Applied Catalysis A: General 439– 440 (2012) 111– 124
Contents lists available at SciVerse ScienceDirect
Applied Catalysis A: General
j ourna l ho me page: www.elsev ier .com/ locate /apcata
ydrodeoxygenation of guaiacol over carbon-supported molybdenum nitrideatalysts: Effects of nitriding methods and support properties
. Tyrone Ghampsona,b , Catherine Sepúlvedac , Rafael Garciac , Ljubisa R. Radovicd,e , J.L. García Fierro f ,illiam J. DeSistoa,g,∗, Nestor Escalonac,∗∗
Department of Chemical and Biological Engineering, University of Maine, Orono, ME 04469, United StatesUnidad de Desarrollo Tecnológico, Universidad de Concepcicn, Casilla 4051, Concepcicn, ChileUniversidad de Concepcion, Facultad de Ciencias Quimicas, Casilla 160c, Concepcicn, ChilePenn State University, University Park, PA 16802, United StatesUniversidad de Concepcicn, Facultad de Ingenieria, Dept. Ing. Quimica, Concepcicn, ChileInstituto de Catalisis y Petroquimica, CSIC, Cantoblanco, 28049 Madrid, SpainForest Bioproducts Research Institute, University of Maine, Orono, ME 04469, United States
r t i c l e i n f o
rticle history:eceived 25 April 2012eceived in revised form 25 June 2012ccepted 28 June 2012vailable online 5 July 2012
eywords:
a b s t r a c t
Molybdenum nitride catalysts supported on activated carbon materials with different textural and chem-ical properties were synthesized by nitriding supported Mo oxide precursors with gaseous NH3 or N2/H2
mixtures using a temperature-programmed reaction. The supports and catalysts were characterized byN2 physisorption, XRD, chemical analysis, TPD, FT-IR and XPS. Guaiacol (2-methoxyphenol) hydrodeoxy-genation (HDO) activities at 5 MPa and 300 ◦C were evaluated in a batch autoclave reactor. Molybdenumnitrides prepared using a N2/H2 mixture resulted in more highly dispersed catalysts, and consequently
more active catalysts, relative to those prepared using ammonolysis. The HDO activity was also related topore size distribution and the concentration of oxygen-containing surface groups of the different carbonsupports. Increased mesoporosity is argued to have facilitated the access to active sites while increasedsurface acidity enhanced their catalytic activity through modification of their electronic properties. Thehighest activity was thus attributed to the highest dispersion of the unsaturated catalyst species and thehighest support mesoporosity. Surprisingly, addition of Co did not improve the HDO activity.
Due to long-term economic and environmental concerns, bio-oilerived from pyrolysis of woody biomass has received consider-ble attention as an alternative renewable feedstock to crude oilor the production of fuels and value-added chemicals [1]. Its uti-ization as fuel is limited, however, by high viscosity, low heatingalue, incomplete volatility and thermal instability, which stemrom the relative abundance of oxygenated organic compounds
2]. Catalytic hydrodeoxygenation (HDO) reactions are typicallyerformed to refine bio-oil and increase its quality as transporta-ion fuel. There are two significant challenges in this process: (i)
∗ Corresponding author at: Department of Chemical and Biological Engineering,niversity of Maine, Orono, ME 04469, United States. Tel.: +1 207 581 2291;
prevention of coke formation/catalyst deactivation and (ii) selec-tive removal of oxygen without excessive hydrogenation ofaromatic and olefinic compounds [2,3].
Model compounds have been used to mimic HDO studies ofbio-oil components in an effort to understand the role and fate ofdifferent functional groups present in the feed, as well as provideadditional insight into the development of improved catalysts andprocesses [3]. Guaiacol (2-methoxyphenol) is commonly used as amodel compound for HDO studies to represent the large numberof mono- and dimethoxy phenols present in bio-oil [4]; it is knownto be a precursor to catechol but it also subsequently forms coke[5,6]. Also, guaiacol possesses two different oxygenated functionalgroups ( OCH3 and OH) which make it challenging to achievecomplete deoxygenation [7].
Heterogeneous catalysts commonly studied for HDO of guaia-col (and many other model compounds) are conventional sulfided
Co(Ni)Mo/�-Al2O3 [5,8] and supported noble metal catalysts suchas Ru, Rh and Pd [9,10]. The initial interest in the metal sulfideswas driven by high cost and lack of selective HDO activity ofthe noble metal catalysts. Despite the high catalytic activity for
uaiacol conversion, there are some drawbacks associated withhese sulfide catalysts: (i) The alumina support can be unstable inater at processing conditions. (ii) The sulfide catalyst can oxi-ize under processing conditions, requiring in situ regenerationith a sulfiding agent to prolong catalyst activity; this regenera-
ion can contaminate products [5,6,11,12]. (iii) The acidic naturef the alumina support is known to be the cause of substantialoke deposition and rapid catalyst deactivation [13]. These draw-acks prompted interest in less acidic materials such as silica [14],irconia [9,15] and activated carbon [16,17] as catalyst supports.enteno et al. [14] reported that despite the lower activity ofetal sulfides supported on silica and carbon compared with con-
entional alumina-supported counterparts, the use of alternativeupports led to negligible coke formation. Furthermore, studiesnvolving carbon- and zirconia-supported catalysts indicated directlimination of the methoxy group which favored direct productionf phenol from guaiacol [14,18]. Particularly, based on their supe-ior performance in hydrodesulfurization processes, carbons arenown to be promising supports for the HDO of bio-oil [19–21].
Interest in carbon supports has increased mainly due to itsemarkable flexibility and the ability to recover active metal afteratalyst deactivation [20]. For HDO reactions, such deactivation inhe presence of water could be limited due to the hydrophobic char-cter of the carbon surface [22]. On the other hand, the weakernteraction between the support surface and the active metal mayesult in a lower dispersion of the sulfide phase [14,20]. The car-on surface can be ‘decorated’ with oxygen functionalities; this
mproves catalytic activity by facilitating a higher dispersion of thective phase [23,24]. For example, oxidative treatments with HNO3odified the carbon surface chemistry and promoted the formation
f small, well-dispersed crystals of the molybdenum precursor onhe support [21,25] although this led to lower phenol yields duringuaiacol HDO [23]. Additional studies further confirmed that HDOhemistry can be controlled by modifying the surface chemistry ofhe carbon support and consequently the dispersion of the catalyst16,22]. This adds to the potential use of carbon-supported systemsn rational catalyst design [17,19].
A wide variety of active phases – metals [4,9,10,26], transitionetal phosphides [7] and transition metal nitrides [27] – have been
mployed for HDO reactions in order to obviate the need to add sul-ur to the feed. In particular, transition metal nitrides show greatotential as catalysts due to their ceramic-like physical propertiesoupled with chemical properties resembling platinum-group met-ls [28]. These materials are also responsible for unique catalyticathways, leading to desirable product selectivities [28,29]. Con-equently, they offer a cheaper and more selective alternative tooble-metal catalysts such as Ru, Pd and Pt. Our previous studyhowed high activities and a high phenol/catechol ratio for bulkolybdenum nitride catalysts in the HDO of guaiacol [30]. How-
ver, supported catalysts are preferred in commercial applicationsor mechanical and morphological stability. The addition of Co as aromoter is known to improve the activity of bulk and supportedo2N catalysts for HDS and HDN reactions [31,32]. Furthermore,
o-promoted MoS2 catalysts have been reported to exhibit sig-ificantly higher HDO activity compared to non-promoted MoS2atalysts for HDO of guaiacol [8]. Although addition of Co improvedhe yield of deoxygenated products, the overall activity was notnhanced compared to the monometallic nitride [30]. This moti-ated us to investigate also the effect of Co promoters on theatalytic properties of supported nitrides in HDO reactions.
Here we report on the behavior of molybdenum nitridesispersed on four different activated carbon supports. The
upports were both microporous/mesoporous and meso-orous/macroporous carbons. The catalysts were synthesizedy impregnation of an aqueous salt and its subsequent conver-ion to the nitride; thermal conversion was achieved by either
: General 439– 440 (2012) 111– 124
ammonolysis or reduction/nitridation using a N2/H2 mixture.The effects of the synthesis procedure, support properties and theaddition of Co as a promoter on the HDO of guaiacol were examinedin terms of catalytic activity and phenol/catechol selectivity.
2. Experimental
2.1. Preparation of catalysts
Four commercial activated carbons (Norit Americas, Inc.) wereused as supports: NORIT GAC 1240 Plus (0.42–2.00 mm particlesize, SBET = 976 m2 g−1, total pore volume = 0.56 cm3 g−1), NORITGCA 1240 Plus (0.42–1.70 mm, 1132 m2 g−1, 0.51 cm3 g−1), DarcoMRX (0.60–2.00 mm, 613 m2 g−1, 0.62 cm3 g−1), and NORIT CGran(0.50–1.70 mm, 1402 m2 g−1, 1.15 cm3 g−1). Prior to their use, theactivated carbon materials were treated with 1 M HNO3 at 90 ◦Cfor 6 h. The solution was then filtered and extensively washed withdistilled water to bring the pH to ca. 7. The samples were driedovernight under vacuum at 120 ◦C. The supported molybdenumoxide precursors were prepared by incipient wetness impregnationusing aqueous solutions of ammonium heptamolybdate (Fischer,AHM, (NH4)6Mo7O24·4H2O, A.C.S. grade). After the impregnation,the samples were kept at room temperature for 24 h, followedby drying overnight at 110 ◦C. The bimetallic oxide precursorswere prepared by sequential impregnation: Mo-loaded sampleswere first prepared using the same drying–calcination proceduredescribed above; these samples were then impregnated with aque-ous solution of cobalt (II) nitrate hexahydrate (Acros Organics,Co(NO3)2·6H2O, 99%) and kept overnight at room temperature fol-lowed by drying overnight at 110 ◦C. The supported oxides wereprepared to obtain a nominal loading of 10 wt% Mo for monometal-lic samples and 10 wt% Mo plus 2.4 wt% Co for the bimetallicsamples. All oxide precursors were sieved to obtain 180–450 �mparticle size range.
Molybdenum nitrides were prepared by loading a 10 mm i.d.quartz reactor tube with 2.5 g of the oxidic precursor, while pass-ing ammonia (Matheson Tri-Gas, NH3, 99.99%) or a N2/H2 mixture(N2: BOC Gases, Grade 5; H2: Matheson Tri-Gas, 99.99%) over thesample [33,34]. The reactor was initially purged with nitrogen for30 min and switched to NH3 (300 mL min−1) or a N2/H2 mixture(300 mL min−1, N2/H2 = 5/1 (v/v)). The temperature was linearlyincreased from ambient temperature to 300 ◦C within 30 min(9.33 ◦C min−1), then from 300 to 500 ◦C at 0.6 ◦C min−1, and from500 to 700 ◦C at 2 ◦C min−1. The temperature was maintained at700 ◦C for 2 h. The nitrides prepared using NH3 were cooled to roomtemperature using the same flow rate of NH3 while the nitridesprepared using the N2/H2 mixture were cooled in 300 mL min−1 ofN2. The materials were then passivated in 1% O2/N2 (BOC Gases,UHP grade) for 12 h to avoid oxidation upon exposure to air. Thepreparation condition was adopted after careful review of the lit-erature in addition to previous results to ascertain the optimalcondition for unsupported molybdenum nitrides [30]. Preparationof nitride catalysts using NH3 and N2/H2 is referred to as method1 and method 2, respectively. For notation, Mo nitrides preparedusing method 1 have suffix “A”, while suffix “NH” implies method 2:e.g., MoN/Darco-A and MoN/Darco-NH are Darco activated carbon-supported Mo nitride catalysts prepared using NH3 and N2/H2,respectively.
Molybdenum, cobalt and nitrogen contents in the catalysts wereperformed by Galbraith Laboratories using ICP-AES for the metalsand a combustion method for nitrogen.
2.2. Nitrogen porosimetry
Nitrogen adsorption/desorption isotherms were obtained at77 K using a Micromeritics ASAP-2020 instrument to evaluate the
ysis A:
ptarsawpvtA2tb(t
2
pe(sca
2
o(wtctwAIabr
(
wta
2s
imTt9mtteR
I.T. Ghampson et al. / Applied Catal
orous structure of the support and catalyst samples. Prior tohe measurements, the samples were outgassed under vacuumt 200 ◦C for 12 h. The isotherms were collected within a broadelative pressure range, 10−6 < P/P0 < 0.995, and using a low pres-ure incremental dosing (3 cm3 g−1 STP) in order to obtain andequate characterization of the micropore region. The isothermsere used to calculate the BET specific surface area (SBET), totalore volume (TPV), average pore diameter (dpore), and microporeolume (V�). SBET was obtained from the adsorption branch inhe range 0.04 ≤ P/P0 ≤ 0.14 and TPV was recorded at P/P0 = 0.995.verage pore diameters were calculated from the equation dpore =
· TPV/SBET , assuming slit-shaped pores. The pore size distribu-ions (PSD) for 0.4–100 nm were determined from the adsorptionranch of the isotherm using the nonlocal density functional theoryNLDFT) method [35,36]. Micropore volume was calculated fromhe NLDFT cumulative volume of pores whose size was below 2 nm.
.3. X-ray diffraction
Wide-angle X-ray diffraction (XRD) patterns of powdered sam-les were obtained using a PANalytical X’Pert Pro diffractometerquipped with a graphite monochromator and Cu K� radiation45 kV, 40 mA) in a parallel beam optical geometry. The standardcan parameters were 15–85◦ 2� with a step size of 0.02◦ and aounting time of 10 s per step. Identification of the phases waschieved by reference to the JCPDS data files.
.4. X-ray photoelectron spectroscopy (XPS)
X-ray photoelectron spectra of reduced catalysts were obtainedn a VG Escalab 200R electron spectrometer using a Mg K�1253.6 eV) photon source. The samples were pre-reduced ex situith H2 at 450 ◦C for 6 h. After reduction, the samples were cooled
o room temperature, flushed with nitrogen and stored in flasksontaining isooctane (Merck, 99.8%), then transferred to the pre-reatment chamber of the spectrometer. The binding energies (BE)ere referenced to the C 1s level of the carbon support at 284.9 eV.n estimated error of ±0.1 eV can be assumed for all measurements.
ntensities of the peaks were calculated from the respective peakreas after background subtraction and spectrum fitting by a com-ination of Gaussian/Lorentzian functions. Relative surface atomicatios were calculated from
Mo/C)atomic ratio = (SMo 3d/fMo 3d)(SC 1s/fC 1s)
; (Co/C)atomic ratio
= (SCo 2p/fCo 2p)(SC 1s/fC 1s)
; (N/C)atomic ratio = (SN 1s/fN 1s)(SC 1s/fC 1s)
here S is the corresponding peak areas and f is the respectiveabulated sensitivity factor [37] and the factor for the device. Thetomic ratios were calculated at a precision of 7%.
TPD analyses of the carbon supports were carried out in ann-house apparatus which consisted of a U-shaped quartz tube
icro-reactor, placed inside a programmable electrical furnace.he TPD profiles of CO, CO2 and H2O were obtained from roomemperature to 1040 ◦C, at 10 ◦C min−1 under helium (AGA Chile,9.995%) flow of 50 mL min−1. Evolution of desorbed gases wasonitored by a thermal conductivity detector (TCD). In addition,
o quantify the gases produced during thermal decomposition ofhe surface functionalities, TPD was coupled with mass spectrom-try (Altamira AMI-200 R-HP instrument equipped with a SRSGA-300 mass spectrometer). About 0.2 g of the activated carbon
General 439– 440 (2012) 111– 124 113
sample was first pretreated at 100 ◦C for 4 h in He (50 mL min−1) toremove most of the weakly adsorbed water, and cooled to roomtemperature in He. The pretreated sample was then heated ina flow of He (50 mL min−1) from room temperature to 800 ◦C at10 ◦C min−1.
FT-IR analyses of the supports were performed on a NicoletNexus FTIR in the wavenumber range (4000–400 cm−1) and with ascan of 64. The samples were prepared using a 1:100 mg of activatedcarbon and KBr support.
2.7. Acidity measurements
Acid site concentrations and acid strength measurements of thesupports and selected catalysts were determined using a potentio-metric method [38], whereby a suspension in acetonitrile (Merck,99.9%) was titrated with n-butylamine (Merck, 99%). The varia-tion in electric potential was registered on a Denver InstrumentUltraBasic pH/mV meter.
2.8. Reaction characterization
Reactivity studies were performed in a 300 mL stirred-batchautoclave set-up (Parr Model 4841) at 300 ◦C and under a hydrogenpressure of 5 MPa. Prior to catalyst testing, the passivated sampleswere activated ex situ under H2 (AGA Chile, 99.99%) at a flow rateof 60 mL min−1 and 450 ◦C for 6 h, conditions that were shown inprevious studies to maximize HDO conversion [27]. Approximately0.25 g of freshly pre-treated catalyst was added to the reactorcharged with 80 mL of decalin (Merck, 99%), 2.53 mL of guaiacol(0.232 mol L−1, Merck, 99.5%), and 700 �L of hexadecane (Merck,99%). Hexadecane was used as an internal standard for quanti-tative GC analysis. The sealed reactor was flushed with nitrogen(AGA Chile, Grade 5) for 30 min to evacuate air from the system.While continuously stirring the mixture, the reactor was heated to300 ◦C under N2. Once the reaction temperature was attained, N2was replaced with H2 and then pressurized to 5 MPa. This pressurewas maintained for the entire duration of the experiment by addingH2 to the reactor whenever necessary. Liquid samples were peri-odically withdrawn during the course of the reaction after purgingthe sampling line by withdrawing a small amount of the reactantmixture. The samples were analyzed by a Perkin Elmer (Clarus400) gas chromatograph equipped with a flame ionization detector(FID) and a CP-Sil 5 CB column (Agilent, 30 m × 0.53 mm × 1.0 �mfilm thickness). The injector and FID were held at 275 and 180 ◦C,respectively. (The GC oven program consisted of an initial isother-mal operation at 30 ◦C for 6 min, followed by heating to 70 ◦C at30 ◦C min−1 with an isotherm of 22 min, and a subsequent rampto 275 ◦C at 30 ◦C min−1.) The product distributions were identifiedby their column retention time in comparison with available stan-dards. The initial concentration of guaiacol was taken as 100% inorder to ignore slight conversion before isothermal condition wasachieved. The catalytic activity was expressed as the initial reactionrate, calculated from the slope of the conversion vs. time plot, aswell as by the intrinsic activity (in moles of guaiacol consumed permole of Mo per second). A number of repeated runs under the sameconditions were performed to ensure satisfactory reproducibility ofthe data. The uncertainty in the calculation of reaction rates fromthe GC peaks is 3%. The phenol/catechol ratios were determined at10% conversion of guaiacol.
The stability of selected nitride catalysts during the HDO reac-tion was compared to that of a commercial reference catalystin a stainless steel continuous-flow micro-reactor. In a typi-cal experiment, approximately 0.2 g of catalyst was diluted 1:1
114 I.T. Ghampson et al. / Applied Catalysis A: General 439– 440 (2012) 111– 124
NO3 tr
wwtswSC3tl3voccG
3
3
cmrHdaa
Fig. 1. Pore size distribution of H
ith SiC (Soviquim, Chile) and loaded into the reactor tube,hile the remaining reactor space was packed with SiC. Prior to
he reaction, the nitride catalyst was reduced in situ using theame pretreatment conditions employed for the batch reaction,hile the commercial Ni–Mo/Al2O3 catalyst (Procatalyse, HR 346,
BET = 256 m2 g−1) was sulfided in situ using a 10 vol.% H2S (AGAhile, 99.99%) in H2, at a flow rate of 67.5 mL min−1 and 350 ◦C for
h. The liquid reactant mixture and hydrogen were connected tohe reactor inlet where they flowed downward through the cata-yst bed. The conditions for HDO reactions were as follows: 300 ◦C,
MPa, 5.4 g/h of liquid feed corresponding to liquid hourly spaceelocity (LHSV) of 27 h−1, H2 gas hourly space velocity (GHSV)f 3600 h−1, H2/guaiacol molar ratio of 23. Fresh samples wereollected at an hourly interval for 8–9 h with regular flushing pre-eding each collection. The liquid products were then analyzed byC–FID.
. Results and discussion
.1. Textural properties
The pore size distributions (PSD) for the activated carbons, cal-ulated from the adsorption branch of the isotherm using the NLDFTethod, are shown in Fig. 1. All the support materials have a wide
ange of pore sizes, including micropores in the range 0.4–2.0 nm.
owever, the results for CGran and Darco supports reveal a pre-ominance of larger mesopores (up to 100 nm). In contrast, the GCAnd GAC supports are more microporous; the latter also possessesn appreciable amount of larger mesopores (3–100 nm).
eated activated carbon supports.
The BET surface areas, total and micropore volumes of supportsand nitride catalysts are presented in Table 1. Significant differ-ences in these textural parameters among the four supports areobvious. The CGran and GCA carbons both have higher surface areasbut the latter has a larger fraction of micropores (76 vs. 32%). Thelowest SBET displayed by the Darco support is consistent with thismaterial having the lowest microporosity (19%) of all the supports.The porosity of the GAC support is more evenly distributed: 57%mesoporosity and 43% microporosity.
Oxidation pretreatment of the supports with HNO3 producedvarying changes in the textural properties of the original samples.In particular, a significant loss of surface area (28%), total pore vol-ume (25%) and micropore volume (24%) was found for the CGranactivated carbon; this is attributed to an increase in the quantity ofoxygen-containing surface groups on the pore walls and entrances,making them inaccessible to N2 molecules at 77 K [39]. The pre-treatment resulted in significantly less micropore volume decreasein the GAC carbon (5%), corresponding to a 4% loss of surface area; incontrast, there was an increase in surface area and pore volume forthe Darco and GCA activated carbons, possibly due to the removalof impurities from the pores. The pore size distributions of the car-bons were not significantly modified after the treatment, however.Results summarized in Table 1 also show that impregnation of thesupports with Mo and Mo/Co, followed by thermal conversion tothe nitrides, led to a general decrease in surface area, as well as total
and micropore volume, especially in the case of the CGran support.These results suggest that nitride formation occurred inside theporous structure. For the Darco support in particular, they suggestpreferential nitride formation inside the mesopores; conversely, for
I.T. Ghampson et al. / Applied Catalysis A: General 439– 440 (2012) 111– 124 115
Table 1Nitrogen porosimetry results for supports and catalysts.
assigned to anhydrides, quinonic, carboxylic or ether groups,respectively [44]. There is no apparent difference between theGCA, GAC and Darco spectra, indicating, in agreement with TPD
MoN/GCA-A 995
MoN/GCA-NH 1066CoMoN/GCA-A 950
he CGran support they suggest preferential deposition in microp-res or at the entrances to such pores.
.2. Surface chemical properties of the support
The chemical nature of oxygen functionalities present on thectivated carbon surface after HNO3 treatment was determinedrom TPD/MS measurements, as shown in Fig. 2 and summarizedn Table 2: lactonic (190–650 ◦C) [40–42], carboxylic (200–300 ◦C)41,42], phenolic (600–700 ◦C) [40,42], carbonyl (800–980 ◦C)40,42], and quinone groups (700–1000 ◦C) [42,43]. Decomposi-ion of groups whose carbon atom is bonded to two oxygen atomscarboxylic acids, lactones, and carboxylic anhydrides) releasesO2, indicative of the presence of strong acidic sites [21,24]. Allhe supports exhibited pronounced peaks at low temperatures250–400 ◦C) and thus possessed large amount of acidic CO2-esorbing groups, as expected [24]. In the high temperature region,he GCA, GAC and Darco supports presented broader 515–1000 ◦Choulders, which is indicative of the prevalence of phenolic anduinone groups. The CGran support exhibited a 415–730 ◦C peakhich is suggestive of the presence of mostly phenolic groups;ecomposition of these groups, which leads to desorption of CO,
ndicates the presence of weakly acidic, neutral and basic groupshose carbon atom is bonded to one oxygen atom [21,24]. In addi-
ion, integration of the evolved gas TPD profiles shows that CGran
Fig. 2. TPD profiles of the activated carbon supports.
9 0.45 0.359 0.49 0.359 0.44 0.33
contained the greatest amount of CO2- and CO-desorbing groupsamong the activated carbon supports. Table 2 also shows that theGCA and Darco carbons had similar quantities of CO2- and CO-releasing functional groups. (The H2O profiles at low temperature,indicating weakly bound desorbed molecules, were not quantified.)Furthermore, comparison of TPD/MS results of pretreated and as-received activated carbon supports (not shown) indicates greateramounts of CO2- and CO-desorbing groups in the former. This isconsistent with the well documented effectiveness of oxidativeHNO3 treatment [21]. It should be clarified, however, that if advan-tage is not taken of the surface oxygen anchoring sites – by makingsure, during catalyst preparation, that there is attraction betweenthe positively charged support surface and the negatively chargedcatalyst precursor, or vice versa – then the only benefit of surfaceoxidation can be to render the support more hydrophilic (unlesstextural properties are significantly modified).
Fig. 3 shows the FT-IR spectra of the activated carbon sup-ports. The supports displayed bands in the 3500–3300 cm−1,1600–1500 cm−1 and 1500–1000 cm−1 regions which can be
results, that they have similar surface oxygenated species. For the
Fig. 3. FT-IR spectra of the activated carbon supports.
116 I.T. Ghampson et al. / Applied Catalysis A: General 439– 440 (2012) 111– 124
Table 2Surface chemical and acidic properties of oxidized supports.
Support TPD (arbitrary units) Acidity measurements
CO2 CO Acid strength (mV) Total acidity (mequiv. m−2)
Darco 12 8 127 2.3
Cbr3mCesi
tTsndtsEtcptAwd
3
sdldfspntwgaopTwsmrowdphang
spin orbits respectively, assigned to Mo(VI) species [51]. This indi-cates that after the reduction treatment only Mo oxynitrides waspresent on the catalyst surface. In the CoMo nitride catalysts the
CGran 26 29
GAC 16 16GCA 12 8
Gran support, a strong band appears at 1691 cm−1 which cane assigned to lactonic groups [44], in agreement with the cor-esponding TPD profile; another two bands were detected in the124–3037 cm−1 and 2904–2844 cm−1 regions, assigned to aro-atic and aliphatic groups [44]. By comparing the spectra in Fig. 3,
Gran support displayed the greatest intensity of carboxylic orther groups suggesting that it contains the greatest quantity ofurface acidic oxygen groups. This result is consistent with thenterpretation from TPD profiles.
The surface acidity of the supports was estimated from poten-iometric titration curves with n-butylamine as the probe molecule.he results include the maximum acid strength of the surfaceites (derived from the initial electrode potential, E0) and the totalumber of acid sites normalized by the surface area (acid siteensity). The acid strength can be determined according to the cri-erion proposed by Cid and Pecchi [38]: E0 > 100 mV, very strongites; 0 < E0 < 100 mV, strong sites; −100 < E0 < 0 mV, weak sites;0 < −100 mV, very weak sites. Accordingly, the results confirm thathe HNO3-treatment produced strong acid sites on the activatedarbon supports. Table 2 shows that the CGran, Darco and GAC sup-orts displayed very strong acid sites with E0 > 100 mV, whereashe GCA carbon displayed strong acid sites with 0 < E0 < 100 mV.s mentioned, the Darco carbon had the highest acid site density,hile the GAC and CGran carbons had similar densities. The lowestensity of acid sites was measured for the GCA carbon.
.3. Bulk and surface composition of nitrided catalysts
Bulk molybdenum, cobalt and nitrogen contents of passivatedupported catalysts are listed in Table 3. The nitrogen content isue to nitride formation and the creation of pyridine- and pyrrole-
ike functions during ammonolysis and N2/H2 treatment [43]. Toistinguish the nitrogen content due to the molybdenum nitriderom those associated with the carbon support, we nitrided blankupport in a manner similar to that used to prepare the sup-orted Mo and CoMo nitride catalysts, and performed chemicalitrogen analysis: the results indicate that between 52 and 73% ofhe total nitrogen content of the monometallic method 1 samplesere due to contributions from the support’s nitrogen-functional
roups; by comparison, the nitrogen species of the carbon supportccounted for between 27 and 45% of the total nitrogen contentf the monometallic method 2 catalyst; the corresponding sup-ort nitrogen groups for the bimetallic samples ranged 42–90%.his clearly shows that for the same support the nitrogen contentsere consistently higher for samples prepared via method 1 (A-
eries). Ammonolysis was thus more effective than with N2/H2. Theeasured metal content values, in most part, correspond to theo-
etical values (10 wt% for Mo and 2.4 wt% for Co). However, somef the catalysts had metal contents lower than the nominal valueshich could be attributed to the possible loss of volatile molyb-ates and cobaltates during the decomposition steps of catalystreparation such as drying and nitridation, a phenomenon which
as also been observed by other authors [22,45]. The bulk N/Motomic ratios (due to the nitrogen structure of the molybdenumitride and denoted (N/Mo)nitride in Table 3) for the catalysts wereenerally higher than the stoichiometric N/Mo ratio for �-Mo2N
290 1.5119 1.6
61 1.2
(i.e. 0.5) and �-Mo2N0.78 (i.e. 0.39) which suggests nitrogen enrich-ment, possibly residing in interstitial sites. Table 3 also shows thatthe N/Mo ratio was higher for the method 1 samples than method 2samples; this suggests a higher degree of nitridation of the former.
The surface composition of the reduced and passivated nitridecatalysts was determined by XPS and the results are summarized inTable 4. The C 1s binding energies consisted of four peaks between284.8 and 289.2 eV. The peak at 284.8 eV was assigned to C Cand/or C C bonds of aromatic (and aliphatic) carbon [42,46], whileat 286.3 eV it is indicative of C O bonds in phenolic or ether groups[47,48], or may be due to the presence of C N bonds [49]. The peakat 287.7 eV is consistent with quinone-type groups or C N species[49,50], and at 289.3 eV with carboxyl groups and esters [48]. Therelative peak intensities (shown in parentheses) indicate the pre-dominance of aromatic carbon on the surface of all the catalysts.
The XRD patterns of the GAC-, GCA- and Darco-supported metalnitrides (see Supplementary data) revealed only peaks due to theoriginal carbon supports. The absence of Mo nitride diffractionpeaks suggests that the catalysts contained small crystallites ofMo nitride below the detection limit, or that the broad 0 0 2 and1 0 bands from the support may have masked the low-intensityMo nitride peaks. The XRD results of CGran supported nitrides aresummarized in Fig. 4. The MoN/CGran-NH catalyst displayed thecharacteristic peaks for �-Mo2N0.78 (2� = 37.61◦, 62.53◦, 75.53◦)together with broad features (2� = 26◦ and 43◦) associated with thecarbon supports; a barely discernible peak at 2� = 37.41, consistentwith the �-Mo2N (1 1 1) phase, was observed in the MoN/CGran-A catalyst. Also shown in Fig. 4 is the diffraction pattern forCoMoN/CGran-A, which indicated the presence of Co3Mo3N crys-tallites (2� = 40.09◦, 42.59◦, 46.59◦).
XPS results of Mo 3d5/2, N 1s and Co 2p3/2 species are summa-rized in Table 4. The binding energies of Mo 3d and N 1s of someselected catalysts are shown in Fig. 5. The Mo 3d spectra, shown inFig. 5(a), presented two spectral lines centered at 232.6 ± 0.3 eVand 235.7 ± 0.3 eV corresponding to the Mo 3d5/2 and Mo 3d3/2
Fig. 4. XRD patterns of CGran carbon-supported nitrides.
I.T. Ghampson et al. / Applied Catalysis A: General 439– 440 (2012) 111– 124 117
Table 3Chemical analysis of metal nitride catalysts.
a Calculated by multiplying the distribution of N 1s in the region of 396.3–396.9 eb N/Mo = (N/C)nitride/(Mo/C).
o 2p3/2 binding energy was 781.4 eV; BE characteristic of Co(II)pecies is normally observed at 781.8 eV [52], showing a decreasen positive Co2+ charge, indicating partial replacement of oxygen
ith a less electronegative nitrogen forming oxynitrides [53]. Theesult confirms that despite reduction of the passivated catalyst at50 ◦C, its surface is mainly an oxynitride rather than a nitride. Thebsence of nitride species on the external surface is attributed to aigher concentration of oxygenated functionalities on the supporturface (as a result of the HNO3-pretreatment): negatively charged
eptamolybdate units interact more strongly with the positivelyharged carbon surfaces which may have inhibited the formationf fully nitrided groups during nitridation. In contrast to theseesults, XRD analyses showed detectable nitride amounts on the
he total N/C atomic ratios.
CGran support, suggesting that nitride forms inside the pores ofthe support rather than on the external surface of the supportedcatalyst particles. This could be related to the heterogeneity in thedistribution of oxygenated functionalities of the carbon supportduring HNO3-treatment. Due to diffusional effects the amount ofoxygen groups inside the pores is expected to be less than thoseat the exterior [25]; consequently, the inner pores of the sup-port are less positively charged and thus adsorbed less strongly tothe heptamolybdate anion, leading to relatively easier decomposi-
tion and transformation of the precursor to nitride. This proposalgoes further by demonstrating that the decrease in support textu-ral parameters after impregnation and thermal conversion showsthat the XRD-detected Mo2N, Mo2N0.78 and Co3Mo3N species were
ysis A: General 439– 440 (2012) 111– 124 119
pdo
spgiogptsfhsmnm
TrotrpltfpoimsfeCndttopldstbtsmmadr
acMpagisso
I.T. Ghampson et al. / Applied Catal
resent on internal surfaces of the particles. Thus, the results implyifference in composition between internal and external surfacesf the carbon support.
The XPS peaks in the N 1s region shown in Fig. 5(b) and pre-ented in Table 4, 398.3 ± 0.1, 399.6 and 400.1 eV, are ascribed toyridine, amide and pyrrolic groups created at the edges of theraphene layers by nitridation [47,54]. The component at ∼396.5 eVs attributed to N 1s in the Mo N bond [55], confirming the presencef molybdenum oxynitrides. The relative abundance of each nitro-en species of the catalysts indicates that monometallic catalystsrepared via method 2 contained more N species associated withhe Mo N bonds than catalysts prepared via method 1; the resultuggests that the effectiveness of surface molybdenum oxynitrideormation was related to the nitridation method. At the same timeigher total abundance of nitrogen-functional groups of the carbonupport was obtained for the method 1 samples, compared to theethod 2 samples. This result qualitatively agrees with the bulk
itrogen content analysis in Table 3, confirming that ammonia isore effective at nitriding the carbon support.The atomic Mo/C, N/C and Co/C ratios are also listed in Table 4.
he XPS atomic ratios, estimated from intensity ratios of particleelated peaks and support related peaks, have a strong dependencen catalyst dispersion, and can qualitatively be used to comparehe dispersion of supported catalysts [56]. The differences in Mo/Catios are indicative of dispersion differences when the catalysts arerepared under different conditions. Thus, for example, the cata-
ysts prepared via method 2 (NH series) displayed higher Mo/C ratiohan those prepared via method 1 (A-series). It can be concludedrom Table 4 that the thermal conversion of the oxide catalystrecursor using N2/H2 mixture led to more highly dispersed Moxynitride particles. We can make a hypothesis on the basis of thenformation from the bulk and surface concentration of nitrogen
oieties on the carbon support. Nitridation reduces the quantity ofurface oxygen groups and simultaneously increases the nitrogenunctional groups; it has been confirmed that ammonolysis is moreffective than N2/H2 mixture for the nitridation of carbon supports.onsequently, the greater quantity of negatively charged surfaceitrogen species in the method 1 samples would induce a greateregree of partial distribution of molybdenum oxynitrides towardshe interior of the catalyst. The N/C atomic ratio, which expresseshe dispersion of the total nitrogen species (including nitrogen fromxynitride and nitrogen from organic species) on the carbon sup-ort surface, did not follow any observable trend; these results are
ess informative in relation to the degree of dispersion of the molyb-enum oxynitride phase due to contributions from surface nitrogenpecies on the carbon support. However, the N/C atomic ratio per-aining to the molybdenum oxynitride species can be elucidatedy multiplying the total N/C atomic ratio by the percent distribu-ion, denoted as (N/C)nitride and shown in Table 4. Except for theamples supported on Darco carbon, (N/C)nitride atomic ratios foronometallic samples are higher for method 2 samples than forethod 1 samples. This result indicates the nitridation method is
n essential factor in the control of the relative dispersion of molyb-enum oxynitride species for carbon supported nitrides. The Co/Catio did not follow any observable trend.
The surface N/Mo atomic ratio was evaluated from the XPS Mo/Cnd (N/C)nitride atomic ratios and is summarized in Table 4. Thealculated values were close to the stoichiometric N/Mo value of �-o2N and �-Mo2N0.78. However, because XPS results indicated the
resence of only molybdenum oxynitride, this observation is prob-bly due to excess nitrogen residing in defect sites like irregularrain boundary surfaces. Moreover, the N/Mo atomic ratio was sim-
lar to each other suggesting that the method of nitridation and theurface chemistry of the support had no observable control on theurface Mo oxynitride species formed. This infers that the naturef the nitrogen deficient sites was the same for all the catalysts.
Fig. 6. Surface N/Mo atomic ratio (from XPS) vs. bulk N/Mo atomic ratio (fromchemical analysis) of the carbon-supported catalysts.
A plot of the surface N/Mo atomic ratio (calculated from XPS) vs.bulk N/Mo atomic ratio (calculated from chemical analysis) aftersubtracting nitrogen content on the carbon support is shown inFig. 6, and clearly indicates that for monometallic catalysts the bulkremained enriched with nitrogen relative to the surface compo-sition. This result suggests that molybdenum nitrides/oxynitrideswith higher nitrogen deficiencies were located in the interior of thecatalysts, in line with the interpretation derived from XRD and XPS.In contrast, the CoMo nitride catalysts, except for CoMoN/CGran-Acatalyst, had higher nitrogen concentration on the external surfacethan in the bulk; this result is not yet clear.
The textural and chemical surface properties of the supportcould also influence Mo dispersion as demonstrated abundantly inthe literature [24]. Indeed, the highest dispersion displayed by theMoN/CGran-NH catalyst is attributed to the fact that this supporthas the most abundant oxygen surface groups (most hydrophiliccharacter) and the highest mesoporosity; this facilitates access ofthe aqueous solution to the internal pore structure and allows ahomogenous radial distribution of the metal precursor within theparticles of the support [23]. Conversely, the lowest dispersion ofthe MoN/GCA-NH catalyst is attributed to the lowest concentra-tion of oxygen surface groups (most hydrophobic character) as wellas the predominance of micropores in the support. Furthermore,the intermediate dispersion of the MoN/GAC-NH catalyst corre-lates with this solid possessing intermediate surface chemical andtextural property. It is important to remark that the presence ofoxygen functional groups can render the support surface negativelycharged over a wide range of pH condition, causing electrostaticrepulsions between the support and the heptamolybdate anion[25]. This can result in metal particles aggregation, leading to lesssurface Mo atoms as evident by the low Mo/C atomic ratio of theMoN/CGran-A catalyst.
3.4. Catalytic activity
The conversion of guaiacol and the evolution of reaction prod-ucts are illustrated in Figs. 7 and 8. Periodic samplings of the
liquid mixture in the reactor were analyzed by GC, from which theconcentration of the reactant and the product yields were deter-mined relative to the hexadecane standard. The observed productswere similar for all the catalysts and include phenol, catechol,
120 I.T. Ghampson et al. / Applied Catalysis A: General 439– 440 (2012) 111– 124
F h timeM
dbpaamtdaacqaTt
ig. 7. Variation of the transformation of guaiacol and the yield of products witoN/GAC-NH, and (f) CoMoN/GAC-A catalysts.
eoxygenated compounds (such as cyclohexene, cyclohexane andenzene), and heavy compounds (such as mono- to tetra-methylhenols and dimethyl catechols). Other possible products, such asnisole, m- and o-cresol were not observed. Furthermore, CGran-nd GAC-supported catalysts produced very small quantities ofethylcatechol which were attributed to the higher total acidi-
ies of these catalysts. Fig. 7 illustrates similarities in the productistribution for the GCA-supported Mo nitride catalysts (Fig. 7(a)nd (b)) and the GAC-supported Mo nitride catalysts (Fig. 7(d)nd (e)): guaiacol was mainly converted to phenol, while cate-hol and the deoxygenated compounds were observed in smaller
uantities. Heavy compounds were formed in significant amountnd are attributed to the methylation of the aromatic ring [57].he production of heavy compounds was even more prominent inhe CoMo nitride catalysts supported on GCA and GCA, as seen in
with (a) MoN/GCA-A, (b) MoN/GCA-NH, (c) CoMoN/GCA-A, (d) MoN/GAC-A (e)
Fig. 7(c) and (d), respectively. By comparison, the CGran-supportedcatalysts exhibited some slight differences in the products distri-bution at higher conversion, as shown in Fig. 8(a)–(c): phenol wasfound to be the dominant product over the MoN/CGran-NH cata-lyst while heavy products and phenol were the main competingproducts formed in almost equal amounts with the MoN/CGran-Aand CoMoN/CGran-A catalysts. As seen in Fig. 8(d)–(f), the Darco-supported catalysts showed similar behavior in the evolution ofproducts. The product evolution observed indicates that guaiacolHDO followed the reaction scheme proposed by Bui et al. [57]shown in Fig. 9. The two general pathways are: (i) initial demethy-
lation (DME) to form catechol, followed by dehydroxylation to formphenol; or (ii) direct demethoxylation (DMO) to form phenol. (Lightproducts such as methane and methanol could not be separatedby the column used under batch conditions, but they are expected
I.T. Ghampson et al. / Applied Catalysis A: General 439– 440 (2012) 111– 124 121
F e with (a) MoN/CGran-A, (b) MoN/CGran-NH, (c) CoMoN/CGran-A, (d) MoN/Darco-A, (e)M
bompcdc[
ratasp
Table 5Catalytic activities of carbon-supported Mo nitride catalysts.
ig. 8. Variation of the transformation of guaiacol and the yield of products with timoN/Darco-NH, and (f) CoMoN/Darco-A catalysts.
yproducts of DME and DMO, respectively.) Further deoxygenationf phenol produces benzene, cyclohexene and cyclohexane, whileethyl-substitution of catechol forms methylcatechol. Continuous
roduction of catechol at longer reaction times indicated that theonversion of catechol to phenol was not prominent although bothemethylation and direct demethoxylation occurred over theseatalysts. The same tendency was observed for bulk metal nitrides30].
Table 5 summarizes the catalytic activities. The initial reactionates calculated from the slopes of the guaiacol conversion curvesre given in Fig. 10. Negligible conversions were obtained using
he bare supports (not shown), indicating that the activities weressociated with the Mo oxynitride and not the support. Thus, theurface oxygen functional groups of the activated carbon did notarticipate in the conversion of guaiacol; however, the interaction
122 I.T. Ghampson et al. / Applied Catalysis A: General 439– 440 (2012) 111– 124
nationA
btitcanrala(rohs
Fig. 9. Hydrodeoxygedapted from Bui et al. [57].
etween the metal precursor compound and the surface groups ofhe support promotes good dispersion, leading to enhanced activ-ties. This is supported by the observed higher activities comparedo our previous results for non-modified carbon supported Mo2Natalysts [27]. The reaction rate was not related to the surface N/Motomic ratio, as a consequence of the catalysts possessing the sameature of surface nitrogen deficient sites discussed previously inegard to the reported XPS N/Mo ratio data. However, the reportedctivities were affected by the method of nitridation: the cata-ysts prepared by method 2 (NH series) had consistently higherctivities. This is related to their higher Mo oxynitride dispersionillustrated in Fig. 11), confirmed by the Mo/C and (N/C)nitride atomic
atio in Table 4. The method 2 catalysts contained a greater numberf active sites associated with molybdenum oxynitride, leading toigher activities. Correlation of HDO activity with dispersion hasimilarly been reported for sulfided Co–Mo/carbon and reduced
Fig. 10. Reaction rates of carbon-supported Mo nitride catalysts.
pathway of guaiacol.
Ni–W/carbon catalysts [22,23]. The higher HDO activity for method2 catalysts coupled with the reported advantages in the large-scalesynthesis of Mo nitride using a N2/H2 mixture as a reactant over theNH3 synthesis makes method 2 attractive for potential industrialapplication [34]; thus, for example, the N2 and H2 reactants can beeconomically recycled by drying, and their use simplifies handlingprocedures as well as eliminate heat transfer problems associatedwith endothermic decomposition of NH3 [34].
Other trends can be observed in Fig. 10 when the catalystsprepared using the same method but dispersed on different sup-ports are considered. The specific activity of the method 2 catalysts(NH series) decreased in the order: MoN/CGran-NH > MoN/GAC-NH > MoN/GCA-NH > MoN/Darco-NH. Thus HDO activity appearsto be favored by a combination of higher Mo dispersion and
higher support mesoporosity (MoN/Darco-NH being the excep-tion): the highest Mo oxynitride dispersion and the easiestreactant accessibility to the support’s mesoporous structure led
Fig. 11. Reaction rates vs. XPS Mo/C atomic ratio.
ysis A: General 439– 440 (2012) 111– 124 123
tTlMtthotpcshl3f[ot
iFacoatTUbSNfMcasotbcpnvtcwsao
taioMltBbosaasM
and manner of the effect of the surface chemistry of the support,dictated by the surface oxygen groups, is unclear. Therefore, thiswarrants additional research.
I.T. Ghampson et al. / Applied Catal
o the highest HDO activity of the MoN/CGran-NH catalyst.he reaction rates of the method 1 catalysts (A-series) corre-ate with Mo oxynitride dispersion and decrease in the order:
oN/GCA-A > MoN/GAC-A > MoN/Darco > MoN/CGran-A. Some ofhe Darco-supported catalysts exhibited inferior activity, comparedo GCA-supported catalysts, despite their higher dispersion andigher mesoporosity. This surprising behavior could be due to anverestimation of the Mo signal obtained by XPS. Considering thathe Darco support possessed the lowest surface area of all the sup-orts, and that all the catalysts were impregnated with a similar Moontent, this catalyst should contain the largest Mo nitride particleizes. However, their measured Mo/C atomic surface ratios wereigher probably due to the inability of X-ray photons to penetrate
arger Mo nitride particles, leading to a higher intensity of the Mod XPS signal and thus an overestimation of the Mo/C atomic sur-ace ratios. Similar behavior was previously observed by Lagos et al.58]. Therefore, the low activity of Mo nitride catalysts supportedn Darco carbon was probably due to the loss of active sites throughhe formation of agglomerates.
The intrinsic activities based on the Mo content are also givenn Table 5. The trends were similar to the total reaction rates.igs. 10 and 11 also show that the addition of Co did not increase thectivity of the catalysts. In fact, GCA- and GAC-supported CoMoNatalysts were not nearly as active as their MoN counterparts. Inur previous, it was shown that addition of Co did not enhance thectivity of unsupported Mo nitride catalyst [30]. This was attributedo incomplete formation of the bimetallic nitride, Co3Mo3N, phase.he same interpretation applies to supported catalysts as well.nlike bimetallic sulfides and oxides, the formation of single-phaseimetallic nitride is not trivial and often exists in multiple phases.ynthesis of pure-phase bimetallic nitride catalysts (Co3Mo3N andi2Mo3N) and its HDO catalysis are warranted. Moreover, the sur-
ace Mo/Co atomic ratio (from XPS analysis) is higher than the bulko/Co atomic ratio (from metal content analysis), indicating that
obalt is preferentially deposited in the interior of the support andre possibly trapped underneath molybdenum. This can be taken touggest that during impregnation cobalt migrated into the interiorf the catalyst and caused the partial migration of molybdenumowards the exterior, in good agreement with the results obtainedy Ferrari et al. [17]. This inhomogeneity in the distribution of cobaltould explain the diminishing influence on activity after Co incor-oration: a higher concentration of surface Co species could createew (or modify) active sites which could enhance activity; con-ersely, a large proportion of Co in the pores will not only hinderheir accessibility but may also limit reactant diffusion into otheratalytic active sites located in the internal surfaces of the supporthich will negatively affect reaction rates. However, a more exten-
ive analysis of promoter effect (utilizing different Co precursorsnd preparative methods such as sequential vs. co-impregnation)n activity requires further investigation.
The selectivity results are summarized in Fig. 12. In con-rast to metal hydrogenation catalysts like Ru [4], metal sulfidend metal nitride hydrotreating catalysts have a higher selectiv-ty for HDO reactions relative to hydrogenation of aromatic andlefinic compounds [8,30]. During conversion of guaiacol overo nitride/carbon catalysts, both demethylation and demethoxy-
ation reactions take place only at the active sites situated onhe metal nitrides due to the inertness of the carbon support.lank experiments with all the supports resulted in negligi-le conversion, suggesting that the physicochemical differencesf the carbon supports had no direct influence on the productelectivity. The dual-pathway behavior of the Mo2N/carbon cat-
lytic system provides evidence that two kinds of active sitesre present on these catalysts. Theoretical studies reveal that theurface of fresh bulk Mo2N consists of coordinately unsaturatedo and N atoms, and 4-fold type vacancies [59]. The surface
Fig. 12. Phenol/catechol ratio calculated at 10% guaiacol conversion.
nature of reduced passivated supported catalysts is inevitablydifferent due to the elimination of the weakly adsorbed NHx speciesby passivation; thus, we infer that the surface of the carbon-supported Mo nitride catalysts expose coordinately unsaturatedMo atoms and 4-fold type nitrogen deficient sites. Hypothesis canbe made that coordinately unsaturated Mo atoms are responsiblefor demethoxylation (Caromatic OCH3 bond cleavage) and dehy-droxylation while demethylation (Cmethyl O bond cleavage) occursat nitrogen deficient sites. This hypothesis is analogous to thedistinction between sites for sulfide catalysts reported by Fer-rari et al. [60]. All the catalysts displayed high phenol productionand this cannot be related directly to the surface acidic proper-ties of the carbon supports; recently, Sepulveda et al. [61] showedthat strong acid sites favor catechol formation. The preferencefor the demethoxylation pathway (Caromatic OCH3 bond cleavage)suggests that unsaturated Mo sites were predominant in the cata-lyst. However, these high phenol/catechol ratios were significantlylower than those observed for unsupported nitride catalysts [30],suggesting that the active sites on nitrides and/or oxynitrides weremodified by the support, rendering them less selective for thedemethoxylation route compared to the unsupported catalysts.This confirms that the generation of oxygen groups on the carbonsurface has an indirect contribution on phenol/catechol selectivitythrough the precursor/support interaction, modifying the nature ofthe active sites and their selectivity. For example, N atoms presenton the surface of bulk Mo nitride catalysts may not be there on sup-ported catalysts due to the metal nitride–support interaction andthis can have an overall effect on the phenol/catechol ratio and theamount of deoxygenated products. Fig. 12 does not reveal any cleartendencies in the phenol/catechol ratio, suggesting that the extent
Fig. 13. Time-on-stream behavior of the MoN/GAC-NH catalysts in terms of totalconversion for HDO of guaiacol at 300 ◦C, 3 MPa H2 pressure, H2/guaiacol ratio of 23.
1 ysis A
ivsicwFfllotp
4
cadnmfstawospopal
A
GI1AcoN
A
fj
R
[[[[
[[
[[[[
[
[
[
[[
[[
[
[[[
[[
[[[[[
[[[
[[[
[[[
[
[
[
[
[[
[[[[[
[
[[
[61] C. Sepúlveda, N. Escalona, R. Garcia, D. Laurenti, M. Vrinat, First International
24 I.T. Ghampson et al. / Applied Catal
Developing a robust HDO catalyst for pyrolysis oil upgradings a considerable challenge. Possible reasons for catalyst deacti-ation during HDO include coking, poisoning and loss of activeites through surface chemistry changes [11,12]. A preliminarynvestigation of time-on-stream behavior of our nitrided catalysts,ompared to a reference commercial sulfided NiMo/Al2O3 catalyst,as conducted in a continuous flow reactor. It is summarized in
ig. 13 and agrees with the results of Monnier et al. [62]. The liquidow rate was chosen to obtain low conversion. The nitrided cata-
ysts displayed higher stability than the sulfided catalyst after 4 hn stream under continuous operation. The gradual deactivation ofhe sulfided NiMo/Al2O3 catalyst could be due to loss of the sulfidehase during HDO reaction.
. Conclusions
Use of activated carbon materials with different textural andhemical surface properties to prepare supported Mo nitride cat-lysts resulted in their different activities in HDO of guaiacol,emonstrating rapid production of significant amounts of phe-ol. This indicated that the transformation of guaiacol proceededostly through the direct demethoxylation route, bypassing the
ormation of catechol. The higher specific activity of carbon-upported catalysts prepared by using a N2/H2 mixture, comparedo similarly supported Mo nitrides prepared by ammonolysis, wasttributed to a higher dispersion of Mo oxynitride. The dispersionas related to the surface chemistry and the textural properties
f the support: high concentration of oxygen surface groups on theupport and high mesoporosity of the support promotes better dis-ersion. The most active HDO catalyst (MoN/CGran-NH) was thene that contained highly exposed Mo species on a highly meso-orous support. Surprisingly, a generally diminishing influence onctivity was observed after incorporation of Co to prepare bimetal-ic nitrided catalysts.
cknowledgements
The authors acknowledge the financial support of DOE Epscorrant #DE-FG02-07ER46373 and the financial support from CON-
CYT Chile, projects PFB-27, PIA-ACT-130 and FONDECYT No.100512 grants. I. Tyrone Ghampson is indebted to NSF Careerward 0547103 for sponsoring a trip to the University of Concep-ión. The authors also gratefully acknowledge valuable discussionsn N2 sorption with Rachel Pollock, and the technical assistance ofick Hill and Manuel Veliz.
ppendix A. Supplementary data
Supplementary data associated with this article can beound, in the online version, at http://dx.doi.org/10.1016/.apcata.2012.06.047.
eferences
[1] A.V. Bridgwater, D. Meier, D. Radlein, Org. Geochem. 30 (1999) 1479–1493.[2] D.C. Elliott, Energy Fuels 21 (2007) 1792–1815.[3] E. Furimsky, Appl. Catal. A 199 (2000) 147–190.[4] D.C. Elliott, T.R. Hart, Energy Fuels 23 (2008) 631–637.[5] E. Laurent, B. Delmon, Appl. Catal. A 109 (1994) 77–96.[6] J. Zakzeski, P.C.A. Bruijnincx, A.L. Jongerius, B.M. Weckhuysen, Chem. Rev. 110
(2010) 3552–3599.[7] H.Y. Zhao, D. Li, P. Bui, S.T. Oyama, Appl. Catal. A 391 (2011) 305–310.[8] V.N. Bui, D. Laurenti, P. Afanasiev, C. Geantet, Appl. Catal. B 101 (2011) 239–245.[9] A. Gutierrez, R.K. Kaila, M.L. Honkela, R. Slioor, A.O.I. Krause, Catal. Today 147
(2009) 239–246.
[
: General 439– 440 (2012) 111– 124
10] Y.-C. Lin, C.-L. Li, H.-P. Wan, H.-T. Lee, C.-F. Liu, Energy Fuels 25 (2011) 890–896.11] E. Laurent, B. Delmon, Appl. Catal. A 109 (1994) 97–115.12] E. Laurent, B. Delmon, J. Catal. 146 (1994) 281–291.13] A. Popov, E. Kondratieva, J.M. Goupil, L. Mariey, P. Bazin, J.-P. Gilson, A. Travert,
F. MaugeÌ, J. Phys. Chem. C 114 (2010) 15661–15670.14] A. Centeno, E. Laurent, B. Delmon, J. Catal. 154 (1995) 288–298.15] P.E. Ruiz, K. Leiva, R. Garcia, P. Reyes, J.L.G. Fierro, N. Escalona, Appl. Catal. A
384 (2010) 78–83.16] M. Ferrari, R. Maggi, B. Delmon, P. Grange, J. Catal. 198 (2001) 47–55.17] M. Ferrari, B. Delmon, P. Grange, Carbon 40 (2002) 497–511.18] V.N. Bui, D. Laurenti, P. Delichère, C. Geantet, Appl. Catal. B 101 (2011) 246–255.19] A. Centeno, V. CH, R. Maggi, B. Delmon, in: A.V. Bridgwater (Ed.), Developments
in Thermochemical Biomass Conversion, Blackie Academic & Professional, Lon-don, 1997, pp. 602–610.
20] A. Centeno, O. David, C. Vanbellinghen, R. Maggi, B. Delmon, in: A.V. Bridgwater(Ed.), Developments in Thermochemical Biomass Conversion, Blackie Academic& Professional, London, 1997, pp. 589–601.
21] G. de la Puente, A. Centeno, A. Gil, P. Grange, J. Colloid Interface Sci. 202 (1998)155–166.
22] S. Echeandia, P.L. Arias, V.L. Barrio, B. Pawelec, J.L.G. Fierro, Appl. Catal. B 101(2010) 1–12.
23] G. de la Puente, A. Gil, J.J. Pis, P. Grange, Langmuir 15 (1999) 5800–5806.24] L.R. Radovic, F. Rodriguez-Reinoso, in: P.A. Thrower (Ed.), Chem. Phys. Carbon,
Marcel Dekker, 1997, pp. 243–358.25] G. de la Puente, J.A. Menendez, Solid State Ionics 112 (1998) 103–111.26] T. Nimmanwudipong, R. Runnebaum, D. Block, B. Gates, Catal. Lett. 141 (2011)
779–783.27] C. Sepúlveda, K. Leiva, R. García, L.R. Radovic, I.T. Ghampson, W.J. DeSisto, J.L.G.
Fierro, N. Escalona, Catal. Today 172 (2011) 232–239.28] S.T. Oyama, Catal. Today 15 (1992) 179–200.29] J.G. Chen, Chem. Rev. 96 (1996) 1477–1498.30] I.T. Ghampson, C. Sepulveda, R. Garcia, B.G. Frederick, M.C. Wheeler, N. Escalona,
W.J. DeSisto, Appl. Catal. A 413–414 (2012) 78–84.31] Y. Li, Y. Zhang, R. Raval, C. Li, R. Zhai, Q. Xin, Catal. Lett. 48 (1997) 239–245.32] J.W. Logan, J.L. Heiser, K.R. McCrea, B.D. Gates, M.E. Bussell, Catal. Lett. 56 (1998)
165–171.33] L. Volpe, M. Boudart, J. Solid State Chem. 59 (1985) 332–347.34] R.S. Wise, E.J. Markel, J. Catal. 145 (1994) 344–355.35] J.P. Olivier, Carbon 36 (1998) 1469–1472.36] M. Kruk, M. Jaroniec, K.P. Gadkaree, J. Colloid Interface Sci. 192 (1997) 250–256.37] C.D. Wagner, L.E. Davis, M.V. Zeller, J.A. Taylor, R.H. Raymond, L.H. Gale, Surf.
(1995) 587–595.49] E. Riedo, F. Comin, J. Chevrier, F. Schmithusen, S. Decossas, M. Sancrotti, Surf.
Coat. Technol. 125 (2000) 124–128.50] E. D’Anna, M.L. De Giorgi, A. Luches, M. Martino, A. Perrone, A. Zocco, Thin Solid
Films 347 (1999) 72–77.51] S.W. Yang, C. Li, J. Xu, Q. Xin, J. Phys. Chem. B 102 (1998) 6986–6993.52] J.S. Girardon, E. Quinet, A. Griboval-Constant, P.A. Chernavskii, L. Gengembre,
A.Y. Khodakov, J. Catal. 248 (2007) 143–157.53] M.L. Kaliya, S.B. Kogan, Catal. Today 106 (2005) 95–98.54] R.J.J. Jansen, H. van Bekkum, Carbon 33 (1995) 1021–1027.55] K. Hada, M. Nagai, S. Omi, J. Phys. Chem. B 105 (2001) 4084–4093.56] A.M. Venezia, Catal. Today 77 (2003) 359–370.57] V.N. Bui, G. Toussaint, D. Laurenti, C. Mirodatos, C. Geantet, Catal. Today 143
(2009) 172–178.58] G. Lagos, R. García, A.L. Agudo, M. Yates, J.L.G. Fierro, F.J. Gil-Llambías, N.
Escalona, Appl. Catal. A 358 (2009) 26–31.59] G. Frapper, M. Pelissier, J. Hafner, J. Phys. Chem. B 104 (2000) 11972–11976.60] M. Ferrari, S. Bosmans, R. Maggi, B. Delmon, P. Grange, Catal. Today 65 (2001)
257–264.
Congress on Catalysis for Biorefineries (CatBior), Torremolinos-Málaga, Spain,2011.
62] J. Monnier, H. Sulimma, A. Dalai, G. Caravaggio, Appl. Catal. A 382 (2010)176–180.
