation) rate of the unsaturated primary intermediates withincreasing H2/feed.

3.4. Liquid hourly space velocity (LHSV)

There was a very small but gradual increase in the unconvertedtriglycerides in the product stream with increased space velocityfrom 1 to 8 h�1 (Fig. 6). Even at a space velocity of 8 h�1 at 360 �C,the conversion of triglycerides was around 98% and increased to al-most 99.9% at around 1 h�1 (Table 2d.). At a space velocity of 1 h�1,there was a considerable increase in the yield of the light and middlefraction lumps (18%) attributed to enhanced cracking due to an in-creased residence times in the reactor. Two stable intermediateswere observed at 360 �C (Supplementary Figs. 2 and 3). These wereoxygenated intermediates such as esters formed by esterification offatty acids (a reaction intermediate) and alcohols (another reactionintermediate). The IR and NMR analyses showed major peaks as-cribed to the presence of esters and some acids, alcohols, aldehydesand ethers (Supplemental Figs. 4 and 5). The oligomerized productsformed on the CoMo active phase and were not observed on purealumina support under the reported reaction conditions (Table 1).The composition of oligomerized products was constant althoughtheir concentration varied, depending upon the severity of thereaction conditions. These stable oxygenated intermediates werethe same as those reported by Kubicka et al. (2009) at lowertemperatures and Huber et al. (2007) at higher space velocities.No oligomeric products could be found in the gaseous phase, inthe experimental volume between the catalyst bed and the gas–li-quid separator which was confirmed by the gas sample analysisprior to gas–liquid separation.

After observing the trends and the product patterns at differentprocess conditions, it was concluded that the hydroprocessing ofjatropha oil over Co–Mo catalyst for the production of hydrocarbonfuels was to be carried out at a reaction temperature of 360 �C forcomplete conversion of triglyceride molecule. The reactor pressureand the H2/feed should be kept around 80 bar and 1500 Nl/l,

Table 3Overall mass balance and atomic balances for two sets of reactions at 80 bar and 1500Nl/l H2/feed ratio.

Overall mass balance 360 �C, 2 h�1 360 �C, 8 h�1

In, g/h Out, g/h In, g/h Out, g/h

Gas 0.63 1.31 2.59 4.55Liquid 4.42 3.73 17.66 15.7Total* 5.05 5.04 20.25 20.25

Atomic balance Feed, g/h Product, g/h

360 �C, 2 h�1 Gas Liquid Gas + H2O Liquid organic

Carbon 0 3.4 0.32 3.1Total* 3.4 3.42Hydrogen 0.625 0.6 0.67 0.56Total* 1.225 1.23Oxygen 0 0.415 0.33 0.087Total* 0.415 0.417

360 �C, 8 h�1

Carbon 0 13.6 0.24 13.36Total* 13.6 13.6Hydrogen 2.59 2.4 2.9 2.09Total* 4.99 4.99Oxygen 0 1.66 1.41 0.25Total* 1.66 1.66

* Total = gas + liquid (g/h)

M. Anand, A.K. Sinha / Bioresource Technology 126 (2012) 148–155 153

respectively to minimize the formation of oligomerized products,i.e. >C18 molecules. The space velocity could be varied from 1 to8 depending on the product pattern desired. Stearyl stearate wasalso identified earlier as the dominant ester at lower space velocity(1.5 h�1) and lower reaction temperature (<300 �C) (Kubicka et al.,2009). The authors had shown that the selectivity for oxygenatedproducts (as sum of acids, alcohols and esters; considered in thecurrent study as oligomerized product) increased over Co–Mocatalysts with decreasing reaction temperature. Since the spacevelocity studied by Kubicka et al. (2009) was only 1.5 h�1, theseintermediates were virtually absent at temperatures above300 �C. In the current study, very low concentrations of these inter-mediates were observed at higher reaction severities, but largequantities of these intermediate oligomeric oxygenates (selectivityas high as 50%) were observed below 280 �C. Since the space veloc-ities considered in the current study were much higher than thoseof Kubicka et al. (2009), a high amount of these intermediates wasobserved. These intermediates increased rapidly with the increasein space velocity (Supplemental Figs. 1 and 2). Huber et al. (2007)also reported that at a very high space-velocity of 17.5 h�1 at350 �C, a white waxy oxygenated intermediate was formed. Atlow conversions (high space time >10 h�1, low temperature <300 �C) selectivity for the oxygenated product intermediates wasvery high and some waxy products formed, with chemical struc-tures similar to those of the triglyceride and free fatty acid inter-mediates. These products rapidly choked the reactor and madestudies difficult. This result indicated that oxygenated intermedi-ates were formed prior to deoxygenation via the pathwayTG ? hydrogenation ? [waxes, glycerides, fatty acids, esters] ?hydrocarbons (Huber et al., 2007).

Mikulec et al. (2010) used NiW and NiMo catalytic systems andprocessed rapeseed oils at 330–340 �C and 1 h�1 and observed lowamounts of >C18 products. Veriansyah et al. (2012) also worked ata higher reaction temperature of 400 �C and lower space velocity(2 h�1) using soyabean oil as the feed stock and did not report sucholigomeric intermediates. Donnis et al. (2009) reported that in thehydroprocessing of methyl n-dodecanoate (methyl laurate) overNiMoP catalyst at low and intermediate conversion levels, theproducts and intermediates were large amounts of 1-dodecanol,small amounts of 1-dodecanal and dodecanoic acid, n-donedaneand n-dodecenes as well as n-undecane and nundecenes. Addi-tional products like dodecyl dodecanoate and didodecyl ether wereseen. The authors explained that it was the adsorbed enol form ofthe dodecanoate or the dodecanal which was the reactive interme-diate for hydrocarbon formation.

3.5. Mass balance

The overall mass balance of the reacting systems was carriedout with less than 1% error (Table 3). Atomic balance for carbonand hydrogen were also conducted for two sets of reactions, whichvery well correlated with the experimental results detailed by GCanalysis. At the low space velocity of 2 h�1 at 360 �C, i.e. at higherseverity, about 9% of the total carbon by weight fed into the systemin the form of triglycerides was found in the gas phase and theremaining 91% by weight was in the liquid phase. At lower sever-ity, i.e. 8 h�1 at 360 �C, only 2% of carbon by weight was observedin the gas phase and rest in the liquid phase. These results are inagreement with the fact that almost 50% less cracked products(<C9 and C9–C14) were observed at 8 h�1 than at 2 h�1 at 360 �C(Fig. 6). Table 5 describes both overall mass balance and atomicbalance for the two reaction sets. The results justify the use ofthe liquid phase analysis results for mathematical modeling. More-over, the model aims to predict the distribution of liquid phaseproducts while the gasses produced were minor side products ofliquid hydrocarbon cracking. For example – when a liquid product

such as a C13 hydrocarbon is produced it would be accompaniedby a corresponding minor gaseous hydrocarbon product (such asC5 hydrocarbon); when heavy lump (C15–C18 fraction) liquidproducts are formed, C3 hydrocarbon is the corresponding minorgas phase product.

3.6. Model validation

Based on the reaction mechanisms for the hydroprocessing ofplant-oil triglycerides, different lumped kinetic models wereframed. These models were developed considering various productpatterns observed during the reactions. All the deoxygenationproducts which include decarboxylation, decarbonylation and hy-dro deoxygenation were lumbed as one. Also the rate constant ofall the deoxygenation reactions were lumped as one to establishfour-lump and five-lump kinetic models. Various possibilities wereconsidered for the formation of different primary and secondaryreaction products (Scheme 1).

Based on the model fit, the possibility for the A4, A5 and A8models at all temperatures was ruled out due to large deviationin v2 and/or R2 values from the acceptable limits (SupplementalTable 2). It means that the direct conversion of triglycerides intoonly heavy fractions (C15–C18) (1st step) and further conversionof heavy fractions into lighter fractions (<C15) and oligomerizedproducts (>C18) (model A4) is ruled out.

Similarly, conversion of triglycerides directly into heavy andoligomerized products, followed by conversion of the heavy frac-tion into light and middle fractions (model A5) was also ruledout. The conversion of heavy fractions (C15–C18) directly to lighterhydrocarbons in the A8 model was ruled out because the modeldid not satisfy the 95% level of significance at any temperature.

At lower temperatures in the order of 320 �C, the A3 model fitbest (Table 4). There was no formation of lighter and middle distil-lates directly from triglycerides and only deoxygenation wasfavored. There was also formation of oligomerized (>C18) productdirectly from the triglycerides. These facts were supported by theexperimental data, which clearly showed an increase in the forma-tion of lighter and middle fractions only at lower space velocitiesand higher temperatures. At lower temperatures these fractionswere only formed as secondary products (0.5–1% yield) as com-pared to the primary products i.e. heavier and oligomerized

Table 4Rate constant values for accepted lumped models at 80 bars and 1500Nl/l H2/feedratio.

Temperature, kineticmodel

Rate constant Value,h�1

Error

320 �C, A3 k0(Triglycerides) 17 0.82k1 (Heavy) 10 0.50k2 (Oligomerized) 7 0.32k3 (Lighter) 0.3 0.16k4 (Middle) 0.3 0.16

340 �C, A2 k’ (Triglycerides) 17 1.25k1 (Lighter) 1.1 3.70k2 (Middle) 0.2 0.62k3 (Heavy) 3 1.12k4 (Oligomerized) 13 1.41k5 (Oligomerized toheavy)

5 0.83

k6 (Heavy to middle) 0.4 0.49k7 (Middle to lighter) 3 3.83

360 �C, A7 k0 (Triglycerides) 24 2.19k1 (Lighter) 1 0.24k2 (Middle) 2 0.30k3 (Heavy) 19 1.64k4 (Oligomerized) 2 0.26k5 (Heavy to middle) 0.1 0.02k6 (Middle to lighter) 0.1 0.1

Table 5Frequency factor and activation energy values for various lumped products andreactants.

Reactant and Productlumps

Frequency Factor,h�1

Activation energy

Value, kJ/mole

Error%

Triglycerides 31.97E02 26 1.9Deoxygenated product 14.58E04 47 6Lighter products 62.62E05 83 3.9Middle products 4.2E10 127 0.3Heavy products 13.73E04 47 2.6

154 M. Anand, A.K. Sinha / Bioresource Technology 126 (2012) 148–155

(>C18) lumps, which were formed at a much larger scale (35–75%yield fraction).

At intermediate temperatures of 340 �C, A2 was the only modelwhich satisfied all the criteria, i.e. goodness of fit test and thecorrelations with experimental results (Table 4). The lighter andthe middle fraction were not only formed from deoxygenatedproducts but also from the direct conversion of the triglycerides.The oligomerized (>C18) product which was formed at 340 �Cwas converted to heavy (C15–C18) hydrocarbons, indicating thatthese oligomerized products were unstable and prone to conver-sion at long residence times and higher temperatures of 340 �C,leading to less coking and ultimately slower deactivation of thecatalyst.

At 360 �C, the A3 model did not satisfy the criteria for accep-tance as a valid model, but the A2 model was within the accep-tance levels of the required criteria, although it gave a negativevalue for the rate parameter for the conversion of oligomerized(>C18) product to heavy hydrocarbons (k5). Hence another modelwas looked into which took into account the conversion of heavyhydrocarbons to oligomerized products (A6 model). Although thismodel also gave acceptable results, it showed an error of more than100% in the prediction of the k5 rate parameter. Hence both ofthese models were also rejected. The simple A1 model also satis-fied the goodness of fit test, but the A7 model which took into ac-count the internal conversion of heavy deoxygenated products(C15–C18) to middle fractions (C9–C14) and then further conver-sion to lighter fractions (C5–C8) yielded the best results at highertemperatures (Table 4). The error in the prediction of these param-eters was the smallest. Hence this model had the best fit for the

conversion of triglycerides at higher temperatures. This best-fittingmodel predicts the conversion of triglyceride molecules to oligo-merized (>C18) products which were stable and not furthercracked into other products. There was a constant yield of theseoligomerized products at all space velocities and at 360 �C whichclearly indicated that the oligomerized products formed at suchtemperatures were not converted further and no other productscontributed towards their formation other than triglycerides. Thisoligomerized product could be the coke precursor and may accountfor the faster deactivation observed at these conditions.

Reaction temperatures control the pathways because of the dif-ferences in the concentration of acidic intermediates formed at dif-ferent temperatures as well as due to differences in reactionseverity. At lower temperatures, when the condition was less se-vere, the lighter and the middle range fractions were the secondaryproducts in low concentrations from the direct deoxygenation ofprimary products of the reaction. The A3 model also predicted nofurther conversion of the oligomerized (>C18) product whichwould mean accumulation of this product in the reaction system.Operating at this temperature for long durations, the reactor tendsto get choked (an increase in DP in the reactor). Hence such lowtemperatures are not appropriate for continuous operation.

With increase in reaction severity by raising the reaction tem-perature, it was expected that cracking products would startforming directly from the triglycerides along with secondarycracking of heavier hydrocarbons. This hypothesis was validatedby the best-fitting model (A2) at moderate temperatures since itpredicted direct formation of lighter and middle products alongwith deoxygenated and oligomerized (>C18) products by internalconversions. The model also predicted further conversion of theoligomerized product which would minimize accumulation ofthis product in the reaction system and no reactor chokingwas observed at moderate temperature during prolonged contin-uous operation.

At higher temperatures, along with internal conversions and di-rect formation of lighter, middle and deoxygenated fractions, lowconcentrations (<10%) of oligomerized (>C18) product was ob-served. Model A7 predicted that, despite cracking of most of theoligomeric intermediate by increase in reaction temperature, itwas not possible to avoid the presence of these oxygenated com-pounds in the product mixture at this temperature. A certainamount of oligomeric product (e.g., �7% at 360 �C and 4 h�1) re-mained stable despite the increase in reaction temperature.

At the beginning of this study, it was hypothesized thatthe acidic intermediates formed during the reaction over Co–Mocatalyst catalyzed the unexpectedly high cracking. The TAN ofthe products at 340 �C was nearly six times higher than that at360 �C. So, at intermediate temperatures (340 �C), there was an in-creased formation of acidic intermediates which promoted the for-mation of cracked products. At higher temperatures (360 �C), theseacidic intermediates were formed in lower amounts (6-times lowerin the product) and hence lower cracking was expected. This factwas supported by higher rate constant values at 340 �C than at360 �C for – (i) triglyceride conversion to lighter products (k1),(ii) triglyceride conversion to oligomeric products (k4), (iii) for con-version of heavy (C15–C18) hydrocarbons to middle (C9–C14)hydrocarbons [k6 (340 �C) > k5(360 �C)] and (iv) middle (C9–C14)hydrocarbons cracking to lighter products (C5–C8) [k7 (340 �C) >k6 (360 �C)]. Thus these observations based on the predictedreaction pathways support our hypothesis. Refined and bettermodels which would consider product intermediates as well asreactant lumps in a more detailed manner should be set up. Look-ing into the possibility of generating more data at varying and verylow conversions would help in better discrimination between themodels. Test reactions with model compound instead of complextriglyceride feeds can be envisioned and may include in situ anal-

M. Anand, A.K. Sinha / Bioresource Technology 126 (2012) 148–155 155

ysis of reactants, intermediate and product species in the reactionand this approach would help in obtaining better fits and thus bet-ter discrimination between models.

After the models were validated, apparent activation energyand frequency factors for the formation of various products fromthe triglyceride molecules were calculated (Table 5). These valueswere predicted from the Arrhenius plot. The apparent activationenergy for the conversion of triglycerides was low (in the orderof 31 kJ/Mole at close to 60% conversion), as compared to that forthe formation of other products. This result was in accordance withthe high conversion rates achieved even at the high space velocityof 8 h�1. A much broader range of global triglyceride conversiondata is desirable, in order to better discriminate between models.There was a limitation in working at very low triglyceride conver-sions because very low concentrations of cracked products wereobserved. For example, the amount of cracked products, C9–C14,was negligibly low at the lowest temperature of 320 �C (Fig. 2).At very low conversions, an insoluble waxy product formed whichchoked the reactor. The apparent activation energy for the forma-tion of deoxygenated products and heavier lump was higher thanthat for the conversion of triglycerides, i.e. around 47 kJ/Mole.

The apparent activation energy for lighter and middle distillateswas higher, 85–127 kJ/Mole, and showed that the formation ofthese products required a larger amount of energy by the hydro-treating catalyst, Co–Mo/Al2O3. The apparent activation energiesfor the heavy, middle and the lighter lumps were closer to real val-ues as compared to those for global triglyceride conversion be-cause they were calculated at comparatively lower yields. Thehydrotreating catalyst’s main function is the removal of oxygenatoms present in the triglyceride molecules; the higher values forthe activation energies for the lighter and middle fractions justifiedthis role. If these products are to be more selectively favored, anacidic support, which promotes cracking, should be used. Hence,the considerable amount of cracking observed at higher tempera-ture (360 �C) could be attributed, to a certain extent, to the roleof acidic intermediates (Kumar et al., 2010; Morgan et al., 2010)that catalyze the cracking activity over Co–Mo system.

4. Conclusion

The conversion of jatropha triglyceride molecules to hydrocar-bon fuels could be best achieved at pressures of around 80 barsand H2/feed ratios of 1500 or higher. The reaction temperaturesand space velocity, respectively, should be around 360 �C and1.0–2 h�1. A reduction in pressure or H2/feed ratio resulted inincreased oligomerization reactions leading to more rapid catalystdeactivation. Mechanistic models indicated a strong dependence ofthe conversion of triglycerides to hydrocarbons on temperatures.

The changing reaction pathways with temperatures suggest tolook into similar, but more refined, lumped kinetic models forother catalytic systems and the use of model reactants.

Acknowledgements

Director IIP is acknowledged for approving the research. Mr. Ra-kesh Kumar, R. K. Joshi, T. Khan, P. Alam, IIP are acknowledged forhelping with catalytic evaluations DST, India is acknowledged forresearch funding.

Appendix A. Supplementary data

Supplementary data associated with this article can be found,in the online version, at http://dx.doi.org/10.1016/j.biortech.2012.08.105.

References

Alvarez, A., Ancheyta, J., 2008. Modeling residue hydroprocessing in a multi-fixed-bed reactor system. Appl. Catal. A Gen. 351, 148.

Bezergianni, S., Kalogianni, A., Vasalos, I.A., 2009a. Hydrocracking of vacuum gasoil–vegetable oil mixtures for biofuels production. Bioresour. Technol. 100,3036.

Bezergianni, S., Voutetakis, S., Kalogianni, A., 2009b. Catalytic hydrocracking of freshand used cooking oil. Ind. Eng. Chem. Res. 48, 8402.

Callejas, M.A., Martinez, M.T., 2000. Hydroprocessing of a Maya residue. II. Intrinsickinetics of the asphaltenic heteroatom and metal removal reactions. EnergyFuels 14, 1309.

Donnis, B., Egeberg, R.G., Blom, P., Knudsen, K.G., 2009. Hydroprocessing of bio-oilsand oxygenates to hydrocarbons understanding the reaction route. Top. Catal.52, 229.

Fu, J., Lu, X., Savage, P.E., 2010. Catalytic hydrothermal deoxygenation of palmiticacid. Energy Environ. Sci. 3, 311.

Huber, G.W., Connor, P.O., Corma, A., 2007. Processing biomass in conventional oilrefineries: Production of high quality diesel by hydrotreating vegetable oils inheavy vacuum oil mixtures. Appl. Catal. A Gen. 329, 120.

Krár, M., Kovács, S., Kalló, D., Hancsók, J., 2010. Fuel purpose hydrotreating ofsunflower oil on CoMo/Al2O3 catalyst. Bioresour. Technol. 101, 9287.

Kubicka, D., Simacek, P., Zilkova, N., 2009. Transformation of vegetable oils intohydrocarbons over mesoporous-alumina-supported como catalysts. Top. Catal.52, 161.

Kumar, H., Froment, G.F., 2007. Mechanistic kinetic modeling of the hydrocrackingof complex feedstocks, such as vacuum gas oils. Ind. Eng. Chem. Res. 46, 5881.

Kumar, R., Rana, B.S., Tiwari, R., Verma, D., Kumar, R., Joshi, R.K., Garg, M.O., Sinha,A.K., 2010. Hydroprocessing of jatropha oil and its mixtures with gas oil. GreenChem. 12, 2232.

Lappas, A.A., Bezergianni, S., Vasalos, I.A., 2009. Production of biofuels via co-processing in conventional refining processes. Catal. Today 145, 55.

Liu, Y., Sotelo-Boyas, R., Murata, K., Minowa, T., Sakanishil, K., 2009. Hydrotreatmentof Jatropha oil to produce green diesel over trifunctional Ni–Mo/SiO2–Al2O3

catalyst. Chem. Lett. 38, 552.Melis, S., Mayo, S., Leliveld, B., 2009. Vegetable oil co-processing in diesel

hydrotreaters. Biofuels Technol. 1, 43.Michaelsen, N.H., Egeberg, R., 2009. Consider new technology to produce renewable

diesel. Hydrocarbon Process., 41.Mikulec, J., Cvengros, J., Jorikova, L., Banic, M., Kleinova, A., 2010. Second generation

diesel fuel from renewable sources. J. Cleaner Prod. 18, 917.Morgan, T., Grubb, D., Jimenez, E.S., Crocker, M., 2010. Conversion of triglycerides to

hydrocarbons over supported metal catalysts. Top. Catal. 53, 820.Murata, K., Liu, Y., Inaba, M., Takahara, I., 2010. Production of synthetic diesel by

hydrotreatment of jatropha oils using Pt�Re/H-ZSM-5 catalyst. Energy Fuels 24,2404.

Peng, B., Yao, Y., Zhao, C., Lercher, J.A., 2012. Towards quantitative conversion ofmicroalgae oil to diesel-range alkanes with bifunctional catalysts. Angew.Chem. Int. Ed. 51, 2072.

Šimácek, P., Kubicka, D., Šebor, G., Pospíšil, M., 2009. Hydroprocessed rapeseed oil asa source of hydrocarbon-based biodiesel. Fuel 88, 456.

Tiwari, R., Rana, B.S., Kumar, R., Verma, D., Kumar, R., Joshi, R.K., Garg, M.O., Sinha,A.K., 2011. Hydrotreating and hydrocracking catalysts for processing of wastesoya-oil and refinery-oil mixtures. Catal. Commun. 12, 559.

Veriansyah, B., Han, J.Y., Kim, S.K., Hong, S.A., Kim, Y.J., Lim, J.S., Shu, Y.W., Oh, S.G.,Kim, J., 2012. Production of renewable diesel by hydroprocessing of soybean oil:effect of catalysts. Fuel 94, 578.

Verma, D., Kumar, R., Rana, B.S., Sinha, A.K., 2011. Aviation fuel production fromlipids by a single-step route using hierarchical mesoporous zeolites. EnergyEnviron. Sci. 4, 1667.

Zhang, S., Yan, Y., Li, T., Ren, Z., 2009. Lumping kinetic model for hydrotreating ofbio-oil from the fast pyrolysis of biomass. Energy Sources, Part A: Recovery,Utilization, and Environmental Effects 31, 639.