Fuel Processing Technology 123 (2014) 11–18

Contents lists available at ScienceDirect

Fuel Processing Technology

j ourna l homepage: www.e lsev ie r .com/ locate / fuproc

Hydrodeoxygenation of fast-pyrolysis bio-oils from various feedstocksusing carbon-supported catalysts☆

Yaseen Elkasabi a,⁎, Charles A. Mullen a, Anna L.M.T. Pighinelli a,b, Akwasi A. Boateng a

a Eastern Regional Research Center, Agricultural Research Service, U.S. Department of Agriculture, 600 E. Mermaid Lane, Wyndmoor, PA 19038, United Statesb Brazilian Agricultural Research Corporation, Embrapa Agroenergy, Parque Estacao Biologica — PqEB s/n°— Av. W3 Norte (final), Asa Norte, Brasilia, DF 70770-901, Brazil

☆ Mention of trade names or commercial products in tpurpose of providing specific information and does notdorsement by the U.S. Department of Agriculture. USDA iand employer.⁎ Corresponding author. Tel.: +1 215 836 3797; fax: +

E-mail address: yaseen.elkasabi@ars.usda.gov (Y. Elka

0378-3820/$ – see front matter. Published by Elsevier B.Vhttp://dx.doi.org/10.1016/j.fuproc.2014.01.039

a b s t r a c t

a r t i c l e i n f o

Article history:Received 25 September 2013Received in revised form 31 January 2014Accepted 31 January 2014Available online 19 February 2014

Keywords:BiofuelsPyrolysisBio-oilHydrotreatingFeedstock

In this paper,we sought to elucidate the relationships between biomass feedstock type and the suitability of theirfast-pyrolysis bio-oils for hydrodeoxygenation (HDO) upgrading. Switchgrass, Eucalyptus benthamii, and equinemanure feedstocks were pyrolyzed into bio-oil using a continuous fast-pyrolysis system. We also synthesizedvariations of switchgrass bio-oil using catalytic pyrolysis methods (HZSM-5 catalyst or tail-gas recycle method).Bio-oil samples underwent batch HDO reactions at 320 °C under ~2100 psi H2 atmosphere for 4 h, using Pt, Ru, orPd on carbon supports. Hydrogen consumption was measured and correlated with compositional trends. Theresulting organic, aqueous, and gas phaseswere analyzed for their chemical compositions.Mass balances indicatelittle coke formation. Switchgrass bio-oil over Pt/C performed the best in terms of hydrogen consumption effi-ciency, deoxygenation efficiency, and types of upgraded bio-oil compounds. Eucalyptus feedstocks consistentlyconsumed more than twice the normal amount of hydrogen gas per run, primarily due to the elevated syringolcontent. Catalytically pyrolyzed bio-oils deoxygenated poorly over Pt/C but hydrogenated more extensivelythanother oils. Although the relative deoxygenation (%DOrel) varied based on feedstock and catalyst, the absolutedeoxygenation (%DOabs) depended only on the overall yield. The total extent of upgrading (hydrogenation+ de-oxygenation) remained independent of feedstock and catalyst.

Published by Elsevier B.V.

1. Introduction

As the global rate of energy consumption increases, the need for sus-tainable renewable fuels escalates. The global supply of fossil fuels can-not meet the demand that is set by developing countries [1]. Hence,alternative and efficient energy sources are needed to mitigate theshortage. Although energy sources such as wind and solar are environ-mentally friendly, they still are impractical for many user applications;they also cannot replace the abundance of commodity chemicals andhigh-demand coproducts that emanate from fossil fuels. Furthermore,while corn grain ethanol has proven to be an effective alternative fuel[2], the U.S. Energy Independence and Security Act (EISA) sets a restric-tion on themaximum corn starch-based fuel obtained by the year 2022.Of all themethods investigated, thermochemical conversion of biomassshows the most promise for scalability, efficiency, reduction of green-house gases, and production of coproducts. Methods such as liquefac-tion [3,4], pyrolysis [5,2], and gasification [6] all fall under this category.

his publication is solely for theimply recommendation or en-s an equal opportunity provider

1 215 233 6559.sabi).

.

In particular, fast pyrolysis of biomass is highly effective at breakingdown macromolecular structures into smaller organic compounds,the mixture of which has an appearance resembling liquid crude oil(i.e. “bio-oil”). Pyrolysis is one of the very fewprocesses that is inherent-ly carbon-negative [7], if all of its products are utilized to their advan-tage. Depending on the biomass feedstock used, fast pyrolysis bio-oilcomposition can drastically vary. Woody [8] and proteinaceous [9] bio-masses, grasses [10], and animal waste [11] have been pyrolyzed intocrude bio-oil. This apparent process neutrality with regard to feedstockmakes fast pyrolysis easily translatable to a wide variety of agriculturalclimates and economies. Bio-oils from lignocellulosic biomasses yieldhighly aromaticmixtures, which carry greater potential for fuel efficien-cy. Due to the abundance of acid, high viscosity, and heteroatom con-taminants, bio-oil must be catalytically upgraded and hydrotreated inorder to serve as a fungible fuel. Hydrodeoxygenation (HDO) reactionsare employed in order to increase the H:C and decrease the O:C atomis-tic ratios. Supported Ru, Pt, and Pd are catalysts which are commonlyemployed for this purpose. Elliot et. al. [12] demonstrated a two-stagecontinuous HDO configuration for the upgrading of pine bio-oil.Mercader and coworkers extracted the acids from bio-oil before HDO,to examine the efficacy on catalytic upgrading [13]. Several other stud-ies show ruthenium catalysts to be effective for deoxygenation [14,15].Despite the abundance of HDO studies [16], very few have gone beyondthe upgrading of wood-based bio-oil [17]. Since feedstocks have a largeeffect on the bio-oil quality, so too theywould affect theHDOefficacy. In

12 Y. Elkasabi et al. / Fuel Processing Technology 123 (2014) 11–18

order for fast pyrolysis to become a viable fuel source, both the pyrolysisand upgrading steps should translate well to different feedstocks andhence, to different global economies. A thorough investigation is neededin order to elucidate the relationship between biomass feedstock andupgradability.

Also, certain process conditions affect the interactions of bio-oilcomponents with the catalyst sites. For example, Subramanian and co-workers [18,19] proposed a shift in catalytic pathways by changingthe reaction solvent for carbon-supported catalysts. The benefits ofaqueous solvent effects have yet to be realized in situ for HDO of bio-oil, in contrast to model compounds. Catalytic pyrolysis processes alsohave a strong effect on the upgradability of bio-oil. Our goal was toelucidate the effect of various feedstocks and feedstock processing onHDO, both with regard to efficiency and quality of upgraded product.The biomasses we investigated cover a range of biomass types, underprocessing conditions that allow for greater contact between the bio-oil components and catalyst sites.

2. Materials and methods

2.1. Bio-oil production

All feedstockswere ground anddried prior to processing. Switchgrassfeedstock was obtained from the McDonnell Farm (East Greenville, PA,USA), horsemanure from the Equine Rehabilitation Center at MorrisvilleState College (Morrisville, NY, USA), and Eucalyptus benthamii fromEmbrapa Forestry (Candoi, Parana, Brazil). Feedstock fast-pyrolysis wascarried out in the ERRC fluidized bed fast pyrolysis system, as describedpreviously [10]. Briefly, 2 kg/h of feedstock is fed through a fluidizedsand bed reactor under nitrogen at 500 °C. Catalytic pyrolysis was simi-larly carried out by replacing the sand bedwithHZSM-5 catalyst (ZeolystInternational, USA). The resulting vapors then pass through a cyclonewhich separates out solid char particles. Subsequently, liquid phasesare condensed from the vapor by 4 condensers in series. Then the oilphase is precipitated from the non-condensable gases (NCGs) by twoelectrostatic precipitators (ESPs) in series. All experiments used the oilobtained from the ESPs. For recycle experiments [20], a fraction of theNCG stream was mixed with the N2 stream and recycled into the fluid-ized bed, using a preheater and gas blower.

2.2. Hydrodeoxygenation

Batch hydrodeoxygenation reaction experimentswere carried out ina Parr Series 4598 100 mL bench-top reactor. A Parr 4848B controllerwas used to monitor the vessel pressure and to control the internaltemperature and impeller mixing speed. The reactor can withstand amaximum pressure of 5000 psi at 500 °C. Hydrogen was suppliedfroma reservoir tank via a pressure regulator. A constant pressure of hy-drogenwas delivered from the regulator to the reactor through a prima-ry valve, which remained open for the duration of the experiment.Hydrogen consumption over timewasmeasured bymonitoring the res-ervoir pressure drop, while taking into account gas production fromHDO reactions. To ensure adequate gas distribution, a dip-tube intro-duced hydrogen internally from the primary valve to the reactor vessel'sbottom. The reactor was modified to accommodate a manifold appara-tus for injecting liquid samples into the reactor through a secondaryvalve, all while the reactor is pressurized. Liquid flow through the man-ifold is controlled with amodified HPLC pump (Scientific Systems, Inc.),and the manifold piping was continuously warmed with heating tape.

All HDO catalysts (Johnson Matthey, Inc., UK) consisted of a powdercharcoal support with 5% metal loading (Pt, Ru, or Pd). For all HDO ex-periments, 0.5 g catalyst and 40 mL deionized water were first loadedinto the reactor vessel. The vessel was then clamped shut and flushedwith hydrogen 3 times to displace air. After charging the vessel with1100 psi hydrogen, the water/catalyst slurry was lightly stirred at300 rpm while a furnace heated the vessel to an internal temperature

of 320 °C. The hydrogen reservoir pressure regulator was then in-creased to 2100 psi. After providing sufficient time for catalyst reduc-tion, ~9 g of bio-oil was injected into the reactor, and the mixer speedwas increased to 1200 rpm. After 4 h, the primary hydrogen valve wasclosed off, the heater was removed, and themixer speed was decreasedto 300 rpm. After gradually cooling the reactor and halting the mixer,non-condensable gases were vented into a collection bag and furtheranalyzed. The aqueous phase was decanted from the vessel and centri-fuged to remove any partially miscible components. The vessel and thereactor parts were washed with acetone to collect any accumulated oil.The acetone washings were filtered through a 0.45 μm PTFE filter to re-move the catalyst, and the acetonewas removed by rotary evaporation.Experiments were performed in duplicate.

2.3. Characterization

Elemental analysis (CHNS) was conducted via a Thermo EA1112CHNS analyzer. Oxygen content was calculated by difference, andwater content was subtracted. Moisture content was measured withKarl-Fischer titration in methanol with Hydranal Karl-Fischer Compos-ite 5 (Fluka) as the titrant. Total acid number (TAN) was measuredusing a Mettler T70 autotitrator using 0.1 M KOH in isopropanol as thetitrant and wet ethanol as the titration solvent. Gas chromatographywith mass spectroscopy (GC–MS) analysis of liquid products wasperformed on a Shimadzu GCMS QC-2010. The column used was aDB-1701, 60 m × 0.25 mm, 0.25 μm film thickness. The oven tempera-ture was programmed to hold at 45 °C for 4 min, ramp at 3 °C/min to280 °C and hold at 280 °C for 20 min. The injector temperature was250 °C, and the injector split ratio set to 30:1. Helium carrier gas flowedat 1 mL/min. GC–MS analysis of gaseous products was performed on anAgilent 6890N gas chromatograph equipped with an Agilent 5973mass selective detector. Analyses were performed with a fused silicacapillary column, CP-PoraBOND Q, 25 m × 0.25 mm (Varian, Palo Alto,CA). The programmed method held the temperature at 35 °C for3 min, then increased to 150 °C at 5 °C/min, then increased to 250 °Cat 10 °C/min. The final temperaturewas held for 30min. Gas concentra-tionswere calculated relative to calibration curves of a gasmixture stan-dard: CO, CH4, CO2, C2H4, C2H6, C3H6, C3H8, and C4H10 mixed in helium(Scott Specialty Gases, Plumsteadville, PA).

3. Results and discussion

3.1. Bio-oil composition

The experimental study was restricted to three types of biomassfeedstock. We examined switchgrass (“SwG”), E. benthamii, (“Eucal”)and equine manure (“Manu”) as representatives of herbaceous grasses,woody biomass, and animal waste, respectively. These particular feed-stockswere selected for their abundance, importance to U.S. agriculture,and fundamental differences in feedstock characteristics. Alongsidenormal switchgrass bio-oil, two variations in partially deoxygenatedswitchgrass bio-oil were produced: 1) switchgrass catalytically pyro-lyzed with HZSM-5 in situ (“cat-SwG”) and 2) switchgrass pyrolyzedunder an atmosphere produced by recycling the exhaust gas into thepyrolysis fluidized bed (“SwG-rec”). For HDO reactions, a moderate cat-alyst loading ratio was used in order to eliminate effects of catalystoverloading, such as excessive coking or cracking. Using water as a sol-vent for carbon-supported catalysts produces three beneficial effects:1) highly acidic and other low-value oxygenated components transferto the aqueous phase 2) beneficial components which are partially mis-cible in water are not excluded from HDO and 3) for particularly strongcatalysts, water regulates the catalyst site interactionswith bio-oil com-ponents [19]. Although some higher activation energies are satisfied byhigher reaction temperatures (N320 °C), a consequently higher systempressure is necessary in order that water remains as a compressedliquid. Hence, one limitation of using water as a solvent is that higher

Table 1Mass balance closures for HDO experiments.

Weight (g) Bio-oil H2 Gas pdt Organic Aqueous % Closure

SwG-Ru 9.08 0.17 1.45 3.56 2.11 76.9SwG-Pd 8.37 0.17 2.53 3.90 1.73 95.5SwG-Pt 9.53 0.17 1.99 3.45 1.96 76.4Manu-Pt 10.25 0.16 1.00 6.42 1.19 82.7Manu-Ru 8.85 0.12 0.85 5.17 1.46 83.4Eucal-Pt 8.82 0.27 3.76 3.86 0.84 93.2Eucal-Ru 8.67 0.25 2.76 4.19 0.42 82.6cat-SwG-Pt 9.25 0.15 1.35 5.93 1.43 92.6SwG-rec-Pt 9.03 0.15 1.27 5.46 1.09 85.3

13Y. Elkasabi et al. / Fuel Processing Technology 123 (2014) 11–18

temperatures quickly become expensive and hazardous, consideringthe maximum pressure limitations of the reactor system used. Anotherreason for using the selected reaction temperature was to prevent ex-cess cracking of hydrocarbons at higher temperatures [21,16].

To assess the efficiency of the reaction system, both the organicand aqueous phases post-HDO were weighed and analyzed for theirchemical compositions and moisture content (Table 1). Mass balanceclosures ranged from 75 to 95%, with closures typically above 90%. Pos-sible sources of mass loss include: escape of volatile liquids with thenon-condensable gases, entrapment of some organic compounds andcoke during solids filtration, and loss of volatile components during ro-tary evaporation of acetone washings. Transfer of organic compounds(mainly organic acids and some phenol) to the aqueous phase playeda relatively significant role, since the aqueous volume was large com-pared to the bio-oil. Smaller organic yields are associated with more ef-ficient deoxygenation. This is because a greater deoxygenation mustdecrease the mass yield.

The most direct way to evaluate HDO performance is through ele-mental analysis (CHNS-O). Through elemental analysis, one obtainsthe H:C and O:C elemental ratios. High H:C and low O:C ratios are criti-cal for integration of upgraded pyrolysis bio-oils into liquid (hydrocar-bon) fuels. A high H:C ratio ensures combustibility, whereas a low O:Cratio eliminates water and acidic components, both of which reducethe energy content and harm downstream equipment. Elimination ofnitrogen heteroatoms also reduces concern about NOx emissions uponcombustion of the products. Table 2 shows the characterization of bio-oil HDO over Pt and Ru, with respect to the three different feedstocks.HDO of switchgrass bio-oil over Pt showed the greatest reduction inwt.% organic oxygen. Consequently, this combination produces themost hydrophobic oil phase product leading to the greatest reductionin dissolved water. The total acid number (TAN) indicates the amountof Brønsted acids present in the sample, on an absolute basis. Hence,the TAN indicates the corrosiveness of the sample. While some smalldifferences exist in the final TAN values, the greatest reduction in TANnumber came from HDO of manure bio-oil over Ru, where the TANwas reduced from 97 mg KOH/g to 32 mg KOH/g. Table 3 displays theelemental results for HDO of switchgrass with Pd/C, as well as Pt/CHDO of catalytically-pyrolyzed bio-oils. One trend that can be seen isthe difficulty of deoxygenating bio-oils with significantly less oxygen

Table 2Elemental analyses (wt.% dry basis), moisture, and acidity of bio-oils from different feedstocks

Switchgrass Eu

Bio-oil Pt Ru Bi

N 0.52 1.08 0.87 0C 53.81 75.40 71.64 59H 5.36 8.48 7.99 5O 33.21 15.04 19.50 34H/C (mol) 1.196 1.349 1.339 1O/C (mol) 0.463 0.150 0.204 0Karl-Fischer moisture (wt.%) 7.98 2.10 2.245 6Total acid number (mg KOH/g) 73 35 44 70

pre-HDO (catalytic/recycle pyrolysis) than feedstocks with significantlyhigher oxygen content (switchgrass, Eucalyptus). Normal pyrolysis bio-oil contains oxygen in the formof various compoundswhere the oxygencleavage is easier, such as aldehydes, alcohols, and acyl derivatives. Cat-alytic pyrolysis oils contain more phenolics and other compounds withoxygen directly bonded to a benzene ring, which require higher activa-tion energies to break [22]. In addition, TAN values of catalytic pyrolysisbio-oils were not significantly reduced, if at all. However, the viscositiesof eucalyptus and HZSM5-catalyzed switchgrass bio-oils were signifi-cantly reduced from that of the original bio-oil, indicating some break-down of higher molecular-weight lignin structures.

To accurately visualize and extrapolate trends from elemental anal-ysis, the elemental composition was plotted in Van Krevelen-type dia-grams (Fig. 1). The coordinates for commercial petroleum are addedfor comparison. While both switchgrass and eucalyptus bio-oils pos-sessed very similar elemental ratios, as indicated by their startingmate-rial coordinates, the HDO for all three feedstocks (including manure)traversed the exact same path towards the direction of the petroleumcoordinate. These results indicate that, for this set of unprocessed bio-oils, the HDO of normal pyrolysis oils depends primarily on the initialelemental composition and is mostly independent of the feedstockcompounds, provided that HDO processing conditions remain identical.Fig. 1b takes into account both oxygen and nitrogen heteroatoms,whereas Fig. 1a does not. Negligible changes occurred between thetwo diagrams, indicating that nitrogen was largely unaffected by theHDO. This is because very little nitrogen existed in the initial bio-oils,with the exception of that derived fromhorsemanure. Surprisingly, nei-ther of the pre-catalyzed pyrolysis bio-oils showed any degree of rela-tive deoxygenation, even such that the SwG-Pt HDO outperformedthem both. In contrast, the catalytic pyrolysis oils were hydrogenatedmore strongly, as indicated by the vertical movement of the coordinatepost-HDO. Tables 2 and 3 reflect this phenomenon, since the (H/C) ratioincrease due to HDO of cat-SwG and SwG-recycle bio-oils doubles thatof regular bio-oils (0.22 increase vs. 0.10–0.15). In the end, the final co-ordinate(s) post-HDO are similar to that attained by SwG-Pt.

The chemical compounds present in the bio-oil affect their overallusefulness, so their identification is necessary. Table 4 displays theconcentrations (wt.%) of various compounds measured via GC–MS.While the GC–MS measurements only capture a small portion of thechemical identities present in bio-oil, there exist undetectable com-pounds of larger molecular weight, which undergo the same phenome-na as those detectable components. Hence, GC–MS observations andtrends provide critical information and are reflective of the whole sam-ple. For all runs, the water-miscible components effectively transferredto the aqueous solvent and remained dissolved post-HDO (Fig. 2).Although all bio-oils began with a significant amount of acetic acid,the amount of acetic acid leftover in the oil varies with no discernibletrend, indicating that both phase transfer and catalytic activity eachplay a role in reducing acidic components. For switchgrass bio-oil,decalin was produced in relatively high amounts. As discussed belowfor the cases of catalytically produced pyrolysis oils, naphthalene hydro-genation is not highly favored under these conditions. This suggests that

before and after HDO over Pt/C or Ru/C catalysts.

calyptus Manure

o-oil Pt Ru Bio-oil Pt Ru

.13 0.36 0.29 2.31 2.09 1.96

.45 70.87 69.24 67.35 72.14 73.99

.99 7.70 7.30 6.82 7.91 7.91

.43 21.07 23.17 23.52 17.87 16.14

.209 1.303 1.266 1.215 1.315 1.283

.434 0.223 0.251 0.262 0.186 0.164

.90 4.31 6.40 7.31 2.95 2.1837 33 97 48 32

Table 3Elemental analyses (wt.% dry basis), moisture, and acidity of switchgrass bio-oil HDO using other types of deoxygenation catalysts. The three samples from left to right are: regular bio-oilover Pd/C, HZSM-5 catalytic pyrolysis oil over Pt/C, and tail-gas recycle pyrolysis oil over Pt/C.

SwG cat-SwG SwG-recycle

Bio-oil Pd Bio-oil Pt Bio-oil Pt

N 0.52 0.84 0.73 0.99 1.96 1.36C 53.81 72.04 76.89 75.04 73.67 72.86H 5.36 7.83 6.10 7.36 5.90 7.13O 33.21 19.29 16.27 16.61 18.46 18.65H/C (mol) 1.196 1.306 0.953 1.177 0.961 1.174O/C (mol) 0.463 0.201 0.159 0.168 0.188 0.192Karl-Fischer moisture (wt.%) 7.98 2.35 4.17 1.72 3.08 1.38Total acid number (mg KOH/g) 73 40 12 16 35 21

14 Y. Elkasabi et al. / Fuel Processing Technology 123 (2014) 11–18

decalin and other higher carbon number hydrocarbons observed formthrough intermolecular reactions and not direct HDO/hydrogenationof naphthalenol type structures. Selectivity towards particular com-pounds would make the bio-oil properties easier to predict. Further-more, distillation into high-purity fractions would be facilitated by thepredominance of a few compounds. The same can be mentioned forphenols for non-fuel purposes. Upgraded and refined fuels require apredominant concentration of saturated hydrocarbons, in addition tosingle aromatic ring structures. Switchgrass bio-oil HDO increasesboth types of molecules in relatively significant amounts. More optimi-zation of process conditions should be pursued to elevate saturated hy-drocarbons to the desired level.

Tables 5 and 6 display the GC–MS compositions of the remainingfeedstock bio-oils, before and after HDO. Again, the platinum catalystproduced more decalin and tetralin for eucalyptus bio-oil. Ru resultedin more phenols and cresols post-HDO, as it did with switchgrass.

Fig. 1. Van Krevelen-type plots of pyrolysis bio-oil compositions before and after HDO,considering (a) oxygen only and (b) oxygen and nitrogen. Feedstock and catalysts are in-dicated by the legend and by insets, respectively. Petroleum composition is added forcomparison.

However, both Ru and Pt-catalyzed bio-oil contain similar amounts ofacetic acid and acetol. This happens despite the large abundance ofacetic acid in Eucalyptus bio-oil compared to switchgrass bio-oil.Levoglucosan and acetol were effectively removed from Eucalyptusbio-oil, just as they were removed from switchgrass bio-oil. The GC–MS percentages represent a small fraction of the total bio-oil composi-tion, as a result of the large abundance of compounds which are unde-tectable by GC–MS. This includes moderately small molecular weightcompounds, oligomeric lignins, and aggregates of larger molecularweight. Keeping all components in the bio-oil during HDO allows foran increased yield of smaller compounds through the catalytic break-down of larger compounds and aggregates. This is reflected in the in-creased percentages of phenols and cresols. With catalytic pyrolysisbio-oils, the concentrations of saturated hydrocarbons showed a greaterincrease, which is reflective of the high degree of hydrogenation. TheHZSM-5 pyrolysis bio-oil moderately decreased in naphthalenes con-tent, whereas recycle pyrolysis bio-oil increased in naphthalenes. TheHZSM-5 catalyzed bio-oil naphthalenes decrease correlates well withthe tetralin increase, but significant amounts of naphthalene remainsuggesting that naphthalene (and subsequent tetralin) hydrogenationreactions are not highly favored under these reaction conditions. How-ever, with recycle oil, the naphthalenes, decalin, and tetralin concentra-tions increase. This suggests that, unlike the case of switchgrass oilproduced in an inert environment, the ultimate source of the fusedsix-membered ring structures may be naphthalenols or other oxidizednaphthalene derivatives. Conversion of an oxidized naphthalene todecalin is not a hydrogen efficient process in terms of HDO, utilizing6 mol of hydrogen to eliminate one oxygen atom.

Past hydrotreating experiments using cresol as a model compound[14,19,23] resulted in formation of methylcyclohexane and methyl-cyclohexanol as the major products. However, neither compound wasformed from any of the experiments described herein, despite the useof higher H2 pressures, higher catalyst loading ratios, and higher reac-tion temperatures. The relatively dry nature of the ESP oil fractionused (3–8%) is a contributing factor, as it effectively decreases the cata-lyst loading ratio relative to other studies. Still, other studies [14] usedeven stronger reaction conditions and demonstrated a modest deoxy-genation conversion with model compounds. Also, our experimentalconditions facilitate hydrogenation alongside deoxygenation, as evi-denced by the increased H:C ratios.

For all feedstocks, the aqueous phase contained various organic acidsand a small amount of partially-miscible phenols. The aqueous-solublecomponents altogether comprise ~3–6% of the aqueous phase viaKarl-Fischer moisture analysis. While the aqueous phases remainedstable post-HDO, this was not the case with that from the manure bio-oil HDO. The color of aqueous phases began as a clear light-yellow and

Table 4GC–MS concentrations (wt.%) of switchgrass bio-oils treatedwith different HDO catalysts.

Bio-oil Ru Pd Pt

Acetic acid 4.83 1.70 0.19 0.20Acetol 2.04 0.87 0.03 0.01Levoglucosan 6.02 0.02 0.00 0.02Furfural 0.04 0.00 0.00 0.00Xylenes 0.00 0.03 0.01 0.02Decalin 0.00 2.27 0.52 3.23Tetralin 0.00 0.14 0.03 0.19Indene 0.01 0.00 0.00 0.00Indane 0.00 0.16 0.03 0.06Phenolsa 0.42 6.43 2.20 2.83Cresols 0.17 2.20 0.65 0.96Guaiacolsb 0.03 0.16 0.07 0.18Syringolsc 0.00 0.00 0.01 0.01Naphthalenes 0.04 0.25 0.07 0.06

a Includes phenol, 4-ethylphenol, and 2,4-dimethylphenol.b Includes guaiacol, vanillin, and 2-methoxy-4-methylphenol.c Includes syringol and syringaldehyde.

Table 5GC–MS concentrations (wt.%) of eucalyptus andhorsemanure bio-oils treatedwith differ-ent HDO catalysts.

Eucalyptus Manure

Bio-oil Pt Ru Bio-oil Pt Ru

Acetic acid 7.33 0.32 0.27 7.91 0.67 0.79Acetol 3.06 0.03 0.09 0.13 0.05 0.00Levoglucosan 2.97 0.00 0.00 0.01 0.01 0.01Furfural 0.06 0.01 0.00 0.00 0.00 0.00Xylenes 0.01 0.03 0.03 0.07 0.03 0.04Decalin 0.01 0.16 0.02 0.00 0.00 0.00Tetralin 0.00 0.47 0.13 0.00 0.06 0.05Indene 0.02 0.00 0.00 0.18 0.01 0.01Indane 0.00 0.16 0.13 0.00 0.17 0.17Phenols 0.31 0.75 0.94 2.25 3.42 3.48Cresols 0.33 0.66 2.12 1.76 3.08 2.92Guaiacols 0.18 0.18 0.12 0.01 0.01 0.05Syringols 0.64 0.13 0.07 0.02 0.01 0.00Naphthalenes 0.04 0.25 0.11 0.34 0.40 0.44

15Y. Elkasabi et al. / Fuel Processing Technology 123 (2014) 11–18

remained as such, whereas the manure bio-oil HDO aqueous phasebegan as a light pink and progressively became darker with time,resulting in precipitation of a solid over time. This is attributable tothe greater ratio of phenols to acid present (Fig. 2), which undergo con-densation reactions and precipitate from solution. Thus, the yield fromthemanure bio-oil systemcould be better optimized through a differentsolvent system or by otherwise minimizing transfer of components tothe aqueous phase.

3.2. Non-condensable gas production

The composition of the HDO non-condensable gas (NCG) phase pro-vides valuable information on the types of reactions that likely occurred,

Fig. 2. GC–MS chromatograms of aqueous

extent of reaction, and the selectivity of reaction pathways. Both the rel-ative and absolute gas-phase compositions are necessary to provide thisinformation (Fig. 3a and b, respectively). While only 8 gas-phase com-pounds were measured, these 8 are the most predominant in con-centration. Other higher molecular weight carbon-based gases werepresent in trace amounts. Both switchgrass bio-oil over Pt (SwG-Pt)andmanure bio-oil over Pt (Manu-Pt) yielded the highest selectivity to-wards CO2 production (Fig. 3a). SwG-Pd gave the highest selectivity to-wards CO, but it produced more NCGs, which accounts for its similarityto SwG-Ru elemental compositions. CO production is unfavorable sinceit does not eliminate oxygen as efficiently as CO2. All eucalyptus HDOexperiments produced significantly larger amounts of NCGs, amountingto at least twice the normal amount inmost cases. Eucalyptus HDO pro-ducedmuch greater levels of CH4. This phenomenon is attributed to thedifference in feedstock characteristics. All lignocellulosic feedstocks

phases after HDO of bio-oil over Pt/C.

16 Y. Elkasabi et al. / Fuel Processing Technology 123 (2014) 11–18

contain varying ratios of three lignin monomers [5]: phenols, guaiacols,and syringols, with guaiacol and syringol being the most predominant.

Fig. 3. HDO gas product compositions as measured by gas chromatography, calculatedbased on (a) relative comparisons and (b) absolute percentages of the final gas phase.

Guaiacol and syringol differ by the number of methoxy groups presenton the ring structure.

Methoxy groups are easily cleaved by transitionmetal catalysts [24],and as such, more NCGs form, especially CH4. Eucalyptus lignin containsa much higher syringol:guaiacol (S:G) ratio than does switchgrass [25].Since syringol possesses twice the methoxy groups as guaiacol, theEucalyptus NCG production will increase accordingly. This reasoning isfurther confirmed by GC–MSmeasurements in Tables 4–6. The syringoland guaiacol concentrations in switchgrass and eucalyptus bio-oils re-flect these characteristics. Furthermore, all eucalyptus experimentsshow significant reductions in syringol concentrations.

While both SwG-Pt and Manu-Pt exhibit high CO2 selectivity, thisfactor alone does not explain the high deoxygenation rate of SwG-Pt,since Manu-Pt did not deoxygenate as well as Manu-Ru. One possiblereason for the discrepancy is the excess formation of high-MW paraffincompounds in SwG-Pt, which increases carbon retention and decreasestheO:C ratio. This phenomenon does not occur significantly inManu-Pt.Also, manure bio-oils lose more carbon to the aqueous phase, which in-creases its effective O:C ratio. Since the greater carbon loss in Manu-Ptdoes not occur in the gas phase, it must occur in the aqueous phase.Karl-Fisher measurements confirm a greater loss of total organic tothe aqueous phase with manu-Pt than with manu-Ru (2.2 g vs 1.6 g).One way of quantifying these relative differences is by comparing therelative and absolute deoxygenation values in Fig. 4a and b, respectively.The equations defining these values are shown in Eqs. (1) and (2):

%DOrel ¼ 1− xOxO;i

!� 100 ð1Þ

%DOabs ¼ 1− xOmoil

xO;imoil;i

!� 100 ð2Þ

Table 6GC–MS concentrations (wt.%) of catalytic and recycle pyrolysis switchgrass bio-oils treat-ed with Pt/C HDO catalyst.

cat-SwG SwG-recycle

Bio-oil Pt Bio-oil Pt

Acetic acid 0.44 0.18 3.05 0.10Acetol 0.13 0.01 0.11 0.02Levoglucosan 3.56 0.00 0.01 0.01Furfural 0.00 0.00 0.02 0.00Xylenes 0.60 0.31 0.15 0.18Decalin 0.00 0.45 0.00 0.07Tetralin 0.01 3.93 0.01 2.57Indene 0.86 0.02 1.97 0.11Indane 0.12 1.53 0.08 1.88Phenols 1.42 4.24 5.38 7.78Cresols 1.06 3.01 3.92 5.85Guaiacols 0.09 0.05 0.00 0.02Syringols 0.02 0.01 0.01 0.01Naphthalenes 12.79 9.46 2.35 3.56

The variables xO and xO,i represent final and initial wt.% oxygen, re-spectively. Similarly, moil and moil,i represent the final and initial massof oil phase, respectively. Eq. (1) only indicates the percent change inoxygen relative to other atoms, whereas Eq. (2) is an absolute massbalance that indicates percent change in oxygen atoms. The two figuresillustrate that SwG-Pt gives the best deoxygenation and carbon reten-tion. Manu-Pt exhibits a greater loss in carbon compared to Manu-Ru(Fig. 4a), even though the absolute oxygen losses are similar (Fig. 4b).Both catalytic pyrolysis and recycle switchgrass bio-oils over Pt(catSwG-Pt and SwG-rec-Pt, respectively) exhibited negative %DOrel

values, indicating that deoxygenation was small compared to otheratomic losses, primarily carbon.

3.3. Hydrogen consumption and efficiency analysis

HDO hydrogen consumption analysis provides information on boththe mechanism of HDO and viability of continuous scale-up. In orderto maintain isobaric conditions, a secondary reservoir tank providedconstant H2 pressure with real-time pressure readings of the reservoir.Thus, hydrogen was consumed from either the reaction atmosphereor directly from the reservoir tank. Using the ideal gas law, the totalmoles of hydrogen consumed is calculated according to Eq. (3):

nconsumed ¼ Δnreservoir þXni¼1

xingas ð3Þ

where xi is the mole fraction of gas component i. When hydrogen con-sumption is normalized with respect to the starting bio-oil mass, theresults are plotted with respect to %DOrel (Fig. 5). The dotted line [26]represents the stoichiometric equivalent of consuming one hydrogen

Fig. 4. (a) Relative deoxygenation and (b) absolute deoxygenation percentages are plottedwith respect to the associated product yield of each HDO reaction. Switchgrass with plat-inum performs the best, in terms of maximizing deoxygenation and minimizing carbonloss.

17Y. Elkasabi et al. / Fuel Processing Technology 123 (2014) 11–18

molecule for every oxygen atom lost, representative of a direct con-version to water. Since oxygen loss will include physical transfer tothe aqueous phase, and since hydrogen consumption includes direct hy-drogenation, the relative efficiency of hydrogen consumption does notpurely reflect HDO reactions. The most efficient processes will land to-wards the lower right section of the plot. Although the Manu-Ru andSwG-Pt combinations are the most hydrogen efficient, SwG-Pt givesthe best relative deoxygenation for the associated H2 consumption.

In contrast, the most inefficient reactions with regard to hydrogenconsumption all pertained to eucalyptus bio-oil. Just as the S:G ratioproduced twice as much NCG as the other feedstocks, nearly twice as

Fig. 5. Plot of hydrogen consumption for each experiment, with respect to the associateddeoxygenation percentage (relative). The dotted line represents the theoretical stoichio-metric consumption of one hydrogen molecule for every oxygen atom removed.

much hydrogenmust be consumed for that to happen. Consequentially,high-syringol feedstocks like eucalyptus will require different catalyststhat can transalkylate methyls from the methoxy groups to the ring.Ru/TiO2 is one example of such a system [27,28]. This will better ensurethe carbon efficiency of the overall feedstock conversion (pyrolysis +HDO). Likewise, feedstocks with higher fractions of phenols and guai-acols can be upgraded by carbon-supported catalysts in a more cost-effective manner.

Since deoxygenation is not the only desired reaction, we computedand plotted a total extent of upgrading which takes into account bothhydrogenation and absolute deoxygenation. Degree of hydrogenation(DH) is defined by Eq. (4):

%DH ¼ H : Cð Þ− H : Cð ÞiH : Cð Þmax− H : Cð Þi

� �� 100 ð4Þ

where theoretical (H:C)max = 2 for a desired end product (e.g. gaso-line), and (H:C)i is the initial ratio. The total extent of upgrading (TEU)is then defined by (5) as an average and plotted in Fig. 6:

TEU ¼ %DOabs þ %DH2

: ð5Þ

Regardless of the yield, catalyst, or feedstock, all experiments fell ona single linear trend line. These experimental data points include cata-lytic pyrolysis experiments where HDO was ineffective. AlthoughSwG-Pt gives the highest TEU, its associated yield is the lowest. An opti-mal set of conditions will compromise between high yields and highdeoxygenation rates. It will be necessary to conduct a full economic op-timization that incorporates the overall efficiency from pyrolysis toHDO, including the feedstock losses during pyrolysis. This is also neces-sary so that the yields and deoxygenation rates of catalytic pyrolysistechnologies are factored into the optimization. Future work shouldalso investigate the rates of catalyst deactivation and coke formation,which could significantly shift the efficiency of operations. Also as partof future work, NMR studies will aid in determining compound identi-ties post-HDO.

4. Conclusions

A systematic hydrodeoxygenation study was carried out with re-spect to fast-pyrolysis bio-oils from various feedstocks, as well as bio-oils produced under catalytic process conditions. HDO reactions usedtransitionmetal catalysts on carbon supports. Critical trendswere relat-ed to the feedstock composition and processing parameters. Out ofswitchgrass, eucalyptus, and horsemanure bio-oils studied, switchgrassbio-oil catalyzed with platinum shows the most promise for overall

Fig. 6. Total extent of reaction plottedwith respect to the associated yield of upgraded bio-oil. The trend line is nearly independent of the feedstock type and/or catalyst.

18 Y. Elkasabi et al. / Fuel Processing Technology 123 (2014) 11–18

upgrading efficiency. This combination yielded the highest degrees ofdeoxygenation for its yield and consumed low amounts of hydrogen.Organic yields correlated well with waste product yields (gas product+ aqueous organics) for all runs, primarily due to HDO oxygen loss. De-viations from this trend were caused by either excessive gas volumesand/or selectivity towards CO production. Coke comprised very littleof the product, since HDO mass balances totaled N80% closure withoutcoke. Analysis of the gas and liquid phases for switchgrass over plati-num catalyst indicated greater carbon sequestration. All unmodifiedfeedstocks followed the same upgrading pathway on the Van Krevelendiagram, making the atomistic composition a greater factor indeoxygenating normal bio-oil than the actual chemical identities. Cata-lytic pyrolysis over HZSM-5 and/or tail-gas recycling prevented any sig-nificant deoxygenation post-pyrolysis, suggesting some temperatureincrease to be necessary. Eucalyptus feedstock exhibited poor HDO effi-ciency, as indicated by a large excess of hydrogen consumption andnon-condensable gas production. Feedstocks with high syringol:guaiacol ratios are expected to perform just as poorly as eucalyptus, sug-gesting that a different catalyst systemaltogether is necessary for reduc-ing their associated process costs. Platinum catalysts generallydeoxygenated more efficiently than ruthenium, except in the case ofmanure bio-oil, in which greater losses to the aqueous phase occurred.Even though several of the feedstock bio-oils deoxygenated very well,some exhibited poor deoxygenation relative to other atoms present.Variables which depend on feedstock include the relative percent deox-ygenation and the hydrogen consumption rate. In contrast, the absolutepercent deoxygenation and total extent of upgrading relied only on thetotal yield, regardless of feedstock or catalyst choice. The aforemen-tioned observations can make HDO techno-economic analyses morebroadly applicable to various feedstocks.

Acknowledgments

The authors would like to thank Charles Coe (Villanova University)for his work in constructing the HDO reactor system, and we alsothank Johnson Matthey Inc. (UK) for providing the catalysts used inthis study.We also thank Craig Einfeldt for the operationalmaintenanceof the fluidized bed pyrolysis system and Gerry Senske for use of thecentrifuge. USDA-NIFA-BRDI grant 2012-10008-20271 is herebyacknowledged.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.fuproc.2014.01.039.

References

[1] S. Sorrell, J. Speirs, R. Bentley, A. Brandt, R. Miller, Energ Policy 38 (2010) 5290–5295.[2] G.W. Huber, S. Iborra, A. Corma, Chemical Reviews 106 (2006) 4044–4098.[3] P.L. Thigpen, W.L. Berry, in: D.L. Klass (Ed.), Energy from Biomass and Wastes VI, In-

stitute of Gas Technology, Chicago, 1982, p. 1057.[4] D.C. Elliot, G.F. Schiefelbein, American Chemical Society Division of Fuel Chemistry

Preprints 34 (1989) 1160–1166.[5] D. Mohan, C.U. Pittman Jr., P.H. Steele, Energy and Fuels 20 (2006) 848–889.[6] P. McKendry, Bioresource Technology 83 (2002) 55–63.[7] J.W. Lee, B. Hawkins, D.M. Day, D.C. Reicosky, Energy and Environmental Science 3

(2010) 1695–1705.[8] D.J. Mihalcik, A.A. Boateng, C.A. Mullen, N.M. Goldberg, Industrial and Engineering

Chemistry Research 50 (2011) 13304–13312.[9] C.A. Mullen, A.A. Boateng, Bioenergetics Research 4 (2011) 303–311.

[10] A.A. Boateng, D.E. Daugaard, N.M. Goldberg, K.B. Hicks, Industrial and EngineeringChemistry Research 46 (2007) 1891–1897.

[11] M.I. Schnitzer, C.M. Monreal, G.A. Facey, P.B.J. Fransham, Journal of EnvironmentalScience and Health B 42 (2007) 71–77.

[12] D.C. Elliot, E.G. Baker, Energy from Biomass & Wastes X, 1986. 765–784.[13] F.M. Mercader, M.J. Groeneveld, S.R. Kersten, C. Geantet, G. Toussaint, N.W. Way, C.J.

Schaverien, K.J. Hogendoorn, Energy and Environmental Science 4 (2011) 985–997.[14] J. Wildschut, I. Melian-Cabrera, H.J. Heeres, Applied Catalysis B: Environmental 99

(2010) 298–306.[15] J. Wildschut, F.H. Mahfud, R.H. Veenderbosch, H.J. Heeres, Industrial and Engineer-

ing Chemistry Research 48 (2009) 10324–10334.[16] P.M. Mortensen, J.D. Grunwaldt, P.A. Jensen, K.G. Knudsen, A.D. Jensen, Applied

Catalysis A: General 407 (2011) 1–19.[17] W. Kiatkittipong, S. Phimsen, K. Kiatkittipong, S.Wongsakulphasatch,N. Laosiripojana,

S. Assabumrungrat, Fuel Processing Technology 116 (2013) 16–26.[18] T.P. Vispute, H. Zhang, A. Sanna, R. Xiao, G.W. Huber, Science 330 (2010) 1222–1227.[19] H. Wan, R.V. Chaudhari, B. Subramaniam, Topics in Catalysis 55 (2012) 129–139.[20] C.A. Mullen, A.A. Boateng, N.M. Goldberg, Energy and Fuels 27 (2013) 3867–3874.[21] D.C. Elliot, T.R. Hart, G.G. Neuenschwander, L.J. Rotness, A.H. Zacher, Environmental

Progress 28 (2009) 441–449.[22] E. Furimsky, Applied Catalysis A: General 199 (2000) 147–190.[23] C.R. Lee, J.S. Yoon, Y.W. Suh, J.W. Choi, J.M. Ha, D.J. Suh, Y.K. Park, Catalysis Commu-

nications 17 (2012) 54–58.[24] T. Nimmanwudipong, R.C. Runnebaum, D.E. Block, B.C. Gates, Catalysis Letters 141

(2011) 779–783.[25] C. Li, L. Sun, B. Simmons, S. Singh, Bioenergetics Research 6 (2013) 14–23.[26] R.H. Venderbosch, A.R. Ardiyanti, J. Wildschut, A. Oasmaa, H.J. Heeres, Journal of

Chemical Technology and Biotechnology 85 (2010) 674–686.[27] S. Wan, T. Pham, S. Zhang, L. Lobban, D. Resasco, R. Mallinson, AIChE Journal 59

(2013) 2275–2285.[28] S. Boonyasuwat, T. Omotoso, D.E. Resasco, S.P. Crossley, Catalysis Letters 143 (2013)

783–791.