Applied Catalysis B: Environmental 104 (2011) 136143

Contents lists available at ScienceDirect

Applied Catalysis B: Environmental

journal homepage: www.elsevier.com/locate/apcatb

Catalytic hydrotreatment of crude algal bio-oil in supercritical water

Peigao Duan a,b, , Phillip E. Savage b
a b

College of Physics and Chemistry, Department of Applied Chemistry, Henan Polytechnic University, No. 2001, Century Avenue, Jiaozuo 454003, Henan, PR China Department of Chemical Engineering, University of Michigan, Ann Arbor, MI 48109-2136, USA

a r t i c l e

i n f o

a b s t r a c t
We report herein on the catalytic hydrotreatment of crude bio-oil, produced from the hydrothermal liquefaction of a microalga (Nannochloropsis sp.) over palladium on carbon (5% Pd/C) in supercritical water (SCW) at 400 C and 3.4 MPa high-pressure H2 . Inuences of wide ranges of reaction time (varied from 1 to 8 h) and catalyst loading (varied from 5 to 80 wt%) on treated oil composition and yield, gas products composition and yield, and hydrogen consumption were explored. The C, H and energy recoveries were determined. The results demonstrated that longer reaction times and higher catalyst loadings did not favor the treated oil yield due to the increasing amount of gas and coke products formation but did lead to treated bio-oil with higher higher-heating-value (HHV) (4144 MJ/kg) than that of the crude feed. Highest HHV of treated oil (44 MJ/kg) was obtained after 4 h using an 80 wt% intake of catalyst on crude bio-oil. The product oil produced at longer reaction times and higher catalyst loadings, which was a freely owing liquid as opposed to being the viscous, sticky, tar-like crude bio-oil material, was higher in H and lower in O and N than the crude feed, and it was essentially free of sulfur (below detection limits). Typical H/C and O/C molar ratios ranges for the bio-oils treated at different reaction times and catalyst loadings were 1.651.79 and 0.0280.067, respectively. The main gas-phase products were unreacted H2 , CH4 , CO2 , C2 H6, C3 H8 and C4 H10 . Overall, many of the properties of the treated oil obtained from catalytic hydrotreatment in SCW in the presence of Pd/C are very similar to those of hydrocarbon fuels derived from fossil fuel resources. 2011 Elsevier B.V. All rights reserved.

Article history: Received 16 October 2010 Received in revised form 7 February 2011 Accepted 16 February 2011 Available online 23 February 2011 Keywords: Catalytic hydrotreatment Algal bio-oil Treated bio-oil Palladium on carbon Supercritical water

1. Introduction Global energy crisis and environmental concerns have motivated more researchers to think about alternatives for fossil derived liquid transportation fuels. Biomass, as a promising alternative due to its high abundance, renewability and carbon neutral, is being paid more attention worldwide. Bio-oil derived from the direct hydrothermal liquefaction of wet biomass, which has lower oxygen content and higher higher-heating-values (HHV) than that of from fast pyrolysis, shows much promise as feedstock avoids the large energy penalty associated with vaporizing the moisture content [1,2]. Of the different biomass candidate for hydrothermal liquefaction, microalgae are very promising feedstock for alternative energy generation because of their higher photosynthetic efciency, faster growth rate, and higher area-specic yield relative to terrestrial biomass. Moreover, microalgae farming for bio-oil production provide a solution to the food versus fuel debate [35]. However, crude algal bio-oils derived from the noncatalytic or catalytic hydrothermal liquefaction of microalgae [7,8] still contain

Corresponding author. Tel.: +86 0391 3987811; fax: +86 0391 3987811. E-mail address: pgduan@hpu.edu.cn (P. Duan). 0926-3373/$ see front matter 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.apcatb.2011.02.020

heteroatoms (O, N, S) which are much higher than the minimum requirement of ASTAM [6] and thus further chemical or physical upgrading is needed. Up to now, numerous upgrading strategies that improve crude bio-oil quality and reduce oxygen, nitrogen and sulfur content were reported in the literature, including hydrotreatment (e.g., hydrodeoxygenation, hydrodenitrogenation and hydrodesulfuration), and hydrocracking [9,10]. Of these proposed crude bio-oil upgrading strategies, catalytic hydrotreatment is a promising one which involves the treatment of crude bio-oil with hydrogen in the presence of a heterogeneous catalyst. The main advantage of this technology is that it requires infrastructure which is widely available in all existing petroleum renery units. However, this upgrading process usually requires high partial pressure hydrogen (up to 30 MPa) to enhance the intrinsic H2 solubility in the liquid phase. H2 solubility in the liquid phase is inversely proportional to temperature, overcoming this solubility limitation at high temperatures often required for hydrotreatment process. One important development in solving this gas solubility problem in liquid phase at higher temperatures is the use of supercritical uids. With employing supercritical uids, the H2 solubility limitation can be potentially improved or eliminated by bringing the feedstock and hydrogen into a single phase. Another advantage of using supercritical uids is that small tuning in pressure or

P. Duan, P.E. Savage / Applied Catalysis B: Environmental 104 (2011) 136143

137

temperature can achieve an optimal combination of liquidlike densities yet gaslike transport properties so that to extract and transport the coke precursors from the catalyst pores to effectively suppress the coke formation and thus extending the life time of catalyst [11]. Since crude algal bio-oil produced from hydrothermal liquefaction will be formed in an aqueous environment, and it may be advantageous from a process engineering perspective to do oil hydrotreatment in that same environment (water). In addition, water is green, safe to handle, inexpensive, and environmentally compatible. The critical temperature of water (374 C) is within the range of temperatures encountered in many oil renery operations. Furthermore, the tunable properties (e.g., density and polarity properties) of SCW in the critical region have a dramatic impact on the reaction selectivity. Therefore, crude bio-oil hydrotreatment in near- or supercritical water (SCW) may be a path toward crude bio-oil upgrading. There have been previous studies of heavy hydrocarbon treatment in SCW [1215]. Most of these previous studies, however, usually involved uncatalyzed treatment. In our previous articles [16,17] on crude algal bio-oil upgrading and upgrading optimization studies in SCW, we treated a crude bio-oil derived from Nannochloropsis sp. in SCW to discover how the properties of the treated bio-oil depend on the experimental conditions. These results suggested that the SCW reaction medium alone seemingly provided complete desulfurization of the crude algal oil, and the abundance of N and O in the treated oils was considerably reduced, though some N and O still remained in the treated oils. Upgrading optimization study [17] further indicated that temperature was always the most inuential for each of the oil properties examined. The lowest temperature examined (430 C), regardless of other conditions, appeared to be the best for producing a treated oil with highest hydrogen content and energy density. However, the desirable properties of this treated oil are very close to that of obtained at 400 C in our upgrading study [16], suggesting that higher temperature would not guarantee to give treated oil with higher hydrogen content and HHV. It is well known that higher temperatures would favor the condensation reactions between aromatics in the treated oil, which can lead to mesophase formation, and consequently coke. In addition, higher temperature will suffer from more energy input penalty. Therefore, from the considerations of carbon efciency and energy conservation, 400 C might be the more suitable temperature to be selected for further studies on the upgrading of crude algal oil. Of the remaining three factors investigated in the upgrading optimization study, catalyst loading had the greatest effect on the HHV and O/C ratio in the treated oil. Reaction time had the greatest effect on the H/C and N/C ratios. Thus, these two factors in achieving desirable effects on the product oil are very important and should be included in the control factors for the further upgrading studies. In present study, crude algal bio-oil was hydrotreated in SCW at 400 C with different reaction times and catalyst loadings. Effect of reaction time (varied from 1 to 8 h) and catalyst loading (varied from 5 to 80 wt%) on the treated oil elemental composition and yield, gas products composition and yield, and hydrogen consumption were examined, respectively. Obviously, catalyst loadings employed in this work sometimes are far greater than the loading typically used for the hydrotreatment reactions. However, our previous study [17] suggested that catalyst loading had the greatest effect on the HHV and O/C ratio in the treated oil. So, rather than refusing to do experiments within the typical catalyst loadings, we opted to increase the catalyst loading beyond the typical scales so as to see the possible highest HHV and lowest O/C ratio in the treated oil. It should be noted that higher catalyst loading is undesirable in view of the economic implication. However, if the catalyst can be recycled, large catalyst loading amount is not a problem. Previously, we used Pt/C as the upgrading catalyst and found

that it had desirable effects on the product oil. In comparison, in this study, we use Pd/C as the hydrotreatment catalyst instead of Pt/C, because previous study [18] suggested that Pd/C catalyst is more active than Pt/C on the deoxygenation of fatty acids. 2. Experimental 2.1. Materials Microalga (Nannochloropsis sp.) paste was purchased from Reed Mariculture (Nannochloropsis 3600TM ). Our earlier articles gave its elemental and chemical composition along with other properties [7,8]. The Pd/C catalyst (5 wt% Pd, metal dispersion = 21%, specic surface area = 863 m2 /g, micropore volume = 0.255 cm3 /g, average particle size = 25 m) was purchased from SigmaAldrich. Both the alga and the catalyst were used as received. Freshly deionized water, prepared in house, was used throughout the experiments. Helium, hydrogen, and argon were obtained from Cryogenic Gases. Gas standards for quantitative analysis were purchased from Air Liquide Specialty Gases. The 400 ml autoclave used for the production of crude bio-oil from the hydrothermal liquefaction of microalgae and the 316-stainles steel batch reactors used for the catalytic hydrotreatment were employed as described in our previous article [16]. 2.2. Procedures 2.2.1. Liquefaction experiment We generated about 12 g of crude bio-oil which resulted in a yield of 34% on the dry microalga base by liquefying 150 mL of the microalga paste in the autoclave reactor. The procedure was nearly identical to that reported previously [16]. In the present experiment, air instead of H2 was charged to prevent the water from vaporizing during the reaction. The dichloromethane in the ltrate was vaporized by using high pressure (140 kPa) nitrogen at room temperature in a fume hood instead of by using a rotary evaporator at 30 C under vacuum in the previous study [16]. It should be noted that the relative proportion of liquid products from the liquefaction of microalgae was dependent on the heating rate, however, variations in the heating rate did not signicantly affect chemical composition of the liquid products [19]. 2.2.2. Upgrading experiments The catalytic hydrotreatment experiments were carried out in a stainless-steel batch mini reactor. The whole procedure was almost identical to our previous publication [16]. In the present study, 0.4 ml of freshly deionized water was added into the reactor. The air inside the reactor was displaced with helium by repeatedly applying vacuum (86 kPa) and charging with 140 kPa He. The reactor was next pressurized to 140 kPa (gauge) with helium and then further charged with hydrogen to 3500 kPa. After the reaction, the reactor was cooled in a refrigerator for 30 min, it was then placed in ambient conditions for at least 15 h to allow the liquidgas system to reach equilibrium prior to performing the gas-phase analysis. The dichloromethane in the separated ltrate was evaporated by using small-pressure (70 kPa) nitrogen at room temperature in fume hood. The treated oil yield was calculated as its mass divided by the mass of crude bio-oil loaded into the reactor. After ltration, the lter paper together with the remained solid residue were dried at 110 C for 12 h in an oven and then weighted. The solid weight was calculated by using the overall weight subtracted the weight of lter paper. The non-catalyst solid residue was dened as coke whose amount was evaluated by subtracting the weight of catalyst from the residue, and its yield was calculated as its mass divided by the mass of carbon loaded into the reactor. It should be noted that the

138

P. Duan, P.E. Savage / Applied Catalysis B: Environmental 104 (2011) 136143

coke product or catalyst alone cannot be recovered. Duplicate independent runs were conducted under nominally identical conditions to determine the uncertainties in the experimental results. Results reported herein represent the mean values for the two independent trials. Uncertainties are reported as the experimentally determined standard deviations. 2.3. Analytical chemistry The gas-phase products except C3 C5 alkanes and composition of bio-oils were analyzed as previously described [16]. For the analysis of C3 C5 alkanes, we used a 6-ft 2.1 mm 1/8-in. i.d. stainless steel column, packed with 80 100 mesh Porapak Q (Supleco) to separate each component in the mixture. For the GCMS analysis, we injected a 2 L volume of bio-oil dissolved in dichloromethane in a split ratio of 6:1. FT-IR analysis was performed as previously described [16], with the exception of using 4000800 cm1 scan range. Elemental analyses included carbon, hydrogen and nitrogen were performed on PerkinElmer Model 2400 Series II Autoanalyzer which utilized a technique based on a modication of the classical Pregl and Dumas methods. Sulfur analyses were performed on Carlo Erba 1108 Sulfur Autoanalyzer. The Dulong formula was used to estimate the HHV of the bio-oils on the basis of their elemental compositions. 3. Results and discussion This section provides information about the treated oil, gas products and hydrogen consumption during the catalytic hydrotreatment of crude algal oil. We rst discuss the inuence of the different reaction time and catalyst loading on the treated oil yields and composition, and then provide a detailed molecular characterization of a representative treated oil from the catalytic hydrotreatment. The nal sections present results from the gas-phase analysis and from overall mass and energy balances for the hydrotreatment feedstock and products. All catalytic hydrotreatment experiments were conducted at 400 C and 3.4 MPa high-pressure H2 . 3.1. Effect of reaction time and catalyst loading on treated oil yields The effect of reaction time on treated oil yields was determined by performing ve separate batch experiments with holding reaction times in the range of 18 h at a catalyst loading of 40%. For the catalyst loading effect, six separate batch experiments were conducted with holding catalyst intake on crude oil in a scale of 580% at a reaction time of 4 h. Both 7 h reaction time and 75% catalyst loading variations are all larger than our previous upgrading optimization study employed [17]. The results are presented in Fig. 1. Obviously, the treated oil yield increases in time when going from 1 h to 4 h of reaction time as shown in Fig. 1(a). The highest yield (79%) was obtained at a reaction time of 4 h. At longest reaction time of 8 h, the treated oil yield was lowered by about 15%. Therefore, from a yield perspective, a 4 h reaction time seems preferred. Similar treated oil yield as a function of reaction time was found in a previous study [20] which used a solvent free hydrotreatment to upgrade the pyrolysis oil at 350 C. In this study, the highest treated oil yield (65%) was also achieved at a reaction time of 4 h, however, this value is smaller than that of obtained from our experiment. Different type of catalyst, catalyst loading and reaction time might lead to different highest treated oil yield. For catalyst loading effect as shown in Fig. 1(b), similar scenario of treated oil yield with that of reaction time effect can be observed. Increasing catalyst intake on crude oil increases the treated oil yield when going from 5% to

20% of catalyst loading. The highest yield (83%) was obtained at a catalyst loading of 20%, and then, decreasing yield was presented when further increasing the catalyst loading. These decreases in treated oil yields were associated with an increasing formation of gas and coke products, which will be discussed later. It is important to note at this point that conditions producing a lower treated oil yield could be desirable if they also produced a bio-oil of higher quality (e.g., if catalytic hydrotreatment were to produce a lower yield of bio-oil because it had a lower oxygen content but higher heating value). Moreover, if the longer reaction time or higher catalyst intake on crude oil promoted hydrocracking reactions wherein large molecules were converted to smaller, more volatile ones, then one could anticipate lower oil yields because more of these light ends would have been lost during the solvent evaporation stage of treated oil recovery. In our previous study [16], closed treated oil yield (77%) was obtained with Pt/C as catalyst at a loading of 25% at 400 C and 4 h, suggesting that Pd/C (5%) and Pt/C (5%) seem to have similar catalytic activity toward the upgrading of algal bio-oil in SCW medium in spite of their different metal dispersion, specic surface area and micropore volume. 3.2. Effect of reaction time and catalyst loading on treated oil composition Table 1 compares the elemental compositions of the crude feed and treated bio-oils produced under different reaction times and catalyst loadings. The crude feed had an elemental composition of 74.3 wt% C; 9.7% H; 11.0% O; 4.5% N; and 0.6% S. This high heteroatom content must be reduced if the material is to be used as a transportation fuel. The HHV of the crude feed was estimated to be 37.1 MJ/kg, which is much higher than that of bio-crude obtained from the pyrolysis of microalgae [21], and is smaller than that of petroleum-derived crude (42 MJ/kg) [22]. This crude feed from the hydrothermal liquefaction of Nannochloropsis sp. is characterized in lower HHV and higher in oxygen content than that of bio-crude we reported in an early article [16]. The only difference in these two experiments was using different liquefaction environments (air vs H2 used previously). It seems that air environment leads to a crude bio-oil with lower HHV and higher oxygen content than that of produced under hydrogen environment. In Table 1, obviously, regardless of experimental conditions, the carbon and hydrogen content in treated oils are always higher than that of the crude feed, whereas the oxygen and nitrogen content were reduced to 6.973.08% and 3.981.99%, respectively. Therefore, the increased C and H levels and reduced O content lead to the treated oils having a higher energy density than the crude feed. The HHVs of the treated oils acquired at different reaction times and catalyst loadings are around 4144 MJ/kg. The highest HHV (43.8) was achieved at a catalyst loading of 80% and reaction time of 4 h, which is very close to that of diesel fuel (44.8 MJ/kg) [23]. No sulfur was detected in any of the treated oils, because supercritical water treatment alone (no catalyst) would also provide complete desulfurization of the crude bio-oil [16,17]. As shown in Table 1, catalyst loading is more inuential than reaction time on the H/C and O/C ratios of treated oils. The H/C ratio increases from 1.57 of crude oil to 1.79 of treated oil when catalyst loading increasing from 5% to 80%, indicating hydrogenation of components in the crude oil took place to a signicant extent. In contrast, different O/C ratio as a function of catalyst loading was observed. The O/C ratio rst decreases when going from 5% to 20% catalyst loading. The lowest O/C (0.028) ratio was obtained at a catalyst loading of 20%. Then, increasing the catalyst loading slightly increases the O/C ratio. One possible explanation for this result is that higher catalyst loadings lead to more light ends and coke formation, suggesting more carbon

P. Duan, P.E. Savage / Applied Catalysis B: Environmental 104 (2011) 136143

139

a
Yield(wt%)

90

b
Yield(wt%)

90

80

80

70

70

60 0 2 4 6 8 10

60 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

Time(h)

Catalyst loading(wt./wt.)

Fig. 1. Yields of treated oil from catalytic hydrotreatment at different reaction time (40% catalyst loading) and catalyst loading (4 h reaction time).

Table 1 Elemental composition (wt%) and heating value of crude and treated bio-oils (400 C, 3.4 MPa H2 ). Experimental conditions Crude bio-oil 400 C, 1 h, 40% 400 C, 2 h, 40% 400 C, 4 h, 40% 400 C, 6 h, 40% 400 C, 8 h, 40% 400 C, 4 h, 5% 400 C, 4 h, 10% 400 C, 4 h, 20% 400 C, 4 h, 60% 400 C, 4 h, 80%
a b

C 74.30 78.73 79.22 80.21 81.22 81.21 78.27 78.86 81.73 80.65 81.02

H 9.71 11.31 11.26 11.75 11.85 11.78 10.78 11.00 11.51 11.71 12.10

Oa 10.97 6.72 6.65 4.85 4.83 4.62 6.97 6.18 3.08 5.38 4.89

N 4.46 3.24 2.87 3.19 2.10 2.39 3.98 3.96 3.68 2.26 1.99

S 0.56 n.d.b n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

H/C 1.57 1.72 1.71 1.76 1.75 1.74 1.65 1.67 1.69 1.74 1.79

O/C 0.111 0.064 0.063 0.045 0.045 0.043 0.067 0.059 0.028 0.050 0.045

N/C 0.052 0.035 0.031 0.034 0.022 0.025 0.044 0.043 0.039 0.024 0.021

HHV (MJ/kg) 37.07 41.56 41.67 43.02 43.51 43.45 40.60 41.26 43.51 43.02 43.79

By difference. None detected.

was lost rather than stayed in the oil phase, thus correspondingly resulting in higher O/C ratio. In general, longer reaction times and higher catalyst loadings produced treated oils with higher HHV, H content and H/C ratio, and lower O/C and N/C ratios, respectively. These ndings indicate that deoxygenation and denitrogenation or hydrodeoxygenation and hydrodenitrogenation proceeded during the supercritical hydrotreatment of crude bio-oil. Thus, only from the higher HHV and H/C ratio perspectives, higher catalyst loadings are more effective than that of longer reaction times with keeping other variables at constant. An explanation for catalyst loading is more powerful than reaction time is that the catalyst deactivation increases with prolonging reaction time when keeping catalyst loading constant. All the properties examined are seemingly insensitive to the variables when reaction time longer than 4 h or catalyst loading higher than 20%, respectively, suggesting that the reaction system was already near the equilibrium. 3.3. Molecular characterization of treated bio-oil In this article, we provide some molecular details about the treated oil produced from crude algal oil processed with 80% Pd/C catalyst loading and 4 h reaction time, which serves as a representative product from the catalytic hydrotreatment. We show just one of the treated bio-oils because they all have similar GCMS spectrum, and the only difference is their relative proportion. 3.3.1. GCMS analysis The crude oil was dark-brown, viscous, sticky, tar-like and it possessed a smoky odor. In contrast, most of the treated oils were freely owing liquid products especially treated at longer reaction times and higher catalyst loadings. We used GCMS to separate and identify several of the molecular components in the crude and treated oils. The inlet temperature of the GC was 325 C, at this temperature about 60% of the crude oil sample was volatilised as the rest had a higher boiling point. This was estimated by perform-

ing a simulated distillation of the oil using vaporization test in a helium atmosphere. Using 80% catalyst loading and 4 h reaction time resulted in treated oil with approximately 80% of the overall oil mass volatilising after 325 C. Fig. 2 compares the total ion chromatograms of crude algal oil and treated oil produced at a reaction time of 4 h and catalyst loading of 80%. Both of them were dissolved in dichloromethane. The differences in the molecular compositions of these two oils are clearly evident in Fig. 2. The chromatogram (top) for the crude oil shows very little material eluting prior to 30 min, and it shows a large, broad signal around 35 min that corresponds to fatty acids. In contrast, the treated oil shows some large peaks at retention times shorter than 30 min and many regularly spaced peaks, which correspond to a series of n-alkanes starting at about C8 . Since this GC column separates largely on the basis of boiling point or volatility, it is clear that the catalytic hydrotreatment process has produced

Fig. 2. Total ion chromatograms of crude oil and treated oil.

140

P. Duan, P.E. Savage / Applied Catalysis B: Environmental 104 (2011) 136143

Fig. 3. Total ion chromatograms of treated oil (4 h, 80% catalyst loading).

treated oil with more low-boiling species. This physical change is also very apparent in the treated oil being a freely owing liquid whereas the crude starting material has more of a tarry consistency. More specically, the peak regions labeled a and b represent fatty acids (palmitoleic acid and palmitic acid) and alkane (pentadecane), respectively (top chromatogram). Under high-pressure H2 , adding the Pd/C catalyst, which has good activity for deoxygenation of fatty acids in HTW [8], leads to treated oil with very tiny peak heights in the fatty acid regions. This result is consistent with the Pd/C catalyst being effective for deoxygenating fatty acids as these molecules were produced during hydrothermal liquefaction. Converting fatty acids to alkanes improves the storage stability of the fuel and also increases its energy density. For a better understanding of the property of the treated oil as shown in Fig. 2, a mass spectral library and computer matching were used to facilitate compound identication. Fig. 3 shows the tentative identities of the major (peak area at least 0.7% of the total) individual molecular components in the treated oil. To display all peaks clearly, the diagrams of total ion chromatogram in the retention time range of 3137 min was enlarged and shown separately (bottom chromatogram). It should be noted that the treated oil likely contained additional compounds that did not appear in the chromatogram. Lighter products were lost during the solvent evaporation and those remaining would go undetected because the solvent delay. At the other extreme, very heavy compounds were not likely to elute from the GC column and be detected because of their high boiling point as mentioned above. For the crude bio-oil, GCMS analysis suggested that its major components include fatty acids, long-chain hydrocarbons

(unsaturated and saturated), naphthalene, pyrrolidine derivatives, piperidine derivatives, pyrrolidinone derivatives, cholestane, and cholestene, amides, and other N-containing compounds. In contrast, the treated oil was primarily a mixture of alkanes and aromatic compounds, with fatty acids in lower proportions than in the crude oil as compared in Fig. 2. Pentadecane was the most abundant alkane, but others ranging from C8 to C32 are also present in the treated oil. A likely pathway for the formation of pentadecane is the decarboxylation of palmitic acid. We did not detect appreciable amounts of alkanes smaller than C15 in the crude oil, so these lighter molecules may have been formed from cracking of the longer chain alkanes during the supercritical hydrotreatment process. The aromatic compounds were primarily benzene derivatives. Small amount of unsaturated compounds were detected. It is likely that they were hydrogenated due to the presence of H2 and catalyst. Derivatives of piperidine, pyrrolidinone, and O-methyloxime, which were present in the crude oil, were not detected in the treated oil. Unfortunately, some new nitrogen-containing compounds such as palmitamide and pentadecanenitrile appeared. A possible pathway for the formation of these amides and nitriles is the reaction of fatty acids with ammonia [24]. GCMS analysis suggested that ammonia existed in the aqueous phase. The overall content of cholesterol derivatives fell to zero in the treated oil. No sulfur compounds were detected in the treated bio-oil, which is consistent with the elemental analysis. The net effects for the catalytic hydrotreatment process on the molecular composition of the bio-oils are that the amount of alkanes and aromatics increased, oxygenated- and nitrogen-containing compounds decreased, and sulfur compounds were removed.

P. Duan, P.E. Savage / Applied Catalysis B: Environmental 104 (2011) 136143

141

Fig. 4. FT-IR spectrum of crude oil and treated oil (4 h, 80% catalyst loading).

3.3.2. FT-IR analysis Fig. 4 shows the FT-IR spectra for the crude and treated oil. We show just one (produced at 4 h, 80% catalyst loading) of the treated oils because they all have similar FT-IR spectra, which suggests that the same types of functional groups exist in each. Both the crude and treated oil display strong absorbance between 2850 and 3000 cm1 , indicating a high content of CH3 and CH2 groups. This result is consistent with the presence of a collection of fatty acids and alkanes in the total ion chromatogram. Absorbance peaks between 1650 and 1760 cm1 represent the C O group stretching vibration in carboxylic acids. The treated oil shows a proportionately lower absorbance in this wave number range, which is consistent with their reduced fatty acids content. In liquid or solid state, carboxylic acids exist as dimers due to strong hydrogen bonding. A free hydroxyl stretching vibration (near 3520 cm1 ) can be observed only in very dilute solution in nonpolar solvents or in vapor phase. We used the liquid-like oil to get the FT-IR spectrum, so the hydroxyl stretching vibration absorption is expected to be very weak or even absent [25]. 3.4. Effect of reaction time and catalyst loading on gas composition and yields Gases formed along with the treated oil during the hydrotreatment processing at different reaction time and catalyst loading. For all experiments tested, the most abundant components were unreacted H2 , CO2 , CH4 , C2 H6 , C3 H8 and C4 H10 . The rich un-reacted H2 presented in the nal reaction gas components indicated that the reactions were not performed under hydrogen starvation conditions. Fig. 5 shows the main gas composition as a function of reaction time and catalyst loading, respectively. The mole fractions shown were calculated on a H2 -free basis, so they reect the relative amounts of the different gases that are solely products of hydrotreatment processing. H2 is also a potential product, but it was present at the start of the reaction, so it would also be a potential reactant. It is clear in Fig. 5(a) and (b) that reaction time would be more inuential than catalyst loading on the gas composition. In addition to the ve main gaseous products presented in Fig. 5(a) and (b), CO and C5 H12 were also formed in low mole fraction (less than 1%) in all experiments. Analysis of CO fraction as a function of reaction time and catalyst loading suggested that shorter reaction times and lower catalyst loadings led to a little bit higher of CO fraction in gas phase, which is consistent with previous work [26,27] on biomass gasication in SCW that the CO yields were always the highest at the early stages of the reaction, and that the CO reacted away as time progressed. No ethylene was detected in all experiments. Presumably, if it formed, it was hydrogenated to ethane. A likely pathway for the formation of some of the CO2 is decarboxylation of fatty acids. CO2 can also be possibly formed

from reactions such as steam reforming and watergas shift. Lighter molecules in the gas phase may have been formed from cracking of the longer chain alkanes. No molecular nitrogen was detected in any of the experiments. Nitrogen atoms removed from the crude oil probably exist in the form of ammonia. Indeed, ammonia peak appeared in the gas chromatogram when using Porapak Q column to separate the gas components. The pH of aqueous phase after the hydrotreatment process revealed that the aqueous phase was basic, and ammonia was evidenced from the GCMS analysis of the aqueous phase as mentioned in Section 3.3.1. Fig. 6 shows the yield of gas of those main components as function of reaction time and catalyst loading. The total gas yield ranged from about 1.13 to1.90 mmol/g with reaction time varied from 1 h to 8 h as shown in Fig. 6(a). In contrast, the catalyst loading had less effect on the total gas yield which has a range of 1.471.72 mmol/g as shown in Fig. 6(b). Our previous study [17] suggested that kinetics control was more prevailed at longer reaction times over higher catalyst loadings. So variation of reaction time is more inuential on the products selectivity and yield than catalyst loading. Thus, total gas yield might be reduced by shortening the reaction time. It should be noted that the production of large amount of gaseous products is undesirable, because these products are less valuable than liquid products. CO and C5 H12 were formed in low yields in all experiments. Experiments at different reaction time and catalyst loading led to modest CO (0.024 mmol/g) and C5 H12 production (0.025 mmol/g). The highest CO yield was only observed at 1 h reaction time and 5% catalyst loading. 3.5. Effect of reaction time and catalyst loading on hydrogen consumption Hydrogen consumption during the catalytic hydrotreatment process will be a major variable cost contributor. Analysis of the gas-phase post-reaction permits calculation of the moles of consumed H2 . Hydrogen consumption was greater than 60% under all of the experimental conditions, which is very close to that of our previous study [16]. Extension of reaction time led to slightly higher hydrogen uptake which ranged from 4.8 mmol/g to 6.2 mmol/g as depicted in Fig. 7(a), and thus improved the product oil quality. This hydrogen uptake as a function of reaction time is consistent with that of hydrogen content in treated oils as suggested in Table 1. For the catalyst loading effect, however, contrary H2 uptake trend ranged from 6.0 mmol/g to 4.8 mmol/g was observed with increasing catalyst loading as presented in Fig. 7(b). In a close reaction system, hydrogen amount in the nal reaction gas phase depends on two opposite effects: hydrogen consumption, which reduces the nal H2 amount, and H2 production, which tends to increase the nal H2 amount. This decrease in H2 consumption with increasing the catalyst intake indicates that H2 production amount was bigger than its consumption. However, the hydrogen content in treated oil increases with increasing the catalyst intake, suggesting that hydrogen was produced from other hydrogen-containing sources. The only hydrogen source in the reaction system is water. Thus, one possible explanation for this hydrogen consumption decrease is that higher catalyst loadings promoted some reactions (such as watergas shift reaction) to generate in situ H2 . This in situ generated H2 was partially consumed instead of the initial charged molecular H2 so that resulted in more H2 left after the reaction. We performed one experiment without the presence of external H2 with keeping other variables at constant. The results suggested that the hydrogen content in treated oil was also higher than that of crude feed, suggesting that watergas shift reaction may contribute to the in situ generation of hydrogen. Our previous work [16] also proved this possibility. We next conducted the other experiment with the same catalyst and same conditions used under inert con-

142

P. Duan, P.E. Savage / Applied Catalysis B: Environmental 104 (2011) 136143

Fig. 5. Inuence of reaction time (40% catalyst loading) and catalyst loading on gas composition (4 h reaction time).

Fig. 6. Inuence of reaction time (40% catalyst loading) and catalyst loading on gas yield (4 h reaction time).

dition (He), but we used the solid residue rather than crude bio-oil. No H2 was present in this experiment, which demonstrates that potential H2 production from coke product can be ruled out. 3.6. Mass and energy balance For those experiments tested, we quantied the amount of products in the treated oil, gas phase, and the non-catalyst solid residue (coke). Knowing the elemental compositions of treated oils and the yields of gases and coke products allows us to calculate the overall C, H atoms and energy recoveries from these experiments. Since the coke product alone cannot be recovered and its yield is less than 20% in all experiments. We make an approximation that the carbon in coke product is roughly equal to the same amount of pure carbon. Table 2 gives the ratios of C, H, and chemical energy (HHV) in the treated oils produced at different reaction time and catalyst loading relative to those quantities in the crude feed, respectively. A ratio of unity denotes recovery of the same number of atoms or amount of chemical energy in the products that was originally present in the feedstock. The C and H recovery in the treated oil increased to 0.90 and 0.98 of the original C and H content in crude oil after 2 h and 4 h reaction

time, respectively, and then decreased. Similar energy recovery as a function of reaction time was also observed. From a carbon and energy perspective, 2 h reaction seems preferred. The variations of C, H and energy recovery with catalyst loading when going from 5% to 20% are similar to that of extending reaction time from 1 h to 2 h. All the highest C, H and energy recoveries were achieved at a catalyst loading of 20%. As catalyst loading continued to increase, the C, H and energy recoveries decreased. Thus, 20% catalyst loading seems the most suitable value to be selected for the hydrotreatment of crude algal oil. Interestingly, comparison of the recoveries of C and H in Table 2 indicates that the ratio of H in the treated oil to H in the crude feed was higher than that of C, suggesting that higher H recovery beneted from the presence of high-pressure H2 or the presence of SCW in the reactor. Thus, it appears that adding high-pressure H2 together with H2 O did signicantly improve the hydrogen uptake and guarantee more H atoms appearing in the treated oil products. Of course, the elemental analysis showed that the oils produced under the high-pressure H2 environment did have a higher H content. Table 2 also shows the atom and energy recoveries with the gaseous and coke products included. The atom and energy recoveries when including gaseous and coke products are higher than

Fig. 7. Inuence of reaction time (40% catalyst loading) and catalyst loading on H2 consumption (4 h reaction time).

P. Duan, P.E. Savage / Applied Catalysis B: Environmental 104 (2011) 136143 Table 2 Atom and energy balances for hydrotreatment of crude feedstock in SCW. Experimental conditions 40% Pd/C, 1 h 40% Pd/C, 2 h 40% Pd/C, 4 h 40% Pd/C, 6 h 40% Pd/C, 8 h 5% Pd/C, 4 h 10% Pd/C, 4 h 20% Pd/C, 4 h 40% Pd/C, 4 h 60% Pd/C, 4 h 80% Pd/C, 4 h
CUBO CCBO CUBO +Cgas CCBO CUBO +Cgas +Ccoke CCBO HUBO HCBO Hoil +Hgas Halgae +H2 feed EUBO ECBO

143

EUBO +Egas ECBO +EH feed

2

0.73 0.90 0.87 0.71 0.73 0.78 0.86 0.92 0.87 0.75 0.72

0.75 0.94 0.92 0.77 0.78 0.83 0.90 0.96 0.91 0.79 0.77

0.93 1.08 1.04 0.91 0.93 0.89 0.98 1.06 1.03 0.93 0.96

0.80 0.97 0.98 0.80 0.81 0.82 0.92 0.99 0.89 0.83 0.82

0.84 0.99 1.00 0.85 0.86 0.87 0.95 1.03 1.00 0.88 0.89

0.78 0.95 0.94 0.77 0.79 0.81 0.90 0.98 0.94 0.81 0.80

0.78 0.95 0.94 0.77 0.80 0.81 0.90 1.00 0.94 0.81 0.81

those for the treated oils alone. A considerable amount of carbon is transferred to gaseous and coke products with increasing the reaction time and catalyst loading. Again, longer reaction times and higher catalyst loadings promoted more side products formation, thus, both the C and H recoveries are all higher when including the gas and coke products. The values of the overall C recovery in third column in Table 2 exceed 100% by only a modest amount, and this amount could also be attributed to systematic errors. The fth column in Table 2 shows the ratio of hydrogen atoms in the treated oil and gas phase (including those in the unreacted H2 ) to hydrogen atoms in crude feed and H2 initially charged to the reactor. This ratio exceeds unity for some experiments. This result is consistent with water molecules being active participants in the chemistry as mentioned above, and H atoms from water being incorporated into the oil and gas products. 4. Conclusions Hydrotreating the crude algal bio-oil in the presence of Pd/C and with high-pressure H2 in SCW at different reaction time and catalyst loading effected several desirable changes in the treated oil. The O/C and N/C atomic ratios decreased whereas the C, H content and HHV increased. Longer reaction times and higher catalyst loadings favor the production of treated oils with higher HHV, H content and H/C ratio, and lower O/C and N/C ratios, but at the same time promoted more side products formation such as gas and coke. Catalytic hydrotreatment led to a low viscosity liquid product that ows easily. All of the hydrotreatment explored in this work showed that the sulfur content of the product oil was reduced below detection limits. The treated bio-oil with the best properties was obtained from processing with Pd/C and 3.4 MPa H2 at a reaction time of 4 h and 80% catalyst loading. The oil has a heating value of 44 MJ/kg, which is comparable to that of diesel fuel (44.8 MJ/kg). An increasing formation of gas and coke products was observed at longer reaction times and catalyst loadings. Finally, the GCMS and FT-IR analysis suggest that the treated oil has a very high content of hydrocarbon molecules including alkanes and a suite of aromatic compounds. Given these encouraging results, we view this catalytic hydrotreatment process in supercritical water in the presence of a Pd/C catalyst as an effective means to improve the properties of crude bio-oil from the hydrothermal liquefaction of microalgae.

Acknowledgements We gratefully acknowledge the nancial support from the Ph.D. Programs Foundation of Henan Polytechnic University (B2011008). The U.S. National Science Foundation provided partial support for this project through grant (EFRI-0937992). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.apcatb.2011.02.020. References
[1] A.A. Peterson, F. Vogel, R.P. Lachance, M. Frling, M.J. Antal Jr., J.W. Tester, Energy Environ. Sci. 1 (2008) 3265. [2] P.E. Savage, R.B. Levine, C.M. Huelsman, in: M. Crocker (Ed.), Hydrothermal processing of biomass, Royal Soc. Chem. Pub., 2010, pp. 190219 (chapter 8). [3] S. Amin, Energy Convers. Manage. 50 (2009) 18341840. [4] L. Brennan, P. Owende, Renew. Sustain. Energy Rev. 14 (2010) 557577. [5] J.A.V. Costa, M.G. de Morais, Bioresour. Technol. 102 (2011) 29. [6] ASTM, ASTM Int. D6751-07a (2007) 18. [7] T. Brown, P. Duan, P.E. Savage, Energy Fuels 24 (2010) 36393646. [8] P. Duan, P.E. Savage, Ind. Eng. Chem. Res. 50 (2011) 5261. [9] J. Wildschut, I. Melin-Cabrera, H.J. Heeres, Appl. Catal. B: Environ. 99 (2010) 298306. [10] N.H. Tran, J.R. Bartlett, G.S.K. Kannangara, A.S. Milev, H. Volk, M.A. Wilson, Fuel 89 (2010) 265274. [11] M.C. Clark, B. Subramaniam, AIChE J. 45 (1999) 15591565. [12] Z.-M. Cheng, Y. Ding, L.-Q. Zhao, P.-Q. Yuan, W.-K. Yuan, Energy Fuels 23 (2009) 31783183. [13] L. Han, R. Zhang, J. Bi, J. Fuel. Chem. Technol. 36 (2008) 15. [14] T. Sato, T. Adschiri, K. Arai, G.L. Rempel, F.T.T. Ng, Fuel 82 (2003) 12311239. [15] M. Watanabe, S. Kato, S. Ishizeki, H. Inomata, R.L. Smith Jr., J. Supercrit. Fluids 53 (2010) 4852. [16] P. Duan, P.E. Savage, Bioresour. Technol. 102 (2011) 18991906. [17] P. Duan, P.E. Savage, Energy Environ. Sci (2011), doi:10.1039/C0EE00343C. [18] M. Snre, I. Kubi ckov, P. Mki-Arvela, K. Ernen, D.Y. Murzin, Ind. Eng. Chem. Res. 45 (2006) 57085715. [19] B. Zhang, M. von Keitz, K. Valentas, Appl. Biochem. Biotechnol. 147 (2008) 143150. [20] J. Wildschut, M. Iqbal, F.H. Mahfud, I.M. Cabrera, R.H. Venderbosch, H.J. Heeres, Energy Environ. Sci. 3 (2010) 962970. [21] X. Miao, Q. Wu, C. Yang, J. Anal. Appl. Pyrolysis 71 (2004) 855863. [22] A. Demirbas, Energy Policy 35 (2007) 46614670. [23] S. Sinha, A.K. Agarwal, S. Garg, Energy Convers. Manage. 49 (2008) 12481257. [24] P.-G. Duan, S. Li, Z.-Z. Wang, L.-Y. Dai, Chem. Eng. Res. Des. 88 (2010) 10671072. [25] R.M. Silverstein, F.X. Webster, D.J. Kiemle, Spectrometric Identication of Organic Compounds, 7th edition, Wiley, 2005. [26] F.L.P. Resende, S.A. Fraley, M.J. Berger, P.E. Savage, Energy Fuels 22 (2008) 13281334. [27] F.L.P. Resende, P.E. Savage, Energy Fuels 23 (2009) 62136221.