Fuel 117 (2014) 1061–1073

Contents lists available at ScienceDirect

Fuel

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

Enhancement of phenol hydrodeoxygenation over Pd catalystssupported on mixed HY zeolite and Al2O3. An approach to O-removalfrom bio-oils

0016-2361/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.fuel.2013.10.011

⇑ Corresponding authors. Tel.: +34 915854769; fax: +34 915854760 (B. Pawelec),Tel.: +34 946017282; fax: +34 946014179 (P.L. Arias).

E-mail addresses: bgarcia@icp.csic.es (B. Pawelec), pedroluis.arias@ehu.es (P.L.Arias).

S. Echeandia a, B. Pawelec b,⇑, V.L. Barrio a, P.L. Arias a,⇑, J.F. Cambra a, C.V. Loricera b, J.L.G. Fierro b

a School of Engineering (UPV/EHU), Chemical and Environmental Engineering Department, c/Alameda Urquijo s/n, 48013 Bilbao, Spainb Instituto de Catálisis y Petroleoquímica, CSIC, Cantoblanco, 28049 Madrid, Spain

h i g h l i g h t s

� Phenol HDO was compared on Pdcatalysts supported on alumina,zeolite HY and mixtures.� The 20%HY–80%Al2O3 support is

beneficial for phenol HDO reactionover Pd catalyst.� The density of active phases on the

support surface influenced on theHDO activity.� Catalyst acidity influence on both

activity and coke formation.

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

0

20

40

60

80

100

Pd/20%HY-A

l

Phen

ol c

onve

rsio

n (%

)

Catalyst

NiMo/Al203-zeosulfided

HDO of phenol:T=250 °CP =1.5 MPaWHSV =0.5 h-1

Pd/Al

Pd/10%HY-A

l

Pd/HY

a r t i c l e i n f o

Article history:Received 1 December 2011Received in revised form 7 October 2013Accepted 8 October 2013Available online 23 October 2013

Keywords:PhenolHDOHY zeoliteAlumina–HYAlumina

a b s t r a c t

This contribution describes the effect of the support (zeolite ultrastable HY, alumina (Al) and mixedHY–Al carriers) on the catalytic activity of Pd catalysts in the phenol hydrodeoxygenation (HDO) reactioncarried out in a flow fixed-bed reactor at T = 523–573–623 K, P = 15 bar and WHSV = 0.5 h�1. Phenoldissolved in n-octane was used as model compound of bio-oil species derived from fast pyrolysis of lig-nocellulosic biomass. The catalysts were characterized by N2 physisorption, XRD, TPR, TPD-NH3, DRIFTspectroscopy of adsorbed CO, HRTEM, X-ray photoelectron spectroscopy (XPS) and TPO/TGA techniques.The largest phenol conversion (63%) achieved at 523 K over the reduced Pd/20%HY–Al catalyst was sim-ilar to that obtained on a commercial NiMo/Al2O3–zeolite hydrocracking sample (HCK) activated by sulf-idation. Regardless of the reaction temperature, the only products detected in the HDO of phenol over allcatalysts studied were four O-free compounds: benzene, cyclohexene, cyclohexane, and methylcylopen-tene. Both reduced Pd/20%HY–Al and sulfided commercial HCK catalysts produced similar yields of O-freeproducts. From the catalyst activity-structure correlation, it can be concluded that the HDO of phenol isfavoured on the bifunctional Pd/20%HY–Al catalyst which possesses moderate acidity and improved Pddispersion on the support surface. The contributions of the acid sites to the catalyst activity and deacti-vation by coke are discussed.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Mitigating greenhouse gases (GHGs) is one of the biggest chal-lenges in the 21st century and requires long-term planning as well

1062 S. Echeandia et al. / Fuel 117 (2014) 1061–1073

as social awareness. Renewable sources of energy can contributesignificantly to the effort of mitigating GHGs. Among differentrenewable energy technologies, biomass-based technologies havehigh potential [1]. Biomass has been proposed as a feedstock for li-quid fuels and chemicals [2,3]. Many processes that transform bio-mass to liquid fuels start with the thermal breakdown of biomassfeedstock producing the so-called bio-oils [4]. The final bio-oilshave higher energy density than the raw biomass precursor, whichmakes it more suitable for industrial applications [5].

Pyrolysis bio-oils obtained from the fast pyrolysis of biomassare complex mixtures of reactive chemical compounds (carboxylicacids, aldehydes, ketones, carbohydrates, degraded lignin), water(up to 25%), and some alkali and alkaline-earth metals (Na, K,Mg, Ca) [6]. The high contents of oxygenated compounds (close to40% oxygen) contribute to some deleterious properties, such as highviscosity, corrosiveness, poor heating value, immiscibility withhydrocarbon fuels, as well as to undesirable formation of carbondeposits in parts of the automotive engines upon combustion, espe-cially in compression–ignition engines, i.e. diesel engines. In order toimprove the thermostability of these bio-oils, the oxygen removalvia hydrodeoxygenation (HDO), which is a similar process to hydro-treatment in a petroleum refinery, is a common practice [7].

Past efforts for HDO catalyst development were focused on alu-mina-supported hydroprocessing catalysts, intensively used inpetroleum refining for several decades and therefore little atten-tion has been paid to development of novel catalysts. Thus, the cat-alysts often studied for HDO were sulfided CoMo/c-Al2O3 andNiMo/c-Al2O3 systems [8–13]. Unfortunately, in the absence of asulfiding agent, the sulfided catalysts deactivated quickly in HDOreaction and the selectivity to different hydrocarbons changedwith time on-stream due to oxidation of the active phases [11].Thus, great efforts have been devoted to enhance the stability ofthese catalysts, but the results are not yet satisfactory [7–13]. Re-cent laboratory and commercial developments in the field of cata-lytic hydroprocessing of biomass-derived liquefaction conversionproducts are presented in a few excellent revisions [14–19].

Recently, the use of noble metal catalysts as an alternative tosulfided catalysts for HDO reaction has been extensively studied[14–19 and references therein]. As pioneering patent demon-strated that Pd catalysts are highly effective for hydrogenation oforganic compounds typically found in bio-oils [20], more studieswere developed on Pd-based catalysts than in other noble metals[14–19 and references within]. Thus, the HDO activity of Pd-basedcatalysts was investigated using acidic substrates such as ZrO2

[21–23], Al2O3 [24,25], SiO2–Al2O3 [25], SiO2 [26–27], SBA-15 andAl-SBA-15 [29], Al2(SiO3)3 [29], SAPO-31 [30], carbon [31–39], zeo-lites Beta and ZSM-5 [37,38], MCM-41 [37], and super acid SO2�

4 /ZrO2/SBA-15 [40] substrates. On the contrary, studies of the HDOreaction over Pd catalysts supported on basic carriers are scarce[41,42]. The reason for this lies in the fact that the HDO reactionover noble metals supported on acidic substrates requires bothmetal site and acid sites, being H2 dissociated on metal sites whileO-containing compounds are adsorbed and activated on either me-tal sites or on the cations/oxygen vacancies located at the metal-support interface [14–19]. Moreover, it is well known that theacidic sites of supports might to catalyse dehydration, isomerisa-tion, alkylation and condensation reactions [43].

The effects of the substrate structure (C, H-Beta, ZSM-5, MCM-41) and its acidity on activity of supported Pd catalysts in thetransformation of benzophenone and benzaldehyde were investi-gated by Cejka et al. [37,38] who found that the substrate structurehas a determining effect on the course of HDO reaction being themost suitable one Pd catalysts supported on large pore zeolite Beta[37,38]. This was confirmed also for Pt catalysts supported on H-Beta [44] and HY [45,46] zeolites. Similarly, Pt catalysts supportedon mesoporous ZSM-5 showed better performance in dibenzofuran

HDO than its counterpart supported on microporous ZSM-5 zeolitedue to the diffusion limitation of dibenzofuran in the microporousstructure of ZSM-5 [44]. Considering the stability of substrate, aparticularly promising approach to the development of novelHDO catalysts lies in the use of highly acidic ultrastable HY zeolitesubstrate. This is because the ultrastable HY zeolite has the advan-tages of higher surface-to-volume ratio than alumina, variableframework compositions and high hydrothermal stability [47].Zeolite acidity, however, drastically leads to fast deactivation ratesand increases the amount of undesirable cracking products, whichaccelerates the rate of coke deposition and the yield of gases [48].In this sense, a large deactivation of bifunctional zeolite-supportednoble metal catalysts in the HDO of phenol was reported [45].

To the best of our knowledge, the only study on HY-supportednoble metal HDO catalysts was reported by Hong et al. [46] whoshowed that bifunctional HY zeolite supported Pt catalysts areeffective for the aqueous phenol HDO using a fixed-bed reactorworking under high H2 pressure (4 MPa). It was found that phenoltransformation over Pt/HY catalysts proceeds via hydrogenation–hydrogenolysis ring-coupling reactions producing monocyclicand useful bicyclic hydrocarbons [46]. The bare HY zeolite demon-strated to be not effective catalyst for HDO of phenol upon reactionconditions employed (T = 473–523 K and PH2 = 4 MPa) confirmingthat bifunctional catalysts are needed for HDO reaction: H+ sitesfor dehydration and metal sites for hydrogenation are needed. Not-withstanding, it is emphasized that Pt/HY catalysts were activatedat high temperature (773 K) [46], thus sintering of Pt crystallites isexpected to occur on the surface of HY zeolite.

Taking into account that future commercial applications of lignin,its conversion will require catalysts with improved resistance todeactivation [44]. In the present work, we aimed to exploit theadvantage of the properties of the alumina and HY zeolite mixturesas support for Pd in order to develop stable and more selective cat-alysts toward the hydrogen-donor species. Indeed, revision of the lit-erature on the use of Pd catalysts for HDO reaction [20–42] indicatesthat hybrid HY–Al substrate has not been used to prepare Pd HDOcatalysts. Thus, we compare here the activity for phenol HDO reac-tion of Pd catalysts supported on alumina, mixed alumina-HY zeolitematerials and alumina-free zeolite, trying to establish a relationshipbetween activity and catalyst structure. Moreover, considering thatthe mechanism of hydrogenation over supported noble metal cata-lyst is still debated [14–19], our selectivity results might contributeto clarify this point. The physicochemical properties of the catalystshave been evaluated by various techniques and their activity com-pared with those of a commercial zeolite-loaded alumina-supportedNiMo sulfide hydrocracking catalyst to provide the catalyst withhydrocracking function arising from substrate acidity.

2. Experimental

2.1. Catalyst preparation

Four supported Pd catalysts were prepared using c-Al2O3 (Gir-dler), HY zeolite (Conteka), and mixed 10%HY/c-Al2O3 as supports.Two mixed HY–Al2O3 supports were prepared by physical mixtureof different amounts (10 and 20 wt.%) of ultrastable HY zeolite(Conteka) and c-Al2O3 (Girdler) in excess of distilled water undervigorous stirring at room temperature for 2 h. The characteristicsof the HY zeolite are as follows: SiO2/Al2O3 mole ratio 5.6, Na2Ocontent 0.14 wt.% and unit cell 2.454 nm. After solid decantation,the solid was dried at 383 K for 12 h and calcined at 523 K for1 h. Depending on the composition, the supports will be abbrevi-ated hereafter as (x)HY–Al, where x is the nominal wt.% of HY(10% and 20%). Before Pd incorporation, all substrates were driedat 523 K for 1 h. Supported palladium catalysts were prepared bywet impregnation of the respective substrate with aqueous

S. Echeandia et al. / Fuel 117 (2014) 1061–1073 1063

solutions of H2PdCl6 (Sigma) with the appropriate concentration toachieve a metal content of 1 wt.%. A known weight of carrier wascontacted with a solution containing the required amount of palla-dium added to 100 mL of water, and the pH was fixed at a valueclose to 7. After the adsorption equilibrium was reached (contact-ing for 12 h at room temperature), the excess of water was re-moved in a rotary evaporator till dryness. Subsequently, theimpregnate was dried at 383 K in air for 12 h, and finally calcinedat 573 K for 4 h. No chloride was detected by XPS and EDX/TEMtechniques.

2.2. Catalyst characterization

2.2.1. Scanning electronic microscopyThe morphology of all calcined catalysts was studied by scan-

ning electronic microscopy (SEM) on a Hitachi Tabletop TM-1000electron microscope, equipped with energy dispersive X-ray anal-ysis EDX QUANTAX 50.

2.2.2. Wide angle X-ray diffraction (XRD)Calcined catalysts were characterized by powder X-ray diffrac-

tometry according to the step-scanning procedure (step size 0.02�;0.5 s) with a computerized Seifert 3000 diffractometer, using Ni-fil-tered Cu Ka (k = 0.15406 nm) radiation and a PW 2200 Bragg–Brent-ano h/2h goniometer equipped with a bent graphite monochromatorand an automatic slit. The assignment of the various crystallinephases was based on the JPDS powder diffraction file cards.

2.2.3. N2 adsorption–desorption isothermsThe textural properties of the oxide catalysts and bare supports

were determined from the adsorption–desorption isotherms ofnitrogen recorded at 77 K with a Micromeritics TriStar 3000 appa-ratus. Prior to measurements, the samples were degassed at 573 Kunder vacuum for 5 h. Specific surface areas of the supports andsupported Pd catalysts were calculated by the BET method fromthe N2 adsorption–desorption isotherms. The normalized SBET

(NSBET) was calculated from Eq. (1):

NSBET ¼ SBET ox=½ð1� yÞSBET sup� ð1Þ

where SBET ox and SBET sup are the values of BET areas of the oxidecatalyst and support, respectively, and y is the Pd loading.

2.2.4. Chemical analysesThe metal loading of the calcined catalysts was determined by

Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES), Perkin Elmer Optima 3300DV. The solid samples were firstdigested (in a mixture of HF, HCl and HNO3) in a microwave ovenfor 2 h. Then, aliquots of solution were diluted to 50 mL usingdeionized water (18.2 mX quality). The metal loadings of calcinedcatalysts are given in Table 1.

2.2.5. Temperature-programmed techniquesTemperature-programmed reduction (TPR) was performed on

the oxide precursors using a semiautomatic Micromeritics TPD/

Table 1Palladium contenta and textural propertiesb,c of pure supports and oxide precursors.

Catalyst Pda (wt%) SBETb (m2 g�1)

Pd/Al 0.81 154 (163)Pd/10%HY–Al 0.77 179 (185)Pd/20%HY–Al 0.80 194 (216)Pd/HY 0.77 476 (594)

a As determined by ICP-AES.b As determined by N2 adsorption–desorption isotherms at 77 K (values in parenthe

volume, d: average pore diameter.c NSBET: normalized SBET calculated from equation NSBET = (SBET of oxide catalyst/[(1 �

TPR 2900 apparatus interfaced to a microcomputer. Prior to reduc-tion, the catalyst (0.05 g) was dried in a TPR cell at 573 K for 1 h ina stream of helium to remove water and other contaminants. TPRprofiles were obtained by passing a 10% H2/Ar flow (50 mL min�1)through the sample. The temperature was increased from roomtemperature to 1273 K at a rate of 15 K min�1, and the amount ofH2 consumed was determined with a thermoconductivity detector(TCD). The effluent gas was passed through a cold trap before theTCD in order to remove water from the exit stream.

The acidity of the fresh reduced catalysts was determined bytemperature-programmed desorption (TPD) of ammonia usingthe same equipment employed for TPR experiments. A sample(0.05 g) was dried in a TPD cell at 573 K for 1 h in a stream of he-lium, reduced with H2 at the same temperature for 1 h, cooleddown in He flow to 373 K, and then exposed to ammonia for0.5 h. TPD measurements were started from 373 K at a heating rateof 15 K min�1 till 873 K using helium as a carrier gas(50 mL min�1). In order to determine the total acidity of the cata-lyst from its NH3 desorption profile, the area under the curvewas integrated. A semiquantitative comparison of the strength dis-tribution was archived by Gaussian deconvolution of the peaks.Weak, medium and strong acidities were defined as the areas un-der the peaks at the lowest, medium and highest temperatures,respectively.

2.2.6. HRTEM measurementsThe reduced (573 K, H2) catalysts were studied by HRTEM

microscopy using a JEM 2100F microscope operating with a200 kV accelerating voltage and fitted with an INCA X-sight (Ox-ford Instruments). Energy dispersive X-ray microanalysis (EDX)system was used to verify the semi-quantitative composition ofthe supported phases. The catalysts were ground into a fine pow-der and dispersed ultrasonically in hexane at room temperature.Then, a drop of the suspension was put on a lacey carbon-coatedCu grid. In order to obtain statistically reliable information, theparticle size distribution was evaluated from several micrographstaken from the same sample. The average particle size was esti-mated using the equation d = Rni�di/Rni, where ni is the numberof particles with diameter di and Rni is the number of particlesused to build the size distribution.

2.2.7. X-ray photoelectron spectroscopy (XPS)The chemical state and surface composition of fresh reduced

(573 K, H2) catalysts were revealed by X-ray photoelectron spec-troscopy (XPS). XP spectra were recorded with a VG Escalab 200Rspectrometer equipped with Mg Ka X-ray source (hm = 1254.6 eV).The procedure followed during XPS measurements was describedelsewhere [48]. The binding energies were calculated with respectto the C–(C,H) component of the C1s peak fixed at 284.8 eV. Theintensity of Pd 3d peaks was estimated by calculating the integralof each peak after subtraction of an S-shaped background and fit-ting the experimental peak to a Lorentzian/Gaussian lines (90G/10L).

NSBETc Vtotal (cm3 g�1) dc (nm)

0.96 0.49 (0.33) 12.7 (7.5)0.99 0.41 (0.44) 11.0 (11.4)0.91 0.37 (0.46) 10.2 (12.0)0.81 0.35 (0.29) 8.9 (7.1)

ses correspond to bare support); SBET: specific BET surface area; Vtotal: total pore

y) � SBET of support], where y is the Pd loading determined by ICP-AES.

1064 S. Echeandia et al. / Fuel 117 (2014) 1061–1073

2.2.8. Coke quantificationThe coke deposited on the spent catalysts was quantified by

thermogravimetry (Mettler Toledo TGA/SDTA851 equipment) bycomparison of the TPO/TG profiles of spent catalysts with thoseof their fresh counterparts. The TPO experiments were carriedout by raising sample temperature up to a final temperature of1173 K at a rate of 10 K min�1 in a 20% O2/N2 gas mixture(50 mL min�1).

2.3. Catalytic activity testing

Activity tests were performed using 0.2 g catalyst with particlesize 0.25–0.4 mm. An industrial NiMo hydrocracking (HCK) cata-lyst was used as reference. Prior to activity tests, catalysts wereactivated by in situ reduction under a 1:3 H2:N2 gas mixture witha hydrogen flow of 2.5 L h�1 (NTP), and at 1 bar total pressure.The reference NiMo/Al2O3–zeolite catalyst was sulfided with10%H2S/90%H2 gas mixture at 673 K for 4 h. The catalysts wereheated at a rate of 10 K min�1 from 298 K to 673 K, held at thistemperature for 4 h under the reduction gas mixture, and thenpurged in flowing N2 for 1 h. Activity measurements were carriedout in a bench-scale fixed-bed catalytic reactor (9 mm i.d. and300 mm length) placed inside a programmable furnace (Microac-tivity-Reference). A K-type thermocouple was used to measurethe temperature of the catalyst bed. Two mass flow controllersmaintained the hydrogen and nitrogen (diluent) flows. Prior toeach run, the system was kept under an inert atmosphere (N2),while the pressure was increased up to 15 bar. The catalytic bedwas heated to the desired temperature and the hydrogen and theliquid feed mixture were fed to the reactor. The experimental con-ditions were as follows: T = 523–573––623 K, total pressure 15 bar,weight hourly space velocity (WHSV) of 0.5 gphe�(gcat h)�1 and afeed consisting of 1 wt.% phenol (Merck > 99%), in n-octane (FlukaChemika > 99%) used as solvent was introduced at a flow of10 g h�1 with a H2:phenol molar ratio of 100:1. n-octane was usedas solvent because blank experiments proved that, contrary toother solvents, this solvent did not react under the reaction condi-tions employed. Moreover, considering its non-polar character, alower interaction of n-octane with acid sites of support, as com-pared with more polar hydrophilic solvent, is expected [38].

The steady-state was usually achieved within 4 h on stream,and the representative composition of the reaction products wascalculated from several subsequent analyses. The on-line analysisof the effluent from the reactor was performed using a gas chro-matograph (HP 5890, capillary column DB-1) equipped with aflame ionization detector. For product identification a gas chro-matograph (Agilent 5973, capillary column HP-5 MS) equippedwith a mass selective detector was used. The main detected prod-ucts of phenol HDO were benzene, cyclohexane, cyclohexene andmethylcylopentane. The steady-state HDO activities wereexpressed by overall phenol conversion (Xphe), where Xphe wasdefined as 1� nphe=n0

phe and n0phe and nphe are the number of moles

of phenol in the feed and the reaction products, respectively.Duplicate experiments showed Xphe values within ±3% deviation.The carbon balances were calculated on the basis of liquid andgas product analyses. The closures obtained were about 95% onaverage.

3. Results

3.1. Characterization of oxide precursors

3.1.1. Physicochemical characteristicsThe morphology of four Pd oxide precursors was studied by

SEM technique. The SEM images of those samples are shown in

Fig. S1 in the Supporting information (SI). SEM observations ofthe Pd/10%HY–Al and Pd/20%HY–Al samples revealed non-uniformdistribution of HY zeolite on the alumina being Pd deposited pref-erentially on the HY surface rather than on alumina, as confirmedby EDX/SEM elemental analysis (data not shown here). Pd/HY sam-ple showed more uniform and smaller grains than its Pd/Alcounterpart.

Table 1 summarizes the main physicochemical characteristicsof all oxide precursors and bare supports. Regardless of the sup-port, all oxide precursors showed similar palladium loading(0.77–0.81 wt.%) which is close to the nominal one (1.0 wt.%).The textural properties of oxide precursors were evaluated fromnitrogen adsorption–desorption isotherms. Fig. 1(a) and (b) showsthe N2 adsorption–desorption isotherms of the bare supports andoxide precursors, respectively. Bare HY and Pd/HY samples exhib-ited type I isotherm with a hysteresis loop which, according to theIUPAC classification, belong to type H4 [49]. The almost horizontalplateau of Pd/HY isotherm is characteristic of an ideal microporoussolid, although some mesoporosity could be also observed. The lat-ter is probably created during the hydrothermal treatment of theHY zeolite. Contrary to the Pd/HY, the N2 adsorption–desorptionisotherms of the Pd catalysts supported on both mixed HY–Al2O3

substrates are of type IV with H1 hysteresis loop. This means thatboth catalysts are mostly mesoporous. The hysteresis loop H1 has anarrow loop with two branches almost parallel and vertical. Thistype of hysteresis is usually found on solids consisting of agglom-erates or aggregates of particles having a narrow pore size distribu-tion. The N2 adsorption–desorption isotherm of Pd/Al belong totype IV with H2-type hysteresis loop, characteristics of solids with‘‘ink-bottle’’ pores [50].

The values of BET area, total pore volume and the average porediameter are also summarized in Table 1. As it can be seen, the BETsurface area follows the trend: Pd/HY� Pd/20%HY–Al > Pd/10%HY–Al > Pd/Al whereas for the total pore volume the trend isdifferent: Pd/Al > Pd/10%HY–Al > Pd/20%HY–Al > Pd/HY. The high-est BET area and the lowest total pore volume of the Pd/HY catalystis due to the large microporous region of HY zeolite (Fig. 1(a)). As aconsequence, this catalyst showed 3-fold higher BET surface areathan its Pd/Al counterpart (476 vs. 154 m2 g�1). The increase inBET surface area when increasing the HY zeolite content in mixedHY–Al2O3 materials is due to a larger contribution of the HY zeolitemicroporosity system. Considering the relatively low zeolite con-tent in the composite carriers, a relatively small increase in N2

sorption capacity upon raising HY content was observed (Table 1).As expected, impregnation of the carriers with H2PdCl6 salt precur-sor leads to a decrease in BET area of the catalysts as compared tothe bare support indicating some pore blocking by palladium spe-cies. Additionally, one might expect that some dissolution of thesupport occurs during impregnation of carrier with an acidic aque-ous solution of H2PdCl6, similar to what was observed previouslyfor the preparation of the RuCl3/Al2O3 catalysts [51,52].

In order to ascertain the pore modification induced by the addi-tion of HY zeolite to alumina, the Barret–Joyner–Halenda (BJH) for-mula was employed to determine pore size distribution (PSDs) ofthe oxide precursors in both meso- (2–50 nm) and macroporous(>50 nm) regions. Unfortunately, the BJH method is imprecise inthe region below 1 nm, where the supercages of the HY zeolite ap-pear. The BJH pore size distribution of the oxide precursors isshown in Fig. 2. As expected, the pore size distribution changedgoing from Pd/Al to Pd/HY sample. The Pd/Al shows a very broadpore size distribution between 2 and 70 nm, with a maximum at17 nm, however the PSD of Pd/HY catalyst is narrower (2–30 nm)and the maximum peaks at ca. 10.9 nm. For the Pd/10%HY–Aland Pd/20%HY–Al catalysts, PSD showed a bimodal distributionwith maxima peaking at about 14.3 and 13.7 nm, respectively indi-cating that the average pore diameter decreased with an increase

0.0 0.2 0.4 0.6 0.8 1.0

0

50

100

150

200

250

300

350

Al

HY

Volu

me

adso

rbed

(cm

3 /g S

TP)

Relative pressure P/P0

(a)

type I - H4 hysteresis

20HY-Al: type IV/ H2 hysteresis

type IV- H1 hysteresis

10%HY-Al20%HY-Al

0.0 0.2 0.4 0.6 0.8 1.00

50

100

150

200

250

300 Pd/Al

Pd/HY

Volu

me

adso

rbed

(cm

3 /g S

TP)

Relative pressure P/P0

0.41

(b)Pd/10%HY-AlPd/20%HY-Al

Fig. 1. N2 adsorption-desorption isotherms at 77 K of bare supports (a) and oxide catalyst precursors (b).

S. Echeandia et al. / Fuel 117 (2014) 1061–1073 1065

of HY content in mixed HY–Al2O3 substrates. Considering the aver-age pore diameter compiled in Table 1, the pore diameter followsthe trend: Pd/Al (12.7 nm) > Pd/10%HY–Al (11.0 nm) > Pd/20%HY–Al (10.2 nm) > Pd/HY (8.9 nm). This trend clearly indicates thatloading HY zeolite into alumina leads to a decrease of the averagepore diameter, as it could be expected.

To evaluate the influence of Pd loading on the location of Pd spe-cies within the support structure, the normalized SBET (NSBET) wascalculated employing Eq. (1). For the Pd/Al and both Pd catalystssupported on mixed HY–Al2O3 materials, NSBET values in 0.91–0.99range suggest the presence of Pd species on the support surface.For the Pd/HY sample, the NSBET value of 0.8 suggests the presenceof Pd species on the support surface as well as their location withinthe inner support’s porous structure (probably in the supercages ofthe HY zeolite). Thus, the comparison of the normalized SBET valuesindicates that, regardless of the support, most of the palladium spe-cies are accessible for reactant molecule (phenol).

0 20 40 60 80 100 120

dV/ d

log(

D) P

ore

volu

me

(cm

3 /g/n

m)

Pore diameter (nm)

17.0 nm

10.9 nm

14.3 nm

13.7 nm

Pd/20%HY-Al

Pd/HY

Pd/Al

Pd/10%HY-Al

Fig. 2. BJH pore size distribution of Pd-based oxide precursors (from adsorptionbranch) calculated by the Harkins and Jura equation.

3.1.2. Wide-angle XRDX-ray powder diffraction data of the calcined catalysts were

used to identify crystalline palladium species formed on the sup-port surfaces (Fig. 3). Regardless of the support composition, nopeaks of PdO or Pd0 crystallites were observed indicating theiramorphous character or that crystallite sizes are below the detec-tion limit of the XRD technique (<4 nm). The X-ray diffraction pat-terns of the Pd/Al and Pd/HY catalysts were typical of the Al2O3

(JCPDS 00-001-1303) and faujasite (JCPDS 00-0111-0672) phases,respectively. As expected, both Pd/10%HY–Al and Pd/20%HY–Alcatalysts showed both phases simultaneously. Thus, all Al-contain-ing catalysts exhibited typical diffraction lines of Al2O3 material at2h values of 37.5�, 39.4�, 45.8�, 60.8�, 66.9� and 85.0� whereas theHY-containing catalysts showed many narrow peaks at 2h values of6.1�, 10.3�, 12.2�, 15.6�, 18.4�, 20.3�, 23.7�, 27.0�, 31.8�, 34.2�, 38.4�and 54.6�, which are typical of zeolite faujasite. As expected, com-parison of XRD profiles of the zeolite peaks in the HY-containing

0 20 40 60 80 100 120

o

o

o

oo

oo

oo o

o

oo

o o *

**

*

*

*

Al2O3

Pd/Al

Pd/10%HY-Al

Line

ar c

ount

s (a

u)

2 Theta (º)

Pd/HY

Pd/20%HY-Al

66.9º

85.0º60.8º

45.8º

39.4º37.5º

*o

HY

Fig. 3. XRD diffraction patterns of Pd-based oxide precursors.

Table 2Hydrogen consumption during TPR of calcined Pd catalysts.

1066 S. Echeandia et al. / Fuel 117 (2014) 1061–1073

catalysts revealed an increase of their intensity with an increase ofHY content in the mixed HY–Al2O3 supports.

Catalyst Amount of H2 (mmol/gcat)

289 K 358 K 473–573 K Total

Pd/Al – �0.24 1.68 1.44Pd/10%HY–Al – �0.26 1.82 1.56Pd/20%HY–Al – �0.21 1.96 1.75Pd/HY 0.89 – 2.44 3.33

3.1.3. Temperature-programmed reductionTPR profiles of bare supports and oxide precursors are shown in

Fig. 4. As expected, contrary to bare supports, Pd catalysts showedH2-consumption in the temperature range 273–573 K. Thus, theH2-consumption peak at 289 K observed for Pd/HY catalysts isdue to reduction of highly dispersed PdOx species [53,54] whereasthe negative H2 consumption peak at 358 K detected on Pd/Al andthe two Pd(x)HY–Al catalysts is usually ascribed to the desorptionof hydrogen from a bulk palladium hydride formed through hydro-gen diffusion into Pd crystallites [55]. This behaviour is typical forcatalysts with low metal dispersion due to the high palladium con-centration in the large particles. The quantification of H2 consump-tion/evolution is listed in Table 2. From these data, the amount ofhydrogen desorbed from bulk palladium hydride follows the trend:Pd/10%HY–Al � Pd/Al > Pd/20%HY–Al� Pd/HY (none). The ab-sence of bulk b-PdH0.6 species on the surface of Pd/HY catalyst isin agreement with TEM data which indicate the formation of smal-ler size Pd0 particles on this sample.

The Pd catalysts exhibited two types of reduction peaks: one atlow-temperature, which is due to reduction of easy reducible pal-ladium oxide, and the other broad one located at somewhat highertemperature (473–573 K) due to reduction of a more stable PdOspecies interacting strongly with the support [56,57]. For the Pd/Al and Pd/10%HY–Al catalysts, the peak maxima appeared around501 and 511 K respectively, indicating similar crystallite sites. Forthe Pd/20%HY–Al and Pd/HY catalysts, this peak is very broadand shifted to lower temperature when compared with Pd/Al andPd/10%HY–Al catalysts. Indeed, considering that reduced Pd/HYsamples showed the lowest Pd0 particle among the catalysts stud-ied (HRTEM results vide infra) and that bulk Pd-hydride is notformed, the shift of its peak maxima toward lower temperature

200 300 400 500 600

(d)

(c)

(b)

x 10

x 10

x 10

Temperature (K)

H2 c

onsu

mpt

ion

(a.u

.)

Pd/Al

Pd/10%HY-Al

Pd/20%HY-Al

Pd/HY

358

511

424291

497

444

x 10

(a)

Fig. 4. TPR profiles of bare supports (doted line) and Pd–based oxide precursors(solid line) Pd/Al, (b) Pd/10%HY-Al, (c) Pd/20%HY-Al, (d) Pd/HY.

can be explained considering that PdO clusters are firstly reducedto form Pd atoms at lower temperature (H2 consumption at291 K) and then H2 is dissociated over these Pd atoms and thenthe H-atoms spilt over on the surface reduce the highly dispersedPd2+ species (H2 consumption at 424 K). Finally, an enhancedreduction of Pd species on the Pd/HY can be explained consideringthe H2 dissociation on reduced palladium particles and subsequenthydrogen migration to the carrier interface (spill-over effect).

3.2. Characterization of fresh reduced catalysts

3.2.1. TPD-NH3In order to compare the acidity of samples before and after Pd

incorporation, in Fig. 5 the TPD-NH3 profiles of the reduced Pd cat-alysts are compared with that of the corresponding bare supportspre-treated at the same conditions (H2/Ar reduction at 573 K for1 h). These profiles were mathematically fitted using Gaussianfunctions (not shown here). Based on the desorption temperature,acid sites were identified to possess weak (T < 523 K) and mediumstrength (523–773 K). Regardless of support, all bare supports andcalcined Pd catalysts show weak and medium strength acid siteswhich could be identified as Brønsted and structural type, respec-tively [58]. The concentration of acid sites (expressed asmmolNH3=gcat) is shown in Table 2. From these data, it is seen thatPd loading on acidic supports results in the loss of an importantfraction of acid sites. This is due to the loss of –OH groups uponPd2+ ions attachment to the support surface. The total acidity ofthe catalysts follows the trend: Pd/HY > Pd/20%HY–Al > Pd/10%HY–Al > Pd/Al. This trend clearly indicates that the incorpora-tion of HY zeolite into alumina carrier led to an increase acidity.

100 200 300 400 500 600

(c)

TCD

sig

nal (

a.u.

)

Temperature (ºC)

(d)

(a)

(b)

Fig. 5. TPD-NH3 profiles of bare supports (doted line) and reduced Pd-basedcatalysts (solid line): (a) Pd/Al, (b) Pd/10%HY-Al, (c) Pd/20%HY-Al, (d) Pd/HY.

S. Echeandia et al. / Fuel 117 (2014) 1061–1073 1067

3.2.2. HRTEMHRTEM was employed to reveal Pd distribution on the different

supports. The HRTEM images of the fresh reduced catalysts are dis-played in Fig. 6. Examination of HRTEM images show that, in gen-eral, Pd crystallites are heterogeneously distributed on the supportsurface but they become more homogeneous upon increasing zeo-lite content. The surface density of the Pd0 particles, expressed asPd0 particles per 50 nm2, is presented in Table 4. As seen in this ta-ble, the surface density of the Pd0 particles follows the trend: Pd/HY� Pd/Al > Pd/20%HY–Al > Pd/10%HY–Al (Table 4). Unfortu-nately, the poor contrast between HY and alumina substrates doesnot allow to unambiguously discriminate the preferential locationof palladium particles on either HY or alumina phase. In this sense,the EDX/SEM elemental analysis gave information that Pd is depos-ited preferentially on the HY surface rather than on alumina (datanot shown).

The influence of support on the Pd crystallite size distributionwas determined by statistical analysis of various HRTEM images(see Fig. S2 in the Supporting Information). For the all catalysts,the Pd particle sizes are distributed over the range of 2–10 nmyielding the average Pd size in the range 3.2–5.7 nm (Table 4).For the Pd particle size, the observed trend is: Pd/HY(3.2 ± 0.7 nm) < Pd/20%HY–Al (4.6 ± 1.2 nm) < Pd/10%HY–Al(5.0 ± 1.7 nm) < Pd/Al (5.7 ± 2.2 nm). Thus, in good agreement withTPR results (Fig. 4), the Pd/HY catalyst shows the lowest Pd particlesize among the catalysts studied. It is emphasized that informationon the Pd particle sizes obtained by TEM is different from XRD re-sults (Fig. 3). This is due to difference in the detection limit andassociated error to XRD and TEM measurements. The XRD tech-nique is not able detect regular crystal particles with size lowerthan 4 nm whereas the TEM technique can do it. In addition,

Pd/Al

Pd/20%HY-Al

Fig. 6. HRTEM of the pre-

TEM technique does not allow distinguish very small particles(<1 nm) from the zeolite background because of the lack of con-trast. It is also emphasized that the catalyst area irradiated bythe electron beam is several orders of magnitude than in case ofthe XRD analysis, therefore the Pd size calculated by statisticalanalysis may not be representative for the whole sample. We cansay that the number of small Pd clusters is underestimated byTEM and that this leads to an overestimation of the average Pdcrystallite size.

3.2.3. XPS and DRIFT spectroscopy of adsorbed COInformation on the Pd state and the surface composition of the

reduced (573 K, H2) catalysts was obtained by XPS analysis. Fig. 7shows the Pd 3d core-level spectra of all reduced catalysts and Ta-ble 4 summarizes the Pd 3d5/2 binding energies (BEs). For all zeo-lite-containing catalysts, the binding energy of Pd 3d5/2 core levelat 334.9 eV (335.0 eV for Pd/HY) is due to metallic Pd [58,59].The BE of the Pd3d5/2 peak in Pd/Al catalyst was 0.5 eV lower thanthat measured for Pd/10%HY–Al and Pd/20%HY–Al catalysts. ThisBE shift of Pd might involve Pd particle size effect. Consideringthe TPR results, the possible contribution of the metal-supportinteraction to the BE shift can be precluded.

CO adsorption followed by DRIFT spectroscopy was performedin order to identify the nature of surface sites on the supportedPd catalysts after their activation in pure hydrogen at 573 K. TheDRIFT-CO spectra (irreversible CO adsorption at room tempera-ture) are displayed in Fig. S3 in the Supporting Information. Ascompared with Al-containing samples, the Pd/HY catalyst showedsignificant differences in: (i) the frequency of the observed max-ima, (ii) the total integrated intensity of the bands, and (iii) the ra-tio between different components. For the pre-reduced Pd/HY, the

Pd/10%HY-Al

Pd/HY

reduced Pd catalysts.

0

20

40

60

80

100

Phen

ol c

onve

rsio

n (%

)

Temperature (K)

Pd/Al Pd/10HY-Al Pd/20HY-Al Pd/HY sulfided NiMo

Reference(a)

523 K 573 K 623 K

Pd/Al Pd/10%HY-Al Pd/20%HY-Al Pd/HY NiMo0,0

0,2

0,5

0,6

Yiel

ds o

f pro

duct

s at

523

K

Catalyst

Benzene Cyclohexene+cyclohexane Methylcyclopentane

(b)

Fig. 8. (a) Comparison of the steady-state phenol conversions at different temper-atures and yields of products (b) in HDO of phenol at 250 �C over supported Pdcatalysts and NiMo reference sample (P = 1.5 MPa, WHSV = 0.5 h�1).

1068 S. Echeandia et al. / Fuel 117 (2014) 1061–1073

DRIFT spectra of adsorbed CO suggest that palladium would be inan electron-deficient state. However, considering TPR results, theelectronic effects associated with a strong metal-support interac-tion are hardly possible. Thus, the upward shift of a band indicativeof linear CO adsorption on Pd/HY zeolite may be related rather tochanges in the Pd particle size. Interestingly, both Pd/HY and Pd/20%HY–Al showed a largest contribution from CO linearly bondedto metallic Pd species. For the Pd/Al, Pd/10%HY–Al and Pd/20%HY–Al catalysts, the lowering of 34 cm�1 of the CO-bridge frequencywith respect to Pd/HY (see Fig. S3 in the Supporting information(SI)) probably reflects the variation in the coordination of Pd atomsin terraces, steps or kink sites.

Quantitative XPS data of all catalysts are summarized in Table 4.Surface exposure of Pd species as evaluated by the Pd/(Si + Al)atomic ratio follows the trend: Pd/Al � Pd/10%HY–Al < Pd/20%HY–Al < Pd/HY. Therefore, both Pd/HY and Pd/20%HY–Al cata-lysts exhibit a lower palladium surface exposure than Pd/Al andPd/10%HY–Al ones. This difference can be also related to a higheramount of Pd species located within the pore structure of the HYzeolite. An additional evidence for this hypothesis comes fromthe linear correlation found between palladium surface exposure,determined by XPS, and the location of palladium species on thesupport surface, derived from the normalized BET area of catalysts(see Fig. S4 in the Supporting information (SI)).

3.3. Activity measurements

Fig. 8(a) shows the activities of the reduced Pd/Al, Pd/10%HY–Al, Pd/20%HY–Al and Pd/HY catalysts in the HDO of phenol at totalhydrogen pressure of 15 bar and three different temperatures (523,573 and 623 K). The activity of a commercial NiMo sulfide catalystused as reference is also included in this figure. Conversion data arethe average of the conversion achieved along 6 h of time on-stream(TOS) reaction at each temperature (during this time the conver-sion changes were <4%). It was found that the optimal reactiontemperature was 523 K. This may be related to a faster deactiva-tion by coking at higher temperatures. Because a low reaction tem-

330 335 340 345BE (eV)

Pd/HY

Pd/20%HY-Al

coun

ts p

er s

econ

d (a

u)

Pd/10%HY-Al

3d3/2

Pd3d

Pd/Al

3d5/2

Pd0

Fig. 7. Pd 3d3/2 core-level spectra of in-situ pre-reduced (300 �C, H2) Pd catalysts.

perature favours hydrogenation of the aromatic ring of O-containing compounds [25] leading to preferential phenol trans-formation via hydrogenation reaction route, the catalyst activitywas not studied at temperatures lower than 523 K. Indeed, only asmall amount of benzene was formed in the HDO of phenol at473 K and 4 MPa over a Pt/HY catalyst [46].

Regardless of the reaction temperature, the highest phenol con-version was achieved for the Pd catalyst supported on the 20% HY/Al2O3 catalyst. Incorporation of a smaller amount of HY (10 wt.%)into alumina generated less active catalysts, and when pure HYzeolite was used as support activity was the lowest with the excep-tion of the test at 623 K. At 523 K, the Pd/20%HY–Al catalystshowed similar phenol conversion than the sulfided NiMo/Al2O3–zeolite hydrocracking catalyst (63% vs. 62%) selected as reference(HCK).

Table 5 shows the selectivity obtained on the different Pd cata-lysts in HDO of phenol reaction at 523, 573 and 623 K. Regardlessof the reaction temperature, the only products detected in the HDOof phenol over all palladium catalysts were four O-free com-pounds: benzene, cyclohexene, cyclohexane and methylcyclopen-tane. Based on the selectivity data shown in Table 5, the possibleHDO reaction pathways for phenol HDO on Pd-based catalysts isshown in Scheme 1. The proposed reaction scheme is based notonly in product distributions but also in literature information[14–19 and references within]. For all catalysts, the HDO of phenolmight occur via two parallel pathways: a hydrogenation (HYD) ofphenol’s aromatic ring followed by cleavage of the C–O r bondleading to the formation of cyclohexane and a direct cleavage ofthe C–O r bond leading to the formation of benzene [14–19]. Thisreaction scheme suggests that the use of hydrogen for phenoltransformation over HY-containing Pd has two effects with respect

Scheme 1. Scheme of the phenol HDO reaction over Pd/Al, Pd/10%HY–Al, Pd/20%HY–Al and Pd/HY catalysts.

S. Echeandia et al. / Fuel 117 (2014) 1061–1073 1069

to reaction mechanism: saturating double bonds and removingoxygen. Because cyclohexanol was not detected on all catalystsstudied, in the reaction Scheme 1 this product is considered asan intermediate which transforms into cyclohexene and cyclohex-ane. Since all catalysts showed 100% selectivity toward O-freeproducts, the hydrogen consumption for saturation of doublebonds could be deduced from selectivity toward cyclohexane for-mation. In the HDO reaction at 523 K, selectivity toward cyclohex-ane follows the trend: Pd/10%HY–Al (97%) � Pd/20%HY–Al(95%)� Pd/Al (70%)� Pd/HY (55%) indicating that both Pd cata-lysts supported on hybrid substrates show larger hydrogen con-sumption for saturation of double bonds. While testing the Pd/HY catalyst, relatively large quantity of methylcyclopentane (onlyat 573 and 623 K) was detected. This product could be originatedas a result of combined effect of the high hydrogenation/dehydro-genation activity of the Pd and the acidity of the support.

Data collected in Table 5 indicates that there is a strong effect ofsupport (Al2O3, mixed HY–Al2O3 and HY zeolite) on product selec-tivity. Regardless of the reaction temperature, product distributionobserved for the Pd catalysts supported on mixed supports(10 wt.% and 20 wt.% of HY) yielded a high proportions of cyclohex-ane + cyclohexene (P93%), with small quantities of benzene (inthe range 1–7%) that increased at higher temperatures. It seemsthat for these HY loading used in these catalysts there is not signif-icant isomerisation activity to produce methylcyclopentane eventhough they presented high HDO activity. Similarly, NiMo sulfidecatalyst tested at 523 K showed a large selectivity toward cyclo-hexane (50%) and benzene (46%), and a low selectivity towardmethylcylopentane (3%). Comparison of the selectivity of the Pd-based catalysts with that of the sulfide NiMo catalyst indicates alarge benzene formation on the latter catalyst which in turn isslightly higher than that of Pd/HY counterpart (46% vs. 38%). Tak-ing into account the high cost of hydrogen, a large benzene forma-tion should be considered as a positive feature of the Pd/HYcatalyst. Contrary to our study, direct hydrogenolysis of phenolto benzene was inhibited on acidic Pd/C catalysts having the com-bination of a hydrogenation catalyst and a strong acid, which weretested in aqueous-phase HDO of bio-derived phenols to cycloal-kanes [31,32]. However, the comparison of the activity results re-ported in literature for supported Pd catalysts is difficult, if notimpossible, because different reactors, reaction conditions, sol-vents and different catalyst formulations have been employed.

Main products obtained on the Pd/Al catalyst are cyclohexaneand cyclohexene, and upon increasing temperature higheramounts of benzene were formed. This is likely due to the factthat the benzene–cyclohexene–cyclohexane equilibrium suffers

significant thermodynamic limitations depending on the tempera-ture (at higher temperatures benzene formation is favoured). An-other reason could be the variation of the hydrogenation activityof the catalyst sites associated to their deactivation–modificationalong the time-on-stream. Similarly to Pd/Al, Pd/HY catalyst showsstrong selectivity-temperature dependence. On the contrary, thestability in the product distribution of the Pd/20%HY–Al catalystis remarkable even if it suffered some deactivation along the tests.

Another interesting comparison is the product yields obtainedat 523 K for the Pd-based catalysts and the sulfided NiMo/Al2O3–zeolite one (Fig. 8(b)). Both Pd/20%HY–Al and sulfided NiMo cata-lysts produced similar yields of O-free products. However, the Pd/HY catalyst produced exclusively cyclohexene and cyclohexanewhereas the reference one yields equal amount of benzene andcyclohexene + cyclohexane. Thus, it can be concluded that lesshydrogen is required for phenol transformation over commercialNiMo catalyst than on the Pd/20%HY–Al one. Finally, the Pd/HYcatalyst showed the lowest yield of all O-free products linked withthe lowest phenol conversion on this catalyst.

3.4. Catalyst deactivation by coke

In order to shed light on the origin of catalyst deactivation dur-ing phenol HDO over reduced Pd catalysts, the quantity of cokedeposited was evaluated by TGA analysis. The amount of cokeformed in the spent catalysts after 16 h on stream was evaluatedby temperature-programmed oxidation in a 20% O2/N2 mixture.The weight change during oxidation together with DTG profilesare shown in Fig. 9. Regardless of the support, all DTG profilesshow one intense peak at temperatures below 773 K, which isdue to burning of carbonaceous residues. For the Pd/10%HY–Aland Pd/20%HY–Al catalysts, this peak shows a shoulder locatedon the higher temperature region indicating the presence of morepolymerized-type of coke species. On the contrary, the DTG peak ofspent Pd/HY catalyst shows some shoulder at lower temperaturessuggesting the formation of carbonaceous residues with higher de-gree of hydrogenation, which are easier to burn off. The percent-ages of total mass loss corresponding to all types of burnt cokespecies follows the trend: Pd/HY (16.3%)� Pd/20%HY–Al (8.4%) > Pd/10%HY–Al (7.1%) > Pd/Al (6.4%).

3.5. Catalyst activity-structure correlation

In this study, mixed alumina–zeolite HY was successfully usedas supports for the catalytic response of palladium catalysts inthe phenol HDO reaction. Under the reaction conditions employed

0 200 400 600 800 100080

85

90

95

100

Temperature (ºC)

dw/dT [m

g/ºC]

Pd/10%HY-Al

438 ºC

Coke: 7.1%

-0.06

-0.04

-0.02

0.00

0 200 400 600 800 100080

85

90

95

100 Pd/20%HY-Al

Temperature (ºC)

Mas

s lo

ss (%

)M

ass

loss

(%)

Mas

s lo

ss (%

)M

ass

loss

(%)

Coke: 8.4%

-0.06

-0.04

-0.02

0.00

dw/dT [m

g/ºC]

0 200 400 600 800 100080

85

90

95

100

Temperature (ºC)

Coke: 6.4%

436 ºC-0.06

-0.04

-0.02

0.00

Pd/Al

dw/dT [m

g/ºC]

0 200 400 600 800 100080

85

90

95

100

Temperatue (ºC)

Coke: 16.3%

-0.25

-0.20

-0.15

-0.10

-0.05

0.00

dw/dT [m

g/ºC]

Pd/HY

502 ºC

Fig. 9. TPO/TGA profiles of spent palladium catalyst.

1070 S. Echeandia et al. / Fuel 117 (2014) 1061–1073

(T = 523–623 K, P = 15 bar, WHSV = 0.5 h�1), the best catalytic re-sponse was achieved on the Pd catalyst supported on mixed20%HY–80%Al2O3 carrier. At a first glance, three interesting obser-vations can be put forward to describe the most interesting fea-tures of this work:

(1) Upon the reaction conditions studied, the phenol transfor-mation over Pd/Al2O3, Pd/HY–Al2O3 and Pd/HY catalystsusing n-octane as solvent proceeds according to the samereaction routes as those observed for sulfided NiMo/Al2O3

catalysts [14–19]: (i) direct hydrogenolysis to benzene, and(ii) hydrogenation to cylohexanol followed by direct cyclo-hexanol hydrogenolysis to cyclohexane.

(2) All catalysts studied shows 100% selectivity toward O-freeproducts

(3) The large extent of benzene formation on the Pd/HY cata-lysts suggests that the excessive H2-consumption could beavoided using HY carrier. This catalyst shows hydrocrackingand isomerization functions, as revealed by the methylcycl-opentane formation.

Considering the catalyst’s characterization data, several expla-nations can be advanced to explain the greater intrinsic activityof the Pd/20%HY–Al catalyst in the HDO of phenol at steady-state:(i), the optimized support acidity, as evidenced by TPD-NH3

measurements, which led to an increase of the catalyst activitywithout excessive coking; (ii), an increase of specific BET surfacearea; (iii), additional activity from acid sites [60], and (iv) a largepalladium dispersion, as evidenced by HRTEM and TPR results.The catalysts showing the best Pd dispersion, showed the largestactivity per palladium surface atom. This suggests that the phenolHDO reaction could be a structure sensitive reaction, i.e. specificactivity might to depend on the details of the surface structure ofthe palladium crystallites.

Contrary to the basic supports, H-spillover is a common featureobserved on Pd catalysts supported on acidic substrates which con-tributes to enhance HDO activity [25]. This is a key requirementneeded for the effective C–O bond cleavage of O-containing com-pounds. In this sense, a direct correlation between the supportacidity and the HDO activity of Rh catalysts supported on differentcarriers was observed [25]. Since Rh/ZrO2 sample was found to bemuch less active than its counterparts supported on more acidicsubstrates (alumina, silica–alumina and nitric-acid treated carbonblack), the authors concluded that bifunctional catalysts areneeded for the effective O-removal from phenol.

It is generally accepted that the HDO reaction over noble metalssupported on acidic substrates occurs on metal sites as well as onacid sites being H2 dissociated on metal sites, while O-containingcompounds are adsorbed and activated on either metal sites oron the cations/oxygen vacancies located at the metal-support

Table 3Acid properties of pure supportsa and reduced Pd/Al, Pd/HY–Al and Pd/HY catalysts asdetermined by TPD-NH3.

Catalyst Amount of acid sites ðmmol NH3 g�1catÞ

Weak T < 523 K Moderate T = 523–773 K Total

Pd/Al 0.9 (0.9) 0.4 (1.1) 1.3 (2.0)Pd/10%HY–Al 0.9 (1.8) 0.8 (1.9) 1.7 (3.7)Pd/20%HY–Al 1.8 (1.8) 0.7 (3.6) 2.5 (5.4)Pd/HY 1.8 (3.4) 3.0 (7.7) 4.8 (11.1)

a Data of the bare supports are given in parenthesis.

S. Echeandia et al. / Fuel 117 (2014) 1061–1073 1071

interface [14–19]. In this work, the possible contribution of theH-spillover to the overall HDO activity should be obtained from acorrelation between the catalyst acidity and HDO activity. Indeed,with exception of Pd/HY, the total acidity (Table 3) and HDOactivity (Table 5) of the catalysts follows the same trend: Pd/20%HY–Al > Pd/10%HY–Al > Pd/Al. The activity and selectivity re-sults reported in this work are in good agreement with literaturereports on the Pd catalysts supported on acidic carriers indicatingthat a bifunctional mechanism is involved in HDO reaction over allPd-based catalysts. This is also confirmed by analyzing products ofHDO reaction over Pd catalysts supported on non-acidic carriers[41,42]. For example, in the vapour phase hydrogenation of phenolover Pd supported on non-acidic substrates (alumina and LTL zeo-lite), the major products detected in the reaction at T = 523 K and20 bar H2 pressure were cyclohexanone and cyclohexanol withcyclohexene, cyclohexane and benzene as minor products [42].Thus, the catalysts supported on non-acidic carriers are good sup-ports if the objective of HDO reaction is cyclohexanone. The rate-determining step of the target reaction over Pd catalysts supportedon non-acidic carriers is probably a surface reaction between thestrongly bond phenol and the weakly adsorbed H-atom, as it wasderived from a kinetic study of phenol hydrogenation over Pd/MgO catalysts [41].

Table 4Binding energy (eV) and surface atomic ratios (from XPS), and HRTEM parameters (statist

Catalyst XPS

Pd 3d5/2 (eV) Si/(Si + Al) at Pd/(Si + Al) at

Pd/Al 334.5 – 0.0140Pd/10%HY–Al 334.9 0.044 0.0139Pd/20%HY–Al 334.9 0.092 0.0084Pd/HY 335.0 0.654 0.0024

Table 5Influence of temperature on the phenol conversion and selectivity in HDO of phenola ove

Catalyst T (K) Conv. (%) Selectivit

Benzene

Pd/Al 523 28 0573 35 2623 22 13

Pd/10%HY–Al 523 21 1573 22 2623 29 4

Pd/20%HY–Al 523 63 4573 57 7623 46 7

Pd/HY 523 11 38573 20 25623 35 42

Sulfided NiMob 523 62 46

a Reaction conditions were: T = 523–623 K, P = 1.5 MPa, WHSV = 0.5 h�1.b A commercial NiMo/Al2O3–zeolite sulfide catalyst (HCK).

In good agreement with recent reports on the supported PdHDO catalysts [14–19 and references within], product distributionrecorded in the present work suggests that Pd metal sites can beresponsible of phenol hydrogenation to cyclohexanol followed byits transformation into cyclohexane and finally dehydrogenatedto form benzene. The Pd/HY catalyst tested at 523 K shows excep-tionally high selectivity to benzene (38%) indicating a large O-elim-ination from phenol via hydrogenolysis reaction route (seeScheme 1). This result appears quite interesting because such reac-tion required less hydrogen consumption. Considering the highestdensity of Pd sites on the surface of Pd/HY sample (Table 4) and itslargest acidity among the catalysts studied (Table 3), it appearsreasonably assume that the couple of Pd and acid sites in closevicinity are responsible for the higher methylcylopentane forma-tion observed using this catalyst. However, hydrocracking andisomerization are not necessary catalyst functions because totaloxygen elimination was achieved for all catalysts studied.

Finally, TPO experiments confirmed that the main cause of cat-alyst deactivation in phenol HDO is deposition of carbonaceousresidues/coke on the catalyst surface showing the Pd/HY catalystthe largest amount of coke among the catalysts studied. For thiscatalyst, an increase of the selectivity toward benzene formationwith an increase of reaction temperature is accompanied by astrong decrease of the phenol conversion (Table 5). Consideringthe large amount of coke formed (Fig. 9), the increase of the selec-tivity toward benzene is due to a decrease of the catalyst hydroge-nation ability linked with coke formation on the metal sites. Theamount of coke is plotted in Fig. 10 against the catalyst total acid-ity. From this figure it is obvious that the catalyst acidity is a deci-sive factor influencing on coke formation, as expected [61,62].Thus, the larger extent of coke formed on the Pd/HY catalyst couldbe due to its larger surface concentration of acid sites. Thus,although the Pd/HY catalyst can be effective for the HDO reaction,it can be concluded that the density of its acid sites needs to beoptimized; large enough to achieve high conversion of phenolic

ics analysis) of the reduced (573 K) Pd/Al, Pd/HY–Al and Pd/HY catalysts.

TEM

Particle size (nm) Surface density of Pd0 particles per 50 nm2

5.7 ± 2.2 26 ± 55.0 ± 1.7 11 ± 24.6 ± 1.2 20 ± 63.2 ± 0.7 48 ± 6

r Pd/Al, Pd/HY–Al, Pd/HY catalysts.

y (%)

Cyclohexene + cyclohexane Methyl-cyclo-pentane

100 098 085 2

99 097 195 1

95 193 093 0

62 058 1725 33

50 3

0 1 2 3 4 50

5

10

15

20C

oke

(%)

Total catalyst acidity (mmol NH3 gcat-1)

R=0.99

Fig. 10. Influence of total acidity of the oxide precursors (from TPD-NH3) on thecoke formation in HDO of phenol over supported Pd catalysts (T = 250-350 �C,P = 1.5 MPa and WHSV = 0.5 h�1).

1072 S. Echeandia et al. / Fuel 117 (2014) 1061–1073

compounds, but not too high that would result in fast deactivation.The lowering of catalyst activity due to undesirable oxygenationfunction of residual chloride ions could be excluded since bothXPS and EDX/SEM techniques did not detect the chloride impurityin the reduced catalysts.

4. Conclusions

� As a general conclusion, the use of mixed HY–Al2O3 (20 wt.% ofHY) as support is beneficial for HDO reaction over palladium-based catalysts.� Conversion results showed that, upon reaction conditions

employed (flow reactor, T = 523 K, PH2 = 15 bar, solvent:n-octane; WHSV = 0.5 h�1), the HDO of phenol over Pd/20%HY–Al catalyst led to similar yields of O-free products asthose obtained using the sulfided NiMo/Al2O3–zeolite hydro-cracking catalyst.� The largest activity of Pd/20%HY–Al catalyst in the HDO of phenol

was explained as due to the hydrogen spillover phenomenon andlargest active phase dispersion on the carrier surface. Catalystsfurther developed from this one could be used in the final stageof bio-liquid transformation into high quality and O-free bio-oil.� A drastic decrease in activity going from Pd/20%HY–Al to Pd/HY

sample indicated that the high catalyst acidity would result infast catalyst deactivation.

Acknowledgements

Financial support by the Community of Madrid (Spain) andEuropean Union (Project S2009/ENE-1743) is gratefully acknowl-edged. S.E. acknowledges financial support from the Spanish Min-istry of Science and Innovation (ENE2010-21198-C04-03) and theBasque Autonomous Government (IE09-263-IE10-288-VALCAPEF).The authors acknowledge the technical assistance of Dr. M.A. Peñain TPO/TG measurements.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.fuel.2013.10.011.

References

[1] Bridgwater AV, Meier D, Radlein D. An overview of fast pyrolysis of biomass.Org Geochem 1999;30(12):1479–93.

[2] Demirbas� A. Biomass resource facilities and biomass conversion processing forfuels and chemicals. Energy Convers Manage 2001;42(11):1357–78.

[3] Sarkar S, Kumar A. Large-scale biohydrogen production from bio-oil. BioresourTechnol 2010;101:7350–61.

[4] Mohan D, Pittman Jr CU, Steele PH. Pyrolysis of wood/biomass for bio-oil: acritical review. Energy Fuels 2006;20:848–89.

[5] Grange P, Laurent E, Maggi R, Centeno A, Delmon B. Hydrotreatment ofpyrolysis oils from biomass: reactivity of the various categories of oxygenatedcompounds and preliminary techno-economical study. Catal Today1996;29:297–301.

[6] Oasmaa A, Kuoppala E, Solantausta Y. Fast pyrolysis of forestry residue. 2.Physicochemical composition of product liquid. Energy Fuels 2003;17:433–43.

[7] Furimsky E. Catalytic hydrodeoxygenation. Appl Catal A: Gen2000;199:147–90.

[8] Senol O_I, Viljava T-R, Krause AOI. Hydrodeoxygenation of methyl esters onsulphided NiMo/gamma-Al2O3 and CoMo/gamma-Al2O3 catalysts. Catal Today2005;106:186–9.

[9] Laurent E, Delmon B. Study of the hydrodeoxygenation of carbonyl, carylic andguaiacyl groups over sulfided CoMo/c-Al2O3 and NiMo/c-Al2O3 catalyst: II.Influence of water, ammonia and hydrogen sulfide. Appl Catal A1994;109:97–115.

[10] Ferrari M, Bosmans S, Maggi R, Delmon B, Grange P. CoMo/carbonhydrodeoxygenation catalysts: influence of the hydrogen sulfide partialpressure and of the sulfidation temperature. Catal Today 2001;65:257–64.

[11] Senol O_I, Viljava T-R, Krause AOI. Effect of sulphiding agents on thehydrodeoxygenation of aliphatic esters on sulphided catalysts. Appl Catal A:Gen 2007;26:236–44.

[12] Viljava T-R, Komulainen RS, Selvam T, Krause AOI. Stability of CoMo/Al2O3

catalysts: effect of HDO cycles on HDS. Stud Surf Sci Catal 1999;127:145–52.[13] Navarro RM, Pawelec B, Castaño P, Álvarez-Galván MC, Loricera CV, Fierro JLG.

Upgrading of bio-liquid on different mesoporous silica-supported CoMocatalysts. Appl Catal, B 2009;92:154–67.

[14] Huber GW, Iborra S, Corma A. Synthesis of transportation fuels from biomass:chemistry, catalysts, and engineering. Chem Rev 2006;106:4044–98.

[15] Elliot DC. Historical developments in hydroprocessing bio-oils. Energy Fuels2007;21:1792–815.

[16] Mortensen PM, Grunwldt J-D, Jensen PA, Knudsen KG, Jensen AD. A review ofcatalytic upgrading of bio-oil to engine fuels. Appl Catal A 2011;407:1–19.

[17] He Z, Wang X. Hydrodeoxygenation of model compounds and catalyticsystems for pyrolysis bio-oils upgrading. Catal Sustain Energy 2012;1:28–52.

[18] Butler E, Devlin G, Meier D, McDonnell K. A review of recent laboratoryresearch and commercial developments in fast pyrolysis and upgrading.Renew Sustain Energy Rev 2011;15(8):4171–86.

[19] Bu Q, Lei H, Zacher AH, Wang L, Ren S, Liang J, et al. A review of catalytichydrodeoxygenation of lignin-derived phenols from biomass pyrolysis.Bioresour Technol 2012;124:470–7.

[20] Elliott DC, Hu J, Hart TR, Neuenschwander GG. US Pattent WO 2008/151269A2.

[21] Lin Y-C, Li C-L, Wan H-P, Lee H-T, Liu C-F. Catalytic hydrodeoxygenation ofguaiacol on Rh-based and sulfided CoMo and NiMo catalysts. Energy Fuels2011;25:890–5.

[22] Gutierrez A, Kaila RK, Honkela ML, Slioor R, Krause AOI. Hydrodeoxygenationof guaiacol on noble metal catalysts. Catal Today 2009;147:239–46.

[23] Ardiyanti AR, Gutierrez A, Honkela ML, Krause AOI, Heeres HJ. Hydrotreatmentof wood-based pyrolisis oil using zirconia-supported mono- and bimetallic (Pt,Pd, Rh) catalysts. Appl Catal A 2011;407:56–66.

[24] Veriansyah B, Han JY, Kim SK, Hong S-A, Kim YJ, Lim JS, et al. Production ofrenewable diesel by hydroprocessing of soybean oil: effect of catalysts. Fuel2012;94:578–85.

[25] Lee CR, Yoon JS, Suh Y-W, Choi J-W, Ha J-M, Suh DJ, et al. Catalytic roles ofmetals and supports on hydrodeoxygenation of lignin monomer guaiacol.Catal Commun 2012;17:54–8.

[26] Pham T, Lobban LL, Resasco DE, Mallinson RG. Hydrogenation andhydrodeoxygenation of 2-methyl-2-pentenal on supported metal catalysts. JCatal 2009;266:9–14.

[27] Sitthisa S, Resasco DE. Hydrodeoxygenation of furfural over supported metalcatalysts: a comparative study of Cu, Pd and Ni. Catal Lett 2011;141:784–91.

[28] Sitthisa S, Pham T, Prasomsri T, Sooknoi T, Mallinson RG, Resasco DE.Conversion of furfural and 2-methylpentanal on Pd/SiO2 and Pd–Cu/SiO2

catalysts. J Catal 2011;280:17–27.[29] Yu W, Tang Y, Mo L, Chen P, Lou H, Zheng X. Bifunctional Pd/Al-SBA-15

catalyzed one-step hydrogenation–esterification of furfural and acetic acid: amodel reaction for catalytic upgrading of bio-oil. Catal Commun2011;13(1):35–9.

[30] Kikhtyanin OV, Rubanov AE, Ayupov AB, Echevsky GV. Hydroconversion ofsunflower oil on Pd/SAPO-31 catalyst. Fuel 2010;89:3085–92.

[31] Zhao C, Kou Y, Lemonidou AA, Li XB, Lercher JA. Highly selective catalyticconversion of phenolic bio-oil to alkanes. Angew Chem Int Ed2009;48:3987–90.

[32] Zhao C, He J, Lemonidou AA, Li X, Lercher JA. Aqueous-phasehydrodeoxygenation of bio-derived phenols to cycloalkanes. J Catal2011;280:8–16.

[33] Parsell TH, Owen BC, Klein I, Jarrell TM, Marcum CL, Haupert LJ, et al. Cleveageand hydrogenation (HDO) of C–O bonds relevant to lignin conversion using Pd/Zn synergetic catalysts. Chem Sci 2013;4:806–13.

[34] Wildschut J, Mahfud FH, Venderbosch RH, Heeres HJ. Hydrotreatment of fastpyrolysis oil using heterogeneous noble metal catalysts. Ind Eng Chem Res2009;48:10324–34.

S. Echeandia et al. / Fuel 117 (2014) 1061–1073 1073

[35] Sun J, Karim AM, Zhang H, Kovarik L, Shari Li X, Hensley AJ. Carbon-supportedbimetallic Pd–Fe catalysts for vapor-phase hydrodeoxygenation of guaiacol. JCatal 2013;306:47–57.

[36] Snåre M, Kubicková I, Mäki-Arvela P, Chichova D, Eränen K, Yu D. Catalyticdeoxygenation of unsaturated renewable feedstocks for production of dieselhydrocarbons. Fuel 2008;87:933–45.

[37] Bejblová M, Zámostny P, Cerveny L, Cejka J. Hydrodeoxygenation ofbenzophenone on Pd catalysts. Appl Catal A 2005;296:169–75.

[38] Procházková D, Zámostny P, Bejblová M, Cerveny L, Cejka J.Hydrodeoxygenation of aldehydes catalyzed by supported palladiumcatalysts. Appl Catal A 2007;332:56–64.

[39] Elliot DC, Hart TR. Catalytic hydroprocessing of chemical models for bio-oil.Energy Fuels 2009;23:631–7.

[40] Tang Z, Lu Q, Zhang Y, Zhu X, Guo Q. One step bio-oil upgrading throughhydrotreatment, esterification, and cracking. Ind Eng Chem Res2009;48:69–6929.

[41] Mahata N, Vishwanathan V. Kinetics of phenol hydrogenation over supportedpalladium catalyst. J Mol Catal A: Chem 1997;120:267–70.

[42] Talukdar K, Bahattacharyya KG, Sivasanker S. Hydrogenation of phenol oversupported platinum and palladium catalysts. Appl Catal A 1993;96:229–39.

[43] Nimmanwudipong T, Runnebaum RC, Block DE, Gates BC. Cyclohexanoneconversion catalyzed by Pt/gamma-alumina: evidence of oxygen removal andcoupling reactions. Catal Lett 2011;141:1072–8.

[44] Zhu X, Lobban LL, Mallinson RG, Resasco DE. Bifunctional transalkylation andhydrodeoxygenation of anisole over a Pt/HBeta catalyst. J Catal2011;281(1):21–9.

[45] Hong D-Y, Agrawal PK, Miller SJ, Jones CW. Hydrodeoxygenation of phenolover zeolite-supported metal catalysts. 2009 AIChE annual meeting. Catalysisand Reaction Engineering Division.

[46] Hong D-Y, Miller SJ, Agrawal PK, Jones CW. Hydrodeoxygenation and couplingof aqueous phenolics over bifuncional zeolite-supported metal catalysts. ChemCommun 2010;46:1038–40.

[47] Navarro RM, Pawelec B, Trejo JM, Mariscal R, Fierro JLG. Hydrogenation ofaromatics on sulfur-resistant PtPd bimetallic catalysts. J Catal2000;189:184–94.

[48] Egia B, Cambra JF, Güemez B, Arias PL, Pawelec B, Fierro JLG. Hydrocrackingactivity of NiMo-USY zeolite hydrotreating catalysts. Stud Surf Sci Catal1997;106:567–71.

[49] Gregg SJ, Sing KSW. Adsorption, surface area and porosity. 2nd ed. NewYork: Academic Press; 1982.

[50] Leofanti G, Padovan M, Tozzola G, Venturelli B. Surface area and pore texture ofcatalysts. Catal Today 1998;41:207–19.

[51] Lin B, Wang R, Yu X, Lin J, Xie F, Wei K. Physicochemical characterization andH2-TPD study of alumina supported ruthenium catalysts. Catal Lett2008;124:178–84.

[52] Okal J, Zawadzki M. Influence of catalyst pretreatments on propane oxidationover Ru/c-Al2O3. Catal Lett 2009;132:225–34.

[53] Luo MF, Zheng XM. Redox behaviour and catalytic properties of Ce0.5Zr0.5O2-supported palladium catalysts. Appl Catal A 1999;189:15–21.

[54] Chang TC, Chen JJ, Yeh CT. Temperature-programmed reduction andtemperature-resolved sorption studies of strong metal-support interaction insupported palladium catalysts. J Catal 1985;96:51–7.

[55] Vannice MA, Chou P. Benzene hydrogenation over supported and unsupportedpalladium: I. Kinetic behavior. J Catal 1987;107(1):129–39.

[56] Lieske H, Völter J. Palladium redispersion by spreading of palladium(II) oxidein oxygen treated palladium/alumina. J Phys Chem 1985;89(10):1841–2.

[57] Gopinath R, Babu NS, Kumar JV, Lingaiah N, Sai Prasad PS. Influence of Pdprecursor and method of preparation on hydrodechlorination activity ofalumina supported palladium catalysts. Catal Lett 2008;120:312–9.

[58] Pawelec B, La Parola V, Thomas S, Fierro JLG. Enhancement of naphthalenehydrogenation over PtPd/SiO2–Al2O3 catalyst modified by gold. J Mol Catal A:Chem 2006;253:30–43.

[59] Murcia-Mascarós S, Pawelec B, Fierro JLG. Aromatics hydrogenation on Pt–Pdmetals supported on Zr-phosphate. Catal Commun 2002;3:305–11.

[60] Pawelec B, Mariscal R, Navarro RM, van Bokhorst S, Rojas S, Fierro JLG.Hydrogenation of aromatics over supported Pt–Pd catalysts. Appl Catal A2002;225:223–37.

[61] Arias PL, Zugazaga F, Güemez B, Cambra JL, Pawelec B, Fierro JLG.Hydrodesulfurization of diesel oil on Pt and NiW phases supported onalumina–zeolite mixtures. In: Symposium on recient advances in heteroatomremoval presented before the Division of Petroleum Chemistry, Inc. 215thNational meeting, Americam Chemical Society, Dallas, TX, March 29–April 3,1998. ACS symposium series, vol. 43; 1998. p. 63–7.

[62] Castaño P, Elordi G, Olazar M, Aguayo AT, Pawelec B, Bilbao J. Insights into thecoke deposited on HZSM-5, Hb and HY zeolites during the cracking ofpolyethylene. Appl Catal B 2011;104:91–100.