Experimental and Theoretical Insights into the Hydrogen-Efficient Direct Hydrodeoxygenation Mechanism of Phenol over Ru/TiO2

Research Article

pubs.acs.org/acscatalysis

Experimental and Theoretical Insights into the Hydrogen-Eﬃcient

Direct Hydrodeoxygenation Mechanism of Phenol over Ru/TiO₂

Ryan C. Nelson,^†Byeongjin Baek,^‡Pamela Ruiz,^§Ben Goundie,^†Ashley Brooks,^†M. Clayton Wheeler,^∥

Brian G. Frederick,^§Lars C. Grabow,*^,‡and Rachel Narehood Austin*^,†,⊥

^†Department of Chemistry, Bates College, Lewiston, Maine 04240, United States

^‡Department of Chemical and Biomolecular Engineering, University of Houston, Houston, Texas 77204-4004, United States

∥

^§Department of Chemistry and Laboratory for Surface Science and Technology (LASST), Department of Chemical and Biological

Engineering, University of Maine, Orono, Maine 04469, United States

S

* Supporting Information

ABSTRACT: Catalytic reduction of pyrolyzed biomass is

required to remove oxygen and produce transportation fuels,

but limited knowledge of how hydrodeoxygenation (HDO)

catalysts work stymies the rational design of more eﬃcient and

stable catalysts, which in turn limits deployment of biofuels.

This work reports results from a novel study utilizing both

isotopically labeled phenol (which models the most recalcitrant

components of biofuels) with D₂O and DFT calculations to

provide insight into the mechanism of the highly eﬃcient

HDO catalyst, Ru/TiO₂. The data point to the importance of interface sites between Ru nanoparticles and the TiO₂support and

suggest that water acts as a cocatalyst favoring a direct deoxygenation pathway in which the phenolic OH is replaced directly with

H to form benzene. Rather than its reducibility, we propose that the amphoteric nature of TiO₂facilitates H₂heterolysis to

generate an active site water molecule that promotes the catalytic C−O bond scission of phenol. This work has clear implications

for eﬀorts to scale-up the hydrogen-eﬃcient conversion of wood waste into transportation fuels and biochemicals.

KEYWORDS: hydrodeoxygenation, ruthenium, TiO₂, biofuels upgrading, DFT

and makes detailed HDO studies of this feedstock quite

diﬃcult.^1aThus, phenol or substituted phenols, representative

of a large fraction of oxygenated bio-oil compounds, are often

used as model substrates.³Figure 1 schematically shows the two

main reaction pathways by which oxygen can be removed from

phenol. The upper pathway shows the direct deoxygenation

(DDO) of phenol to benzene, which utilizes a single hydrogen

equivalent for reduction. The lower pathway is initiated by a

catalytic hydrogenation (HYD) of phenol to cyclohexanone.

Additional reduction and dehydration steps are necessary to

reach a deoxygenated product, cyclohexene, which under HDO

reaction conditions is rapidly reduced by an additional

hydrogen equivalent to cyclohexane.⁴Overall, this process

consumes four hydrogen equivalents for deoxygenation. For

upgrading biofuels, it is desirable to selectively remove oxygen

because hydrogenation of the double bonds uses expensive

reduction equivalents while decreasing the octane value of the

naphtha fraction of the product.⁵Selectivity in the context of

the present work refers speciﬁcally to the extent to which

reducing equivalents are used to directly remove oxygen, as

opposed to the reduction of double bonds.

INTRODUCTION

■

Concerns about global climate change and energy security

motivate eﬀorts to develop biofuels and biochemicals from

nonedible lignocellulosic biomass. The U.S. DOE and USDA

have estimated that approximately 1.3 billion dry tons, the

equivalent of more than 1/3 of the demand for transportation

fuel in 2005, of this renewable resource can be sustainably

grown and is readily available.¹Furthermore, pyrolysis

technologies that convert biomass into bio-oils have the

potential to diversify the global energy mix and to develop

high-tech jobs in rural locations.²However, current pyrolysis

technologies produce oils that are too oxygenated to serve as

suitable transportation fuels or as platform chemicals that could

substitute for certain petrochemicals.¹Even recent break-

through technologies such as formate-assisted pyrolysis (FAsP)

still produce oils that contain phenolic compounds incompat-

ible with current fuel standards.²Thus, a complete biomass-to-

fuel process requires a catalytic upgrading step in which a

catalyst, a reducing agent, and the crude bio-oil are combined to

catalytically remove oxygen, ideally without removal of carbon.

When the reductant is hydrogen gas and the byproduct is

water, this upgrading process is called hydrodeoxygenation

(HDO).

Received: July 21, 2015

The complex, amorphous chemical structure of lignocellulo-

Revised: September 15, 2015

sic biomass leads to hundreds of oxygenated bio-oil compounds

6509

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

Figure 1. Two schematic HDO pathways for phenol. The top pathway represents the direct deoxygenation (DDO) route and utilizes hydrogen-

reducing equivalents most eﬃciently. The lower pathway (HYD) utilizes a total of four hydrogen-reducing equivalents to arrive at the ﬁnal

deoxygenated, but saturated product.

HDO catalyst design has been inspired by closely related

hydrodesulfurization catalysts (HDS) used for decades in the

petroleum industry to remove sulfur contaminants in crude oil

and to meet low sulfur fuel standards.^{3a,d−h,k,4,6}MoS₂catalysts

are archetypal HDS catalysts whose catalytic activity relies on

the formation of sulﬁde vacancies at the Mo-promoted edge

sites.⁷Although many HDS catalysts are also eﬀective at

catalyzing HDO reactions, they require a continual feed of H₂S

to prevent catalyst degradation. Lignocellulosic biomass

inherently has little sulfur, so adding sulfur during an upgrading

step is not desirable.^3dRecent eﬀorts have turned instead to

metal oxide- and supported metal particle-based catalysts,

which do not require a sulﬁde regeneration step.^{3a,d,h,k,6d,8}

Rational catalyst design relies on understanding the

mechanism by which catalysts operate. In our initial work in

this area, we screened a number of supported Ru catalysts for

HDO of liqueﬁed phenol, which contains 10 wt % water.^3a

Uncalcined Ru/TiO₂catalysts with particle sizes of ∼2 nm

showed the best activity and DDO selectivity of all the catalysts

screened, consistent with several other literature reports.^3m,9

Our ﬁndings supported the hypothesis that metallic Ru(0) was

the primary ruthenium species in the active catalyst.^3aWe and

others hypothesized that TiO₂’s superior support properties

were attributable to its redox activity,^3a,m,9cwhich can be

enhanced by hydrogen spillover. This interpretation was also

Figure 2. Three possible reaction mechanisms for DDO emphasizing

the isotopic signature predicted for each pathway when starting with

D₂as the reductant. Mechanism A is the direct C−O cleavage

suggested in literature reports on reducible ZrO₂and Fe₂O₃

supports.^3a,h,q,10

Conclusive information on the mechanism of catalytic direct

deoxygenation (DDO) of phenolic compounds is scarce.

Although the direct C−O bond scission pathway (Figure 2,

mechanism A) is fully consistent with observed product

distributions, it has been widely discarded because it requires

breaking a strong C−O bond.^3q,11Indeed, density functional

theory (DFT) calculations on the ﬂat Ru(0001) surface support

this assessment.¹²However, a recent DFT study on stepped Ru

reported that C−O activation may occur with a moderate

activation barrier of E_a= 0.78 eV.¹³Similarly, McEwen and co-

workers identiﬁed direct C−O bond scission as the dominant

pathway on Fe and Pd/Fe surfaces.^10,14

As an alternative to direct C−O bond scission, an initial

hydrogenation (HYD) step was proposed to weaken the C−O

bond, followed by an acid-catalyzed dehydration reaction

(Figure 2, mechanism B).^6a,15Resasco’s group, however,

convincingly argues that the product distributions observed

on selective DDO catalysts, such as Ru/TiO₂, Pd/ZrO₂, Fe, and

bimetallic Ni−Fe catalysts are inconsistent with the hydro-

genation/dehydration sequence.^3j,m,qThey suggest instead that

the DDO pathway is initiated by a tautomerization step leading

to a keto intermediate, which undergoes hydrogenation and

dehydration (Figure 2, mechanism C).^3j,q

mechanism with hydrogen attack at the ipso position. Mechanism B

requires initial hydrogenation at the ortho position, followed by

dehydration. Mechanism C is initiated by tautomerization, followed by

hydrogenation and dehydration. In the presence of D₂O, the hydroxyl

can be deuterated through H/D exchange. Reductive H/D equivalents

and H/D atoms exchanged with water are highlighted in red.

principles calculations to resolve the debate regarding the

mechanism of selective DDO catalysts. The addition of polar

and nonpolar additives provides direct evidence for the crucial

role of water in the activity and selectivity of these catalysts, and

a series of isotopic labeling experiments imply that the phenolic

hydroxyl is directly replaced by a single hydrogen atom.

Theoretical calculations indicate that H₂undergoes heterolytic

cleavage at an interfacial site between ruthenium nanoparticles

and a basic bridging hydroxyl at the TiO₂surface. The

heterolytic cleavage of hydrogen generates an active site

consisting of a Brønsted acid on the support in close proximity

to a reductive ruthenium hydride. The lowest energy reduction

pathway is a proton-mediated, direct substitution of the

aromatic hydroxyl with this ruthenium hydride. The exceptional

activity and selectivity of TiO₂is attributed to its amphoteric

character rather than its reducibility. TiO₂can both accept a

proton from H₂to generate the active site and donate a proton

to assist in cleaving the C−O bond of phenol; both steps have

We have expanded our examination of the Ru/TiO₂catalyst

system using a combination of experimental work and ﬁrst-

6510

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

almost identical activation energies. This novel mechanism,

which does not require the formation of an oxygen vacancy site

on TiO₂, is supported by theory and experiments using

isotopically labeled reactants and is capable of explaining the

observed activity and selectivity reported for other metal

catalysts on amphoteric oxide supports.

conducted as described above (section 2.4.1), except that

either anhydrous D₀-phenol or anhydrous D₆-phenol (4.5 g)

was added to the reactor along with an optional addition of

water (deionized or deuterated) or octane. We use the notation

D_xto indicate extent of deuteration, e.g. D₀-phenol is regular

phenol, whereas D₆-phenol is fully deuterated. In all cases, 0.5 g

of additive was used for a ﬁnal concentration of 90 wt %

phenol. A catalytic experiment was also conducted with D₀-

phenol (liquiﬁed, ∼10 wt % water) and 200 μL of NMR quality

D₆-benzene. This reaction was run for a total of 15 min.

Samples from the isotopic labeling experiments were

derivitized with trimethylsilane (TMS) by taking ∼100 μL of

the reaction mixture and mixing it with 50 μL of

hexamethyldisilazine. After shaking the samples for 10 min,

excess silylating agent was removed using a stream of argon gas.

2.4.3. Gas Chromatography Mass Spectrometry (GC/MS).

Two microliters of the reaction product was diluted in 1 mL of

hexadecane. The samples were injected without a solvent delay.

Samples were analyzed using an Agilent 689N Network GC

System with a 6890 series injector and 5973N network mass

selective detector with a HP-5MS cross-linked 5% PH ME

siloxane capillary column (dimensions of 30 m × 0.25 mm ×

0.25 μm) using UHP He. The analysis method contained a 6

min hold at 308 K, followed by a ramp at 5 K/min to 398 K.

This temperature was held for another 10 min before ramping

at 30 K/min to 573 K, which was held for 10 min. The mass

detector was turned oﬀ after the elution time of all potential

products but before the solvent (hexadecane) eluted. Product

yields were obtained from calibrated GC/MS spectra.

2. EXPERIMENTAL SECTION

2.1. Materials. Titanium dioxide (Aldrich, 21 nm particle

size, 33−55 m²g⁻¹surface area), RuCl₃·3H₂O, phenol,

hexadecane, cyclohexane, cyclohexene, cyclohexanol, cyclo-

hexanone, hexamethyldisilazine, trace metal grade nitric acid,

and trace metal grade hydrochloric acid were all purchased

from Sigma-Aldrich. Deuterium oxide, deuterated phenol, and

deuterated benzene were purchased from Cambridge Isotopes.

2.2. Catalyst Preparation. Supported ruthenium catalysts

were prepared by a wet impregnation method, as previously

reported, but without high-temperature calcination.^3aCatalysts

were reduced in a 25 mL Parr reactor. After sealing an

appropriate amount of catalyst in the reactor, the temperature

of the system was increased under ﬂowing H₂gas at a rate of 20

K min⁻¹until it reached the 573 K set point. Upon temperature

equilibration, the system was maintained under ﬂowing H₂for

0.5 h. At that point, the reactor was sealed and pressurized with

H₂to 550 psig and maintained at this temperature for 1 h. At

the end of this reaction time, the reactor was cooled to room

temperature before venting the remaining H₂. This reduced

catalyst was quickly weighed in air before being immediately

transferred back to the reactor along with the phenol.

2.4.4. Isotopic Distribution Analysis. For HDO reactions

conducted with isotopically labeled starting materials, the

isotopologue distribution of the products and remaining

starting materials were determined by non-negative least-

squares ﬁtting of a portion of the experimental MS data with a

reference array of simulated isotopologue MS data.

2.3. Catalyst Characterization. 2.3.1. Metal Determi-

nation. Ru content analysis was performed by Galbraith

Laboratories or done using a Thermo Scientiﬁc iCAP 600 ICP-

OES spectrometer with microwave-assisted digestion. Addi-

tional details on metal analysis are provided in the Supporting

Information (SI).

Experimental MS data were extracted from the GC/MS

spectra of the reaction products. Under our GC/MS

conditions, benzene and cyclohexane elute at very similar

retention times, so the experimental MS intensity data for the

entire elution region containing both compounds was

considered simultaneously. Samples of TMS-protected phenol

were prepared separately (see above), so the MS of this

material was extracted and analyzed independently. The

analysis window was 73−97 m/z for benzene/cyclohexane

and 161−175 m/z for TMS-protected phenol. The exper-

imentally observed intensities for each reaction are shown in

Table S1.

To prepare the reference MS matrices for all isotopologues,

initial reference spectra were obtained from the NIST

Webbook¹⁶for benzene and cyclohexane or from a MS of an

authentic sample of TMS-protected phenol. The intensity

values for the molecular ion and several lighter fragments were

then extracted from this data: 73−78 m/z for benzene, 81−84

m/z for cyclohexane, and 161−166 m/z for TMS-protected

phenol. MS data for heavier deuterium-containing isotopo-

logues were simulated using a simple probability-based

approach for loss of H or D in the molecular fragments,

which was shown to give an approximation very close to the

experimentally observed mass spectrum of D₁-benzene.¹⁷To

account for heavier ¹³C-containing isotopologues, the reference

MS data for each deuterium-containing isotopologue were

convoluted with the ratios of the ¹³C isotopologues calculated

from the natural isotopic abundance of ¹²C and ¹³C. The full

2.3.2. High-Resolution TEM (HRTEM). High-resolution TEM

was carried out at the MIT Center for Material Science and

Engineering (CMSE) using the JEOL 2010 Advanced High

Performance TEM. The catalysts were dispersed in isopropyl

alcohol, and a drop of this suspension was placed on a lacey

carbon Cu grid. Simultaneously with TEM analysis of the

selected samples, energy dispersive X-ray microanalysis (EDX)

experiments were also performed on selected regions of the

catalysts to conﬁrm the presence of Ru metal and to

demonstrate the absence of residual chlorine.

2.4. Catalytic Conversion of Phenol. 2.4.1. Standard

Reaction Conditions. The phenol HDO reactions were carried

out in a 25 mL autoclave reactor operating in batch mode. In a

typical reaction, liqueﬁed phenol (5 g, ∼10 wt % water, Fisher

Scientiﬁc) was introduced into the reactor along with ∼100 mg

of freshly reduced catalyst. The system was closed, and to avoid

any air contamination, H₂was bubbled through the solution for

10 min. This was followed by three reactor purges with 75 psig

of H₂. The closed reactor, still under H₂atmosphere, was

heated to the reaction temperature of 573 K while stirring at

700 rpm. When the reaction temperature was reached, the total

pressure was adjusted to 650 psig (45.8 bar) by regulating the

H₂pressure. The pressure was kept constant during the

progress of the 1 h experiment. After the reaction, the reactor

was cooled to room temperature. Samples were immediately

frozen until they could be analyzed by GC/MS.

2.4.2. Modiﬁed Reaction Conditions for Isotopic Labeling

and the Eﬀect of Additives. These experiments were

6511

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

reference matrices for benzene (73−86 m/z), cyclohexane

(81−96 m/z), and TMS-protected phenol (161−174 m/z) are

found in Tables S2−S4. The matrices for benzene and

cyclohexane were combined before ﬁtting the data. The raw

and normalized non-negative least-squares ﬁt coeﬃcients are

found in Table S5. Plots of the experimental MS data versus the

ﬁt-simulated MS data are shown in Figures S1−S5.

experimental data.²⁶The TiO₂(110) surface was represented in

a 5 × 4 periodic unit cell with four TiO₂layers, of which the top

two layers were fully relaxed, while the bottom two layers were

ﬁxed at their bulk positions. Subsequent slabs were separated by

16 Å of vacuum space in the z direction, a dipole correction was

applied to compensate for the eﬀect of adsorbing molecules

only on one side of the surface,²⁷and spin polarization was

included. The Brillouin zone was integrated using a (2 × 2 × 1)

Monkhorst−Pack mesh,²⁸and geometries were optimized

using a force convergence criterion of 0.05 eV/Å. Convergence

with respect to the k-point set and the force criterion were

conﬁrmed.

2.5. Computational Approach. Density functional theory

(DFT) calculations were carried out using the Vienna Ab-initio

Simulation Package (VASP) in combination with the Atomic

Simulation Environment (ASE).¹⁸The projector augmented

wave (PAW) method¹⁹and the generalized gradient approx-

imation (GGA) with the revised Perdew−Burke−Ernzerhoﬀ

(RPBE) functional were employed to solve the Kohn−Sham

equations.²⁰All DFT calculations were performed with a cutoﬀ

energy of 400 eV and a Gaussian smearing of k_BT = 0.1 eV with

subsequent extrapolation of the total energies to k_BT = 0 eV.²¹

To correct for on-site Coulombic interaction of the localized Ti

3d state, we used the DFT+U approach by Dudarev et al. with

The binding energies of stable intermediate states (E_BE) were

calculated with respect to the clean surface and gas phase

reference molecules according to E_BE= E_{slab+adsorbate}− (E_slab

+

E_gas), where E_{slab+adsorbate}is the total energy of the slab with

adsorbates; E_slabis the energy of the clean surface; and E_gasis a

gas phase reference state based on phenol, water, and hydrogen.

All reported DFT energies are electronic energies at T = 0 K

unless otherwise noted. Where the use of Gibbs free energies is

indicated, those were calculated according to ΔG = ΔE + E_ZPE

− TΔS (ZPE = zero point energy) at T = 573 K, P_phenol= 11.43

bar, P_H= 33.37 bar, and P_{H O}= 1.00 bar.

U

_eff= 2.0 eV.²²This value is in good agreement with the value

obtained by Hu and Metiu.²³To model the Ru/TiO₂interface,

we emulated our previous approach to modeling the Au/TiO₂

interface.²⁴We placed a 10 atom Ru cluster, initially in the hcp

structure, on a fully hydroxylated rutile h-TiO₂(110) surface,

the most stable surface facet of TiO₂, as shown in Figure 3. The

2

The climbing image nudged elastic band method was

implemented to locate the transition state (TS) of each

elementary reaction step with ﬁve or six intermediate images,

which were fully optimized to a force criterion of 0.1 eV/

Å.^24,29,30We conﬁrmed that this convergence criterion is

suﬃcient to obtain transition states within 0.1 eV accuracy, a

generally accepted error bar in DFT calculations. Vibrational

analysis was performed to conﬁrm the existence of a single

imaginary mode, corresponding to the reaction coordinate

along the reaction pathway. For each calculated elementary

reaction step, the hydrogen coverage was kept minimal by

adding or removing 1/2 H₂to or from the surface as needed.

We note that the associated adsorption/desorption energy

changes depend on lateral surface interactions within the local

surface environment to or from which the H atom is added or

removed. These interactions are included in the hydrogen

adsorption/desorption energies, leading to variations of these

values.

To study the eﬀect of water on phenol HDO, we investigated

Figure 3. Side and top views of Ru₁₀/h-TiO₂(110) with enumerations

at the locations used in the DFT calculations to investigate the eﬀect

of water. (1) Water adsorbed in the V_Ovacancy site near the Ru₁₀/h-

three plausible scenarios, as shown in Figure 3: (1) water

br

adsorbed in a bridging oxygen vacancy site (V_O) at the Ru₁₀/h-

TiO₂interface. (2) A gas-phase water molecule used to approximate

the eﬀects of the additive water. (3) Water adsorbed directly on the

Ru₁₀cluster. For visual clarity, the bridging hydroxyl group shown in

black dotted circles was removed in the side view. Color code:

hydrogen, white; oxygen, red, ruthenium, teal; and titanium, light gray.

TiO₂interface, (2) a single gas-phase water molecule near a

V_Osite on h-TiO₂, and (3) water adsorbed on the Ru metal

cluster. The binding energies (E_BE) for water in these positions

are (1) −0.06, (2) −0.11, and (3) −0.40 eV, respectively. For

comparison, the water binding energy on the ﬂat Ru(0001)

surface is E_BE= −0.14 eV.

br

Ru₁₀cluster was placed on three adjacent bridging oxygen

vacancy sites, which can serve as nucleation sites for cluster

growth.²⁵The supported Ru₁₀/h-TiO₂(110) cluster model is

representative for very small particles, whereas the more

frequently used periodic Ru(0001) surface is a model for very

large particles. Comparisons based on these two extreme

models, neither one of which is an exact representation of the

real catalyst, enable a computationally aided interpretation of

the experimental measurements and allow rationalizing the

observed particle size and support eﬀect.

3. RESULTS AND DISCUSSION

There are at least two separate reduction reaction pathways by

which phenol can react on a supported Ru catalyst in the

presence of hydrogen (Figure 1).^4a,31In the hydrogenation

pathway (HYD), phenol is involved in a multistep hydro-

genation sequence to form the ketone cyclohexanone and the

alcohol cyclohexanol. Dehydration of cyclohexanol provides the

ﬁrst deoxygenated product cyclohexene, which is rapidly

reduced to the ultimate deoxygented product cyclohexane.³²

The second pathway is a direct deoxygenation (DDO) that

The optimized lattice constants for rutile TiO₂are a = 4.712

Å, c/a = 0.640, and u = 0.306, in good agreement with

6512

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

Figure 4. High-resolution transmission electron microscopy (HRTEM) images of Ru/TiO₂(a) after H₂reductive activation and (b) after a

subsequent calcination. Both ﬁgures are magniﬁed 200 000× at 200 kV. The black scale bar represents 5 nm.

a

Table 1. Results of Catalytic Tests of Ru/TiO₂for Phenol HDO

a

The product distribution is reported as the average percent of each product formed relative to all phenol products. Percent conversion (% conv)

gives the percentage of phenol converted to products. The percentage of all oxygen-free products (% deox) and the ratio of products that come from

direct deoxygenation (benzene) to those that come from the hydrogenation pathway (all other products) are reported.

converts phenol into benzene without saturating its aromatic

ring.³³

mechanism, we conducted three sets of experiments: a

reference experiment with liqueﬁed phenol (10 wt % water);

an experiment with anhydrous phenol; and an experiment with

a nonpolar additive (octane).

The reference experiments (Table 1, entry 1) were done

under conditions similar to our previously published work, with

only a minor modiﬁcation to the pretreatment conditions.^3a,35

The product selectivity is similar to that reported in our earlier

work; however, slightly higher conversion is reported in this

paper (30% conversion in this paper vs 12% in prior work),

which we attribute to the more aggressive pretreatment

conditions.³⁵

Although the overall eﬃciency of the catalysts to

deoxygenate phenol is similar whether water is present or

absent in the reaction, the dominant deoxygenation reaction

pathway is quite diﬀerent. When water is excluded from the

initial reaction, cyclohexane is the major deoxygenated product,

with a DDO/HYD ratio of 0.8 (Table 1, entry 2). In the

presence of water, benzene is the dominant product, with a

DDO/HYD ratio of 19 (Table 1, entry 1).

When octane (10 wt %) is added to anhydrous phenol at the

start of the reaction to reduce the polarity of the reaction

medium, the reaction shows both lower conversion (13%

conversion) and low selectivity toward DDO (DDO/HYD

ratio of 1.6) (Table 1, entry 3). The presence of 10 wt % octane

alters the reaction trajectory and slows the dehydration of

cyclohexanol. Clearly, the presence of water plays an important

role in the overall reaction, as these experiments done with and

without water and with octane demonstrate.

It has been suggested that HDO catalysts can have three

distinct sites: a site for hydrogenation, a site for dehydration

(these two sites would need to work in concert for HDO to

occur), and a site for hydrogenolysis of the C−O bond for

direct deoxygenation.^32e,34We postulate that HYD preferen-

tially occurs on larger Ru particles and that DDO occurs at the

interface between small Ru particles and an activated interfacial

site. Below, we describe our use of HRTEM, isotopically

labeled substrates, and DFT calculations to test this postulate.

3.1. Catalyst Characterization with HRTEM. EXAFS and

CO chemisorption measurements in our previous report

indicated uncalcined Ru/TiO₂catalysts with a particle size of

∼2 nm showed the best DDO activity and selectivity.^3aIn

contrast, calcined Ru/TiO₂catalysts had a particle size of 33

nm and were not selective for DDO, producing only 20%

deoxygenated products from phenol.^3aTo conﬁrm the eﬀect of

calcination on the particle sizes, we acquired high-resolution

transmission electron microscopy (HRTEM) images on

selected catalysts (Figure 4). These HRTEM images conﬁrm

the presence of small 2−3 nm Ru particles in catalysts treated

only under a high-pressure reducing environment (Figure 4a).

Upon calcination, the Ru particles aggregate to form much

larger clusters of Ru metal (Figure 4b), which is consistent with

our previously reported chemisorption data.^3a

3.2. The Eﬀects of Additives on Catalytic Activity and

Selectivity. To explore the eﬀects of polar and nonpolar

additives as well as the possible role of water on the reaction

6513

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

The primary focus of this work is to develop catalysts that

eﬃciently produce benzene directly from phenol; however, the

pathway where phenol is hydrogenated and then dehydrated

can also lead to products suitable for combustion. Diﬀerent

engines and diﬀerent countries require fuels with diﬀering

compositions, so understanding the structural factors that

control the HYD pathway is also of interest. As described in our

prior work, under standard pressure at 573 K, both formation

of cyclohexanone and reduction to cyclohexanol are thermo-

dynamically unfavorable, but the subsequent steps in the HYD

pathway are favorable.^3aThe dehydration of cyclohexanol has

been used as a probe of surface acidity of heterogeneous

materials.^36,37In this work, the ratio of cyclohexanol to

cyclohexane varies from a low of 0.075 in the reaction with no

water additive to a ratio of 0.63 in the reaction with octane

additive, suggesting that there are fewer acidic sites suitable for

catalyzing the dehydration of cyclohexanol when nonpolar

octane is added as an additive.

3.3. Isotopic Labeling Experiments. To distinguish

between the various proposed mechanisms for DDO, several

reactions were carried out with isotopically labeled substrates.

Figure 2 provides the expected isotopic signature of benzene

from each pathway when the starting material is D₀-phenol and

the reductant is D₂. Mechanism A, a direct substitution of a

hydride for a hydroxide, should generate benzene with a single

reductive hydrogen substituting the phenolic hydroxyl. The

isotopic composition of benzene from mechanism B, ring

hydrogenation followed by dehydration, and mechanism C,

hydrogenation of the ketone tautomer of phenol followed by

dehydration, is a bit more complicated to predict, but in each

case, the isotopic composition of benzene should include more

than one of the reductive hydrogen atoms.

D₀-phenol and a sample of activated Ru/TiO₂catalyst. This

reaction mixture was heated to the normal reaction temperature

(573 K) under an atmosphere of inert gas (N₂). The isotopic

distribution of phenol observed at the end of this reaction is

shown in Figure 5a. D₀-phenol remains the major isotopologue

Figure 5. Isotopic distributions for benzene (blue) and phenol (red)

for two control experiments. (a) The isotopologue distribution of

phenol when a mixture of D₀-phenol and D₂O was used without H₂

showing H/D exchange due to tautomerization. Phenol was derivitized

prior to analysis, so it contains a maximum of ﬁve H + D atoms. (b)

The isotopologue composition of benzene after a reaction in which a

spike of D₆-benzene was added to D₀-phenol with H₂O and H₂at the

start of the reaction. This experiment indicates that H/D metal

hydride exchange for the benzene product (via Scheme 2) is very

minor under these conditions. (Note: D₆-benzene overlaps with D₀-

cyclohexane in the MS. However, very small amounts of cyclohexane

are typically observed in reactions with this catalyst under these

conditions, so it can be assumed that the majority of the intensity

shown here comes from D₆-benzene.).

in this mixture; however, the presence of D₁- (∼22%) and D₂-

phenol (∼4%) suggests that tautomerization occurs and that

care must be taken in the interpretation of other isotopic

labeling experiments.

3.3.2. H/D Ring Exchange in Benzene under Reaction

Conditions. Aromatic H/D exchange is commonly observed in

metal hydride-catalyzed reactions (Scheme 2)^38,39,40and has

The isotopic labeling experiments in this work were

conducted with diﬀerent mixtures of isotopically labeled

substrates, and the ﬁnal isotopic compositions of the products

were determined using simple least-squares ﬁtting methods of

the acquired GC/MS data. Phenol undergoes an inevitable H/

D exchange of the phenolic protons with ambient water during

sample workup and GC/MS analysis.³⁸To minimize the

complications this induces in isotopic analysis, aliquots of each

reaction sample were treated with a silylating agent, which

derivitizes free hydroxyl groups and leaves phenol with a

maximum of ﬁve aromatic protons. These samples were used

for isotopic distribution analysis of phenol.

Scheme 2. Possible Aromatic H/D Scrambling Pathways for

a

Phenol (X = OH) or Benzene (X = H)

3.3.1. H/D Ring Exchange in Phenol from Tautomeriza-

tion. Phenol tautomerization (Scheme 1) provides a route to

a

(a) The H/D exchange of a metal hydride with water.⁴¹(b, c) Metal-

Scheme 1. Deuterium Incorporation into Phenol through

Tautomerization

a

catalyzed H/D ring exchange alters the isotopic composition of the

aromatic ring.

the potential to complicate mechanistic interpretations of

isotopic distributions. To assess the extent of this process under

our reaction conditions, a reaction was conducted in which 200

μL of D₆-benzene was added at the start of a standard HDO

reaction. The isotopic composition of all benzene (spike +

reaction product) at the end of this reaction is shown in Figure

5b.

The results of this experiment indicate that benzene is

eﬀectively inert to H/D ring exchange under these reaction

conditions. D₀- and D₆-benzene are the two most prominent

isotopologues of benzene, each representing >40% of the

product composition. D₀-benzene is the product of the catalytic

a

The initial deuterium is a result of H/D exchange with D₂O.

incorporate the isotopic signature of water into the aromatic

ring.^3qDFT calculations estimate that the water-assisted

tautomerization step at the Ru/TiO₂interface is likely

equilibrated with a moderate activation barrier of E_a= 0.38

eV and ΔE = −0.22 eV (Figure S6 in the SI). The extent of

tautomerization expected under these reaction conditions was

measured by a control reaction consisting of 10 wt % D₂O in

6514

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

DDO of D₀-phenol. The low levels of intermediate

isotopologues are evidence that aromatic H/D exchange is a

minor reaction pathway for benzene under these reaction

conditions. It is also worth noting that this experiment conﬁrms

an assumption made in our earlier work: namely, that benzene

is not fully hydrogenated under these reaction conditions, and

therefore, the cyclohexane produced is entirely derived from

the HYD pathway.^3i,42This is also conﬁrmed by the lack of

heavier cyclohexane isotopologues in this experiment (Table

S5).

3.3.3. H/D Ring Exchange in Phenol from Metal-Catalyzed

H/D Exchange. Direct H/D exchange of the phenolic aromatic

protons, in a manner analogous to that described above for

benzene, would provide another route by which the isotopic

composition of phenol could shift to reﬂect either the isotopic

composition of the reducing gas or the isotopic composition of

water (Scheme 2). An HDO experiment conducted with D₆-

phenol and H₂without water present at the start of the reaction

was used to assess the level of aromatic H/D exchange in

phenol. This reaction was stopped after only 15 min of total

reaction time (rather than 1 h under normal conditions) to

minimize side reactions with water, which is a reaction product.

The distribution of phenol isotopologues in this experiment is

shown in Figure 6. The phenol distribution shifts to lighter

isotopologues as the reaction proceeds, which is indicative of

metal-catalyzed H/D ring exchange.

Figure 7. Isotopologue distributions for benzene (blue) and phenol

(red) for two HDO isotopic labeling experiments. (a) Isotopologue

composition from a reaction with D₀-phenol, 10 wt % D₂O, and H₂.

(b) Isotopologue composition from a reaction with D₆-phenol, 10 wt

% D₂O, and H₂. Phenol was derivitized prior to analysis, so it contains

a maximum of ﬁve H + D atoms.

activity and selectivity but does not directly participate in the

reductive proton transfer.

3.3.5. Mechanistic Insights from Determining the Number

of Reductive Hydrides Transferred to Product. The primary

purpose of the isotope experiments was to elucidate the

reduction mechanism by assessing the total number of

hydrogen equivalents being transferred to benzene (Figure

2). To determine this, we compare the isotopologue

distribution of benzene and phenol in HDO reactions that

use either D₀-phenol or D₆-phenol in the presence of 10 wt %

D₂O and H₂(Figure 7). In both experiments, the isotopologue

distribution of benzene mirrors that seen for phenol. Given that

benzene is inert to H/D exchange under these reaction

conditions, this result implies that the phenolic hydroxyl is

being substituted by a single reductive hydrogen. Other

mechanisms would lead to notably lighter distributions (i.e.

1

more H incorporation) of benzene isotopologues. Figure 7a

shows the nearly identical distribution of benzene and phenol

isotopologues. Taken in total, these data are consistent with the

direct C−O bond cleavage pathway (Figure 2, mechanism A),

but inconsistent with the mechanistic hypotheses that involve

hydrogenation of the tautomerized keto form of phenol or

multiple hydrogenation/dehydration steps (Figure 2, mecha-

nisms B and C), which would be expected to add additional

reductive equivalents to benzene.

Figure 6. Isotopologue distribution for phenol in an HDO experiment

with D₆-phenol and H₂but without water showing that signiﬁcant

aromatic H/D exchange occurs for phenol.

3.4. First Principle Analysis of HDO Pathways on Ru/

TiO₂. To propose a DDO reaction mechanism that is

consistent with our isotopic labeling experiments and can

simultaneously explain the promotional eﬀect of water, multiple

pathways were explored using DFT calculations on a Ru₁₀/h-

TiO₂(110) surface model. The key calculated energetics are

summarized in Table 2. Guided by the experimental

observation that a single reducing hydrogen equivalent is

exchanged, the most probable DDO pathway for phenol in the

absence of water is presented as DDO1 in Figure 8. The

promotional eﬀect of water on DDO activity and selectivity is

then explored for three diﬀerent cases (DDO2−DDO4), which

are also summarized in Figure 9. Our results are contrasted with

data on the thermodynamically most stable Ru(0001) facet,

which is a good representation of larger Ru particles and has

been used in several previous DFT investigations of HDO of

phenolic compounds.¹²All of the most probable DDO

mechanisms explored occur at interfacial sites between small

Ru particles and the TiO₂support. Such interfacial sites are

known to play important roles in heterogeneous catalysis, and

3.3.4. Involvement of Water in H/D Exchange and

Reduction Mechanism. To better understand both the

mechanistic role played by water and its possible involvement

in H/D exchange, an HDO reaction was conducted with 10 wt

% D₂O, H₂, and anhydrous D₀-phenol. The isotopologue

distributions for phenol (red bars) and benzene (blue bars)

from this reaction are shown in Figure 7a. Some deuterium

incorporation into phenol is observed. The levels are slightly

higher than the amount of deuterium incorporation from the

tautomerization experiment (Figure 5a), suggesting that both

tautomerization and metal-catalyzed H/D ring exchange in

phenol (but not benzene) are operative (as illustrated in

Scheme 2) and lead to a change in the isotopic composition of

phenol relative to its initial composition. Deuterium incorpo-

ration is also seen in benzene, but at levels that precisely mirror

that of phenol starting material. The lack of additional

deuterium incorporation into benzene leads to a mechanistic

interpretation that water plays an important role in dictating

6515

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

a

Table 2. Summary of Key Elementary Steps at the Ru₁₀/h-TiO₂(110) Interface Model

b

c

no.

elementary steps

ΔE

E_a

ΔG

G_a

br

I

(C₆H₅)OH_(g)+ V_O+ Ru → (C₆H₅)OH*

C₆H₆−Ru → C₆H_6(g)+ Ru

−1.60

1.40

−0.56

0.41

0.69

e

II

H₂O_(g)+ V_O→ H₂O^br

−0.06

−0.25

0.94

br

III

d

IV

H₂(g) + 2Ru → 2H−Ru

∼0.0

V

H−Ru + HO^br→ H₂O^br+ Ru

1.15

0.11

0.47

0.66

0.42

0.94

0.79

0.18

e

f

VI(a)

VI(b)

VII

VIII

IX

H_2(g)+ Ru + HO^br→ H₂−Ru + HO^br

H₂−Ru + HO^br→ H−Ru + H₂O^br

−0.05

0.22

0.67

0.36

−1.09

−1.25

e

(0.67)

0.51

0.65

0.58

0.90

1.05

0.19

DDO1: (C₆H₅)OH* → (C₆H₅)−Ru + HO^br

DDO2: (C₆H₅)OH* + H₂O^br→ (C₆H₅)−Ru + HO^br+ H₂O_(g)

−1.02

−1.25

−0.83

−0.79

−0.67

g

DDO3: (C₆H₅)OH* + H₂O_(g)→ (C₆H₅)−Ru + HO^br+ H₂O_(g)

DDO4: (C₆H₅)OH* + H₂O−Ru → (C₆H₅)−Ru + HO^br+ H₂O−Ru

(C₆H₅)−Ru + H−Ru → (C₆H₆)−Ru + Ru

X

e

XI

−0.52

a

The table shows the electronic energy change (ΔE), the activation energy (E_a), the Gibbs free energy change (ΔG), and the Gibbs free energy of

br

activation (G_a) for each step. V_Odenotes the bridging oxygen vacancy site on h-TiO₂near the interface. Species X bound only to the Ru cluster are

br

shown as X−Ru. The simultaneous binding of phenol to both the V_Osite and the Ru cluster is indicated with an asterisk, for example, (C₆H₅)OH*.

b

Additional data for the individual contributions of ΔE_ZPEand TΔS are available in Table S6. ΔE and E_aare the total energy change and activation

c

energy barriers without zero point energy (E_ZPE) or entropy correction in electronvolts. Gibbs free energies are calculated as ΔG = ΔE + E_ZPE

−

d

TΔS at T = 573 K, P_phenol= 11.43 bar, P_H= 33.37 bar, and P_{H O}= 1.00 bar. The H₂molecule dissociated spontaneously upon adsorption on a Ru

2

e

f

metal site away from the Ru/TiO₂interface. Values were not calculated. The Gibbs free energy of activation G_ais set equal to ΔG because entropy

g

br

corrections render the calculated Gibbs free energy of the transition state more stable than the ﬁnal state. Instead of binding to the V_Osite, the

phenolic OH group interacts with the H₂O^brmolecule occupying the V_Osite.

br

and begins with its adsorption onto a Ru/TiO₂interface site

br

near a bridging oxygen vacancy (V_O) on TiO₂. The calculated

binding energy of phenol at the interface is exothermic by E_BE

=

−1.60 eV. In comparison, phenol binding to the ﬂat Ru(0001)

surface is weaker, with E_BE= −0.47 eV (Figure S7 in the SI),

indicating a preferential adsorption at the interface site. Binding

occurs primarily by charge transfer from Ru metal to the

adsorbate and strong π interactions between the aromatic ring

and the d-states of the Ru cluster.⁴⁴

The binding of phenol is followed by direct C−O scission,

which requires a moderate activation energy barrier of E_a= 0.66

br

eV. The cleaved OH group heals the V_Osite at the Ru/TiO₂

interface, rendering the reaction exothermic by ΔE = −1.02 eV.

The same step on Ru(0001) is activated by E_a= 1.23 eV

(Figure S7 in the SI), which compares well with previously

calculated values of E_a= 1.15 eV^12aand E_a= 1.29 eV.^12bThe

resulting C₆H₅−Ru intermediate undergoes easy H atom

transfer on the Ru cluster. The reaction C₆H₅−Ru + H−Ru

→ C₆H₆−Ru + Ru has an activation barrier of only E_a= 0.18 eV

and is exothermic by ΔE = −0.67 eV (Figure S8 in the SI). On

Ru(0001), the corresponding activation barrier increases to E_a

= 0.55 eV.¹²

Benzene is also strongly adsorbed on the Ru cluster, and its

desorption requires ΔE = 1.40 eV from the Ru₁₀/h-TiO₂

interface and ΔE = 0.60 eV from the Ru(0001) surface.

Thus, it may at ﬁrst seem as if product desorption could be

rate-limiting, but when Gibbs free energies (ΔG) are

considered under realistic reaction conditions (e.g., conditions

other than T = 0 K and ultrahigh vacuum), the adsorption and

desorption steps are associated with a much smaller Gibbs free

energy change. For example, Chiu et al. have calculated the

ground state desorption energy for phenol and benzene from

Ru(0001) as 1.35 and 1.41 eV and tabulated the Gibbs free

energies at 523 K and 40 bar as 0.50 and 0.55 eV,

respectively.^12bThese values are in good agreement with the

Gibbs free energies of adsorption of phenol and desorption of

benzene shown in Table 2 obtained under the conditions of our

experiments. Temperature and pressure eﬀects on adsorbed

Figure 8. DDO1 pathway of phenol without water at the Ru₁₀/h-

TiO₂(110) interface: (a) calculated geometries of the initial state,

transition state, and ﬁnal state; (b) potential energy surface (PES).

The energy of the initial and ﬁnal state is calculated with respect to

phenol and the Ru₁₀/h-TiO₂model with bridging oxygen vacancy site

(V_O). At the transition state (TS), the activation barrier, E_a, is

indicated in bold face. Color code: hydrogen, white; oxygen, red;

carbon, gray; ruthenium, teal; and titanium, light gray.

br

all of our computational results implicate them in the observed

chemistry reported in this work.⁴³

3.4.1. Direct Deoxygenation (DDO) in the Absence of

Water. The DDO of phenol without water is shown in Figure 8

6516

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

3.4.2. Site Regeneration: Vacancy Formation and Hydro-

gen Activation. After phenol dehydroxylation via DDO1, the

br

V_Osite at the Ru/TiO₂interface is healed with the abstracted

OH group, denoted as HO^br. To close the catalytic cycle and

regenerate the oxygen vacancy, several mechanistic possibilities

were explored, including H₂activation on a Ru site followed by

spillover of atomic H. Although H₂dissociation is nonactivated,

the spillover from Ru to TiO₂requires an activation barrier of

E_a= 1.15 eV (Table 2). The most likely pathway according to

our study is a TiO₂-assisted heterolytic cleavage of H₂, as

illustrated in Figure 10. It is initiated with activated H₂

adsorption on the Ru cluster near the TiO₂support, which

has an E_a= 0.11 eV and ΔE = −0.05 eV (Figure 10a). The

adsorbed H₂species then dissociates heterolytically across the

Ru/TiO₂interface to protonate a bridging hydroxyl group

(HO^br) on TiO₂, as shown in the enlarged transition and ﬁnal

state of Figure 10b. The protic character of the hydrogen being

transferred to the support is evidenced by a fractional Bader

charge of +0.38 e⁻in the transition state. This dissociation

mechanism generates a support Brønsted acid site and an H

adatom, with signiﬁcant negative charge density, on the bridge

site of two Ru atoms. The calculated energy barrier and

reaction energy for this H₂dissociation step are E_a= 0.47 eV

and ΔE = 0.22 eV, respectively. After H₂O^bris formed at the

br

interface, it may desorb as water and recreate the interfacial V_O

site with Lewis acid character. The H atom remaining on the

Ru cluster is required for the hydrogenation of C₆H₅−Ru to

benzene, as described earlier. H₂dissociation on the Ru cluster

followed by atomic hydrogen spillover to TiO₂with or

without the assistance of water as suggested by Xi et al.⁴⁷was

also tested; however, both atomic hydrogen spillover pathways

involved at least one step with a barrier of ∼1.1 eV or more

(Figure S11 in the SI).

3.4.3. HYD Reactivity. Molecular hydrogen adsorbs

dissociatively without barrier on a Ru site of Ru/TiO₂, and

hydrogen delivery can be assumed to be quasi-equilibrated.

Hence, to assess the HYD pathway, only the activation barrier

for the ﬁrst phenol hydrogenation step was calculated. On

Ru(0001), hydrogen addition to the ortho position is more

favorable than addition to either the meta and para positions on

phenol. The calculated activation barrier for H addition at the

ortho position is E_a= 0.91 eV on Ru(0001) (Figure S7 in the

SI) and E_a= 0.81 eV on the supported Ru cluster (Figure S12

in the SI). These energetics are in line with the same reaction

pathway studied on water-solvated Pt(111) and Ni(111), which

fell between 0.84 and 1.00 eV.^6a

Figure 9. Potential energy surface (PES) of the H₂O-assisted DDO

pathways. Initial and ﬁnal state energies in electronvolts are given with

respect to phenol and the respective (water-modiﬁed) Ru/TiO₂model

with oxygen vacancy (see Figure 3). Activation energy barriers (E_a) in

electronvolts are shown in bold face, and the roman numerals refer to

the elementary steps in Table 2. The dotted black line serves as a

reference to the DDO1 pathway in the absence of water, as shown in

Figure 8. The ﬁnal state energies vary slightly as a result of diﬀerent

binding orientations of benzene on the Ru particle and diﬀerent

binding environments, that is, the presence of H₂O.

intermediate and transition states are generally small.^12bAs

shown in Table 2 and contrasted in Figure S9 of the SI, the

Gibbs free energies change, and activation barriers for surface

steps are similar to the ground state electronic energies. This

allows us to neglect entropy contributions for our qualitative

discussion of surface reaction mechanisms. Overall, our analysis

suggests that the kinetics of the DDO pathway on Ru/TiO₂

and Ru(0001) are largely determined by the initial C−O

scission step and that the Ru/TiO₂interface site is more active

than Ru(0001).

3.4.3. DDO/HYD Selectivity and the Eﬀect of Particle Size.

Our calculations show that the Ru₁₀/h-TiO₂interface with a

br

bridging V_Ovacancy site, representative of small Ru particles

on partially reduced TiO₂, has very good DDO activity (E_a=

0.66 eV) and moderate HYD activity (E_a= 0.81 eV). The

Ru(0001) surface, representative of large Ru particles, has low

HYD activity (E_a= 0.91 eV) and even lower DDO activity (E_a

= 1.23 eV). Thus, small Ru particles with a large number of

metal/support interface sites should have higher DDO/HYD

selectivity and activity compared wtih large Ru particles with

fewer interface sites.

This theoretical assessment of the DDO and HYD pathways

and activity/selectivity trends for Ru₁₀/h-TiO₂and Ru(0001) is

consistent with the experimental data. The HRTEM images in

Figure 4 showed fairly uniform 2−3 nm Ru particles on the H₂-

activated catalysts. These small particles have relatively more

Ru/TiO₂interface area in comparison with bulk Ru, which is

br

Phenol DDO on a V_Ovacancy site of the fully hydroxylated

h-TiO₂(110) surface in the absence of a Ru cluster was also

considered. Although vacancy formation by hydrogen activation

and water elimination on h-TiO₂is kinetically unfavorable,⁴⁵it

has been suggested that a vacancy site may form via hydrogen

spillover in the presence of transition metal clusters.⁴⁶If such a

vacancy site exists, then the calculated activation energy barrier

of the direct C−O scission step is E_a= 1.02 eV (Figure S10 in

the SI), which is lower than on Ru(0001), yet 0.36 eV larger

than at the Ru/TiO₂interface. Thus, the Ru/TiO₂interface site

remains the most favorable site for DDO.

6517

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

Figure 10. DFT calculated models for hydrogen binding and heterolytic cleavage at the Ru₁₀/h-TiO₂(110) interface. (a) H₂adsorbs on the Ru

particle: H_2(g)+ Ru → H₂−Ru. (b) Heterolytic dissociation of H₂−Ru occurs across the Ru/TiO₂interface: H₂−Ru + HO^br→ H−Ru + H₂O^br. The

charge density diﬀerence, Δρ = ρ_{H Ru/TiO}− (ρ_RuTiO+ ρ_H), shown with blue and red represent charge accumulation and depletion, respectively. The

2

numerical values indicate the Bader charge at the transition state and ﬁnal state. Color code: hydrogen, white; oxygen, red; ruthenium, teal; and

titanium, light-gray.

consistent with the most favorable DDO mechanism (DDO1).

The aggregated Ru particles (∼30 nm) obtained after

calcination show reversed selectivity toward HYD, in agreement

with our predictions for the Ru(0001) surface model. The

proposed HYD and DDO pathways proceed independently,

that is, ring hydrogenation is not necessary for C−O cleavage,

which is consistent with observed product distributions.^3j

Although this proposed mechanism and active site model is

congruent with the vast majority of experimental observations,

it cannot explicitly account for the promotional eﬀect of water.

3.4.4. The Mechanistic Role of Water: Cocatalyzing C−O

Cleavage. The additive exchange experiments clearly demon-

strate a strong beneﬁcial eﬀect of water on DDO selectivity and

are in agreement with the observations of others.^9cTo

investigate the mechanistic role of water during DDO, three

scenarios were considered in which water might assist the C−O

scission step.

experiments using D₀-phenol, with or without D₂O, which

exclude the possibility of direct water involvement during the

reductive proton transfer, that is, the isotopic signature from

water does not end up in benzene, except to the extent that it is

exchanged into phenol through one of the two possible

mechanisms discussed earlier. As discussed for DDO1, the

benzene radical (C₆H₅−Ru) on the Ru particle reacts rapidly

with the remaining H−Ru adatom.

The precise details of the C−O bond-breaking and C−H

bond-forming steps are diﬃcult to determine from our work

and may depend on the hydrogen surface coverage. Lu and

Heyden have considered the inverse sequence consisting of an

initial hydride attack to the ispo carbon of phenol on Ru(0001)

followed by C−O bond-breaking.^12aTheir reported barriers

along both alternative routes are within 0.1 eV, yet their

microkinetic model predicts a ∼2.5 times faster rate for C−H

bond-forming followed by C−O bond-breaking, presumably

caused by higher hydrogen coverage. In Figure S13 of the SI,

we present a similar sequence of HYD, followed by C−O

scission for the water-assisted DDO2 mechanism at 1/9 ML

hydrogen coverage; that is, one hydrogen atom is adsorbed on

nine exposed Ru atoms. We ﬁnd that hydrogen does not

change the adsorption strength of phenol as long as it does not

In the ﬁrst water-assisted scenario, DDO2, H₂O^brformed

during the site regeneration process does not desorb from the

br

V_Osite, that is, it remains in position 1 in Figure 3. This is

likely to occur in water-rich environments that would shift the

equilibrium to the surface-bound water state. As explained

earlier in section 3.4.2, this motif is created by heterolytic H₂

bond dissociation, which leads to the creation of a Brønsted

acidic H₂O^brgroup. Brønsted acids can catalyze the elimination

of OH groups from saturated reaction intermediates (e.g.,

cyclohexanol) during HDO.^9b,15In the DDO2 pathway shown

in Figure 9, H₂O^brwas retained on the surface and acts as a

proton donor to the hydroxyl group of phenol. Although

phenol binding is weaker at this active site model (E_BE= −1.29

eV), the proton-assisted activation barrier is greatly reduced (E_a

= 0.42 eV) compared with the nonassisted pathway, DDO1

(Figure 8). This mechanism is consistent with data from prior

researchers⁴⁸and supports the speculation that protonation of

the lone pair orbital of oxygen could facilitate C−O bond

cleavage.

br

block the Ru sites near the V_Ovacancy. We also ﬁnd that the

adsorption complex after the ipso C−H bond formation step

has a binding energy of −1.48 eV, which is even 0.19 eV

stronger than phenol binding along the DDO2 pathway.

Furthermore, the activation barrier for H₂O-assisted C−O

scission remains identical at E_a= 0.42 eV. These ﬁndings do not

implicate a strong hydrogen coverage dependence of the

promotional eﬀect of H₂O on the C−O dissociation barrier; yet

a stronger eﬀect at higher coverages cannot be excluded.

Despite several attempts, a direct hydride attack pathway to

generate a transition state, in which C−H bond formation and

C−O bond breaking occur in concert, was not identiﬁed. In

practice, both bond-forming/-breaking sequences are compet-

itive, and the actual mechanism may depend on the extent of

hydrogen coverage on Ru, which in turn depends on reaction

conditions.^12aA full microkinetic model, including all relevant

An important observation about this pathway is that the

proton originating from H₂O^bris not incorporated into the

aromatic ring. This is consistent with our isotopic labeling

6518

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

Table 3. Estimated Relative Reaction Rates at T = 573 K for the Four Diﬀerent HDO Pathways Based on the DFT-Calculated

a

Binding and Activation Energies Tabulated in Table 2

b

no.

ΔG_BE, eV

−0.56

E_a, eV

ΔS^‡, eV K⁻¹

k⁺₁, bar⁻¹s⁻¹

k⁻₁, s⁻¹

k⁺₂, s⁻¹

θ_phenol

r_norm

DDO1

DDO2

DDO3

DDO4

0.66

0.42

0.94

0.79

−0.00012

−0.00021

−0.00007

−0.00011

8.86 × 10⁰³

9.22 × 10⁰⁴

2.06 × 10⁰³

8.59 × 10⁰³

9.70 × 10⁻⁰²

5.06 × 10⁰¹

6.14 × 10⁻⁰⁴

1.22 × 10⁻⁰³

4.28 × 10⁰⁶

2.19 × 10⁰⁸

2.79 × 10⁰⁴

3.62 × 10⁰⁵

2.32 × 10⁻⁰²

4.80 × 10⁻⁰³

4.59 × 10⁻⁰¹

2.14 × 10⁻⁰¹

1.0

10.6

0.1

c

−0.37

−0.74

c

−0.78

0.8

a

ΔG_BEis the Gibbs free binding energy of phenol, ΔS^‡is the entropy change from adsorbed phenol to the transition state of C−O bond scission,

and the fugacity of phenol is taken as f_phenol= 11.48 bar. Rate is normalized to DDO1. Only the entropy change for phenol was considered in the

calculation of ΔG_BE.

b

c

steps, is required to address these details, but our main

conclusion that the presence of H₂O lowers the C−O scission

barrier appears to be independent of hydrogen coverage.

3.4.5. Alternative Role of Water: Phenol Stabilization. In

addition to actively lowering the C−O scission barrier, water

may passively stabilize reaction intermediates and transition

states by pure solvation eﬀects without transferring a proton.^6a

Rather than attempting to fully model the aqueous reaction

environment, two simpliﬁed systems were considered: (1)

physisorption of a single water molecule near the phenolic OH

group (position 2 in Figure 3), and (2) water chemisorption on

the Ru cluster (position 3 in Figure 3). These two pathways,

DDO3 and DDO4, respectively, are illustrated in Figure 9. In

the DDO3 pathway, a single physisorbed (≈ gas-phase) water

molecule is located near the hydroxyl group of phenol, which

cluster, making it diﬃcult to extrapolate the quantitative results

of this pathway to other scenarios.

Calculations on the role of water in the DDO2 mechanism

have led to a number of conclusions. Most importantly, a

Brønsted acidic H₂O^bris able to act as a hydrogen bond donor

to the ruthenium-bound phenol. This interaction decreases the

activation barrier for C−O bond scission, which was thought to

be the rate-determining step in our water-free model pathway

DDO1. Notably, the formation of a vacancy site on TiO₂is not

required. This should lead to increased DDO activity in the

presence of water, which mirrors the experimental increase in

activity seen with water present (Table 1; compare entries 2

and 1). This observation is also in agreement with the

conclusions of Behtash et al., who noted that polar solvents

such as water and n-butanol not only increase the catalytic

activity for HDO of propanoic acid but also stabilize the key

intermediates.⁴⁹When water acts as a hydrogen bond acceptor

from bound phenol or is passively bound to the ruthenium

cluster, decreased binding energies are observed, but the eﬀect

on the C−O cleavage energy barriers is minimal.

3.5. Kinetic Evaluation of Water-Assisted DDO Path-

ways. A simple kinetic model was developed to discriminate

between C−O scission mechanisms in the possible DDO

pathways. This model should not be understood as a

quantitative microkinetic model; it is rather a tool that allows

us to predict whether the favorable binding and higher coverage

of phenol in DDO3 and DDO4 can compensate for the larger

C−O cleavage barrier when compared with DDO2. In this

model, the adsorption of phenol is assumed to occur reversibly

on the (water-modiﬁed) vacancy site at the interface of Ru/

TiO₂denoted as “*”, followed by irreversible C−O scission.¹³

Note that for weak phenol binding, the desorption of phenol

may become faster than the C−O scission reaction, which will

lower the consumption of phenol. Importantly, we do not

assume that phenol adsorption is quasi-equilibrated, but that

br

adsorbs in the interfacial V_Osite (see the schematic of DDO3

in Figure 9). By acting as a hydrogen bond acceptor, the water

molecule stabilizes the adsorbed phenol by −0.25 eV, resulting

in a stronger phenol binding energy of E_BE= −1.85 eV, which is

consistent with hydrogen bonding between water and the

hydroxyl group of phenol. This stabilization, however, does not

translate to a lower energy transition state barrier for C−O

cleavage. The barrier for DDO3 (solid blue line in Figure 9) is

E_a= 0.94 eV, therefore 0.28 eV higher than the barrier for

DDO1 in the absence of water (E_a= 0.66 eV). Hence, this

scenario, which approximates the eﬀect of water on hydrogen

bonding, is not alone capable of explaining the promotional

eﬀect of water on DDO activity.

The DDO3 pathway is, in fact, similar to the proton-assisted

mechanism (DDO2), and the drastic activity diﬀerence

between these two paths can qualitatively be understood in

terms of reversed electronic interactions. In DDO2, hydrogen

bond donation to the departing phenolic hydroxyl decreases

the electron density in the C−O bond, facilitating its cleavage.

In DDO3, however, the phenolic oxygen acts as a hydrogen

bond donor, increasing the electron density in the C−O bond,

making it more diﬃcult to cleave.

Another potential role for water is modulating the density of

states by binding to ruthenium near the active site (DDO4; see

the schematic of DDO4 in Figure 9). Again, the presence of

water stabilizes the initial binding of phenol by −0.22 eV

relative to DDO1 (E_BE= −1.82 eV) while having relatively little

eﬀect on the activation energy (E_a= 0.78 eV for DDO4 vs E_a=

0.66 eV for DDO1). Overall, the potential energy diagram for

DDO4 (solid green line in Figure 9) suggests that water

chemisorbed on the Ru cluster lowers the potential energy of

the initial and transition state equally, without signiﬁcantly

changing the C−O bond breaking barrier. We note, however,

that the eﬀect of adsorbing water molecules on a Ru cluster will

depend on the number of water molecules or size of the metal

the fractional phenol coverage is at steady-state (θ_phenol

=

constant). As indicated by our DFT analysis, the benzene

radical formed after the C−O scission step is quickly

hydrogenated to the ﬁnal DDO product benzene. Because

the hydrogenation barrier of E_a= 0.18 eV is so low, we make

this last assumption irrespective of the presence or absence of

water. Finally, it is suﬃcient for our purposes to consider only

the intrinsic activity of each active site model, and we do not

attempt to estimate the concentration of vacancy sites or water

coverage. This reduces the kinetic analysis to the three

elementary steps, of which only the ﬁrst two are kinetically

relevant. The simpliﬁed mechanism is then given by

phenol + * ⇋ phenol*

with r = k₁⁺f

θ* − k₁⁻θ_phenol

1

phenol

(1)

6519

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

undercoordinated Zr⁴⁺sites with Lewis acid character,^3qwhich

is similar to the role our calculations indicate TiO₂is playing.

As an intrinsic biofuel component and a reaction product,

water is a key player in phenolic HDO reactions, and as such,

its presence on the catalyst surface cannot be avoided. Most

previous studies have neglected to account for the presence of

water, but our results conﬁrm that water greatly enhances HDO

activity and selectivity for the DDO pathway. On the basis of

isotopic labeling studies and DFT simulations, we conclude that

selective DDO of phenol occurs on a bifunctional reaction site

at the interface between Ru nanoparticles and a Brønsted acidic

surface hydroxyl/water on the TiO₂support. In the proposed

proton-assisted DDO mechanism (DDO2, Figure 9), phenol

binds at this interfacial site. It is here that the acid site catalyzes

C−O cleavage while the Ru metal component catalyzes the

hydride attack. In this case, the active site motif is regenerated

through heterolytic H−H bond cleavage, yielding a ruthenium

hydride and a Brønsted acidic water on the surface of

hydroxylated TiO₂(Figure 10). A number of homogeneous

Ru-based hydrogenation catalysts, for example, Shvo’s catalyst,

show a similar ability to simultaneously transfer acidic and

hydridic hydrogens in the reduction of carbonyl substrates and

ultimately regenerate the active site through H₂bond

heterolysis.⁵¹The proposed proton-assisted DDO mechanism

(DDO2) does not require formation of an oxygen vacancy site

on the reducible TiO₂support.

phenol* → C₆H₅−Ru + HO^brwith r₂≈ k₂⁺θ_phenol

(2)

C₆H₅−Ru + HO^br+ H₂→ benzene + H₂O + * (fast)

(3)

Assuming steady state for the fractional coverage of phenol

θ_phenolleads to r₁− r₂= k⁺₁f_phenolθ* − k₁⁻θ_phenol− k₂⁺θ_phenol= 0,

which allows the estimation of θ_phenolas a function of the

fugacity of phenol (f_phenol).

k₁⁺f

phenol

θ_phenol

=

k₂⁺+ k₁⁻+ k₁⁺f

phenol

(4)

The site balance θ_phenol+ θ* = 1 was used to obtain eq 4. The

reaction rate can then be approximated from eqs 2 and 4 as

k₁⁺f

phenol

r = r₂= k₂⁺

k₂⁺+ k₁⁻+ k₁⁺f

phenol

(5)

The fugacity of phenol at T = 573 K is approximated as its

pure vapor pressure of 11.48 bar.⁵⁰The DFT-derived energetics

from Table 2 were used to calculate the expected reaction rates

of the four diﬀerent DDO pathways; the results are summarized

in Table 3. Herein, the preexponential factors for the rate

constants k_i= A_iexp(−E_a,i/k_bT) were calculated as A_i= (k_bT/

h) exp(ΔS^‡_i/k_b).

Our newly obtained insights into the mechanism and active

site location for DDO on Ru/TiO₂allows us to revise the

prevailing notion that the support reducibility is the key

property for selective and hydrogen-eﬃcient HDO catalysts.

None of the intermediate states shown in the PES in Figure 11

The simpliﬁed kinetic analysis clearly shows that the direct

water-assisted pathway, DDO2, is the fastest reaction channel.

Its rate is ∼11 times faster than the reference rate for the

DDO1 pathway without water, even though the phenol

coverage decreases by one order of magnitude (Table 3).

Without overemphasizing the quantitative agreement and

assuming that the HYD rate is not altered by water, we note

that the ca. 11-fold increase is similar to the experimentally

obtained DDO/HYD ratio of 19 in the presence of water

(Table 1, entry 1), whereas the DDO/HYD ratio is

approximately unity in the absence of water (Table 1, entries

2 and 3). The less intuitive result of this analysis is that the

stabilizing eﬀect of water in pathways DDO3 and DDO4 is, in

fact, detrimental because it slows the C−O scission rate, even

though the phenol coverage is increased. The reason is that the

phenol reactant in the initial state is stabilized with respect to

the transition state, and although the surface coverage of phenol

increases, the rate constant, k⁺₂, becomes smaller.

3.6. Active Site Formation for DDO Catalysis: The

Speciﬁc Role of TiO₂. Titania has generated interest as a

HDO support primarily as an alternative to alumina because

alumina has known instability problems under typical HDO

conditions of high temperatures and humidity.^3d,6dIn addition,

titania is also a reducible support, which has been linked to ease

of oxygen vacancy formation,^9cand it has been noted that the

use of reducible supports enhances the catalytic hydrogenation

of carbonyls by group VIII metals.^8dThe calculations presented

in this work support this beneﬁcial role of an oxygen vacancy in

the energetics of C−O bond scission pathways in the absence

of water (DDO1) or with passive water stabilization of phenol

(DDO3 and DDO4). The HDO experiments of McEwen and

Wang also support the hypothesis that DDO selectivity is

enhanced by the combination of a reducible support (Fe₂O₃)

and transition metal (Pd).¹⁰A comparison of Pd/SiO₂, Pd/

Al₂O₃, and Pd/ZrO₂showed that only the reducible ZrO₂

support leads to selective DDO, because it has oxophilic

Figure 11. Full potential energy surface (PES) for the DDO2 pathway.

Binding energies (electronvolts) are calculated with respect to gas-

phase phenol, H₂, H₂O, and the h-Ru/TiO₂model with H₂O^bracid

site (model 1 in Figure 3). Activation energy barriers (E_a) in

electronvolts are shown in bold red font, and the roman numerals refer

to the elementary steps in Table 2. Reactants adsorbing to the surface

are shown in blue; products desorbing from the surface are shown in

green. For consistency with the modeled reaction steps, compensating

hydrogen atoms (1/2 H₂) have been added and removed.

6520

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

for the most favorable DDO2 pathway involves an oxygen

vacancy site. Instead of its reducibility, we propose that TiO₂is

unique among reducible metal oxide supports because of its

amphoteric character, that is, it can act as both an acid and a

base. In the full PES for the water-promoted DDO2

mechanism, shown in Figure 11, the two largest, and almost

identical, activation barriers are obtained for proton-assisted

dehydroxylation of phenol (E_a= 0.42 eV) and the regeneration

of the acid site by heterolytic H₂dissociation across the Ru/

TiO₂interface (E_a= 0.47 eV). These proton transfers to and

from the support are ideally balanced such that neither the acid-

catalyzed dehydroxylation nor the active site regeneration step

are highly activated. If TiO₂had a higher proton aﬃnity (lower

acidity), the dehydroxylation step would require a higher

activation energy; however, on the other hand, if TiO₂was

more acidic (lower proton aﬃnity), the surface protonation

during heterolytic H₂dissociation would be more diﬃcult.

This explanation is congruent with the fact that ZrO₂,

another amphoteric oxide with tunable Brønsted and Lewis

acidity/basicity,⁵²has also been identiﬁed as a suitable support

for transition metal catalysts for selective DDO.^3h,qThe Lewis

acid Zr⁴⁺cations may be transformed to Brønsted acid sites in

the presence of water and form an active site motif similar to

what we propose for Ru/TiO₂. Similarly, Pd/Fe₂O₃has shown

high DDO selectivity, which was attributed to the reducibility

of Fe₂O₃,^10abut the authors did not investigate the role of

water or support hydroxyl species. However, a combined STM/

DFT study has shown that protons on iron oxide surfaces are

highly mobile in the presence of water⁵³and could assist the

dehydration reaction, as proposed here for TiO₂. Finally, the

availability of mobile protons has also been implicated as a key

mechanistic component during water-assisted CO oxidation on

Au/TiO₂.²⁴

The novel mechanism we propose allows us to attribute the

exceptional activity and selectivity of Ru/TiO₂to the ability of

TiO₂to act as a conjugated acid/base pair rather than its

reducibility. Water molecules adsorbed on hydroxylated or

partially reduced TiO₂can easily accept and donate protons

across the Ru/TiO₂interface and lower the C−O scission

barrier by donating a proton during the abstraction of the

phenolic OH group. The OH group is replaced in a hydride

attack with a single reductive proton. Small Ru particles that

create a maximal number of interface sites are necessary for

optimal DDO performance. Because of the strong C_aryl−O

bond, the direct scission pathway has received very little

attention in the past. The DFT results presented here, however,

identiﬁed a bifunctional site with metal and Brønsted acid

functionality at the Ru/TiO₂interface that provides a lower C−

O bond scission barrier (E_a= 0.42 eV) than the most favorable

hydrogenation step (E_a= 0.81 eV). This proposed mechanism

fully explains the beneﬁcial eﬀect of water (maintaining the

Brønsted acid site) and small Ru particles (maximizing the ratio

of interface/bulk Ru active sites) on DDO selectivity and the

observation that a single reductive hydrogen is transferred to

the product. Although we have taken all necessary care in the

pathway and active site determination, we have concentrated

our theoretical eﬀorts on pathways that are consistent with our

catalytic and isotopic labeling experiments, and we acknowl-

edge the possibility that other more complex reaction pathways

or multisite reactions may exist.

Ring hydrogenation (HYD) occurs preferentially on large Ru

particles, as indicated by DFT simulations on the idealized

Ru(0001) surface. In the absence of the Ru/TiO₂interface the

lowest barrier to hydrogenation is E_a= 0.91 eV, whereas the

direct C−O scission is more diﬃcult (E_a= 1.23 eV). This is in

line with previous results obtained using calcined Ru/TiO₂

catalysts, which have aggregated Ru particle sizes of about 30

nm.^3aHigher amounts of cyclohexanone and cyclohexanol were

also seen in this work when octane was added to the reaction

mixture. This is consistent with the hypothesis that octane may

occlude and reduce the number of dehydroxylation sites at the

Ru/TiO₂interface.

4. CONCLUSIONS

This work reports that the DDO selectivity of Ru/TiO₂catalyst

with small Ru particles is enhanced by the presence of water. At

573 K and 550 psig H₂, these catalysts show unprecedented

activity (30% conversion) and selectivity for direct deoxygena-

tion of phenol (DDO/HYD ratio of 19). In the absence of

water or the presence of 10% octane, there is a 10−20 fold

reduction in the ratio of DDO/HYD products and a 30−60%

reduction in the overall conversion of phenol to products.

We propose a novel DDO mechanism for phenol, which is

consistent with the bulk of the experimental and theoretical

literature and explains the cocatalytic eﬀect of water. Isotopic

labeling experiments indicate that a single reductive hydrogen

atom is exchanged along the DDO path. This provides strong

evidence against mechanisms involving multiple

(de)hydrogenation/dehydration steps (Figure 2, mechanism

B) because the overall number of reductive proton transfers

would be expected to be greater than one. On the basis of

product distributions, Resasco and co-workers proposed a

reaction pathway initiated by a tautomerization step followed

by hydrogenation and dehydration (Figure 2, mechanism C).^3j,q

Isotopic labeling control experiments and low calculated

activation barriers for water-assisted tautomerization in this

work conﬁrm the presence of phenol tautomerization. The

hydrogenation/dehydration sequence along the tautomeriza-

tion pathway, however, would increase the number of reductive

H atoms transferred, which is not shown in our isotopic

labeling data.

In conclusion, our detailed understanding of the DDO

mechanism and novel interpretation of the role of the support

material will aid future eﬀorts to design more eﬃcient, less

expensive catalysts or catalytic processes for bio-oil upgrading

by hydrodeoxygenation. Our results suggest that by tuning the

support Brønsted acidity, an optimal balance for proton

acceptance and donation can be found, which is critical for

selective C−O cleavage. Furthermore, the process conditions

can be optimized to yield a desired DDO/HYD ratio by

adjusting the amount of water present or modifying the Ru

particle size.

ASSOCIATED CONTENT

■

S

* Supporting Information

The Supporting Information is available free of charge on the

ACS Publications website at DOI: 10.1021/acscatal.5b01554.

Additional experimental and computational details for

MS and DFT simulations, reaction energetics, illustra-

tions of alternative reaction pathways (PDF)

Relevant geometry ﬁles to facilitate viewing computa-

tional models (ZIP)

6521

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

ACS Catal. 2013, 3, 1774−1785. (i) Romero, Y.; Richard, F.; Brunet,

S. Appl. Catal., B 2010, 98, 213−233. (j) Nie, L.; de Souza, P. M.;

Noronha, F. B.; An, W.; Sooknoi, T.; Resasco, D. E. J. Mol. Catal. A:

Chem. 2014, 388−389, 47−55. (k) Gao, D.; Schweitzer, C.; Hwang, H.

T.; Varma, A. Ind. Eng. Chem. Res. 2014, 53, 18658−18667. (l) Resasco,

D. E.; Crossley, S. P. Catal. Today 2014, 257, 185. (m) Boonyasuwat,

S.; Omotoso, T.; Resasco, D. E.; Crossley, S. P. Catal. Lett. 2013, 143

(8), 783−791. (n) Sun, J.; Karim, A. M.; Zhang, H.; Kovarik, L.; Li, X.

S.; Hensley, A. J. R.; McEwen, J.-S.; Wang, Y. J. Catal. 2013, 306, 47−

57. (o) Song, W.; Liu, Y.; Barath, E.; Zhao, C.; Lercher, J. A. Green

Chem. 2015, 17 (2), 1204−1218. (p) Prasomsri, T.; Shetty, M.;

Murugappan, K.; Roman-Leshkov, Y. Energy Environ. Sci. 2014, 7 (8),

2660−2669. (q) de Souza, P. M.; Rabelo-Neto, R. C.; Borges, L. E. P.;

Jacobs, G.; Davis, B. H.; Sooknoi, T.; Resasco, D. E.; Noronha, F. B.

ACS Catal. 2015, 5 (2), 1318−1329. (r) Lu, J.; Heyden, A. J. J. Catal.

2015, 321, 39−50.

(4) (a) Echeandia, S.; Pawelec, B.; Barrio, V. L.; Arias, P. L.; Cambra,

J. F.; Loricera, C. V.; Fierro, J. L. G. Fuel 2014, 117, 1061−1073.

(b) Laurent, E.; Delmon, B. Ind. Eng. Chem. Res. 1993, 32, 2516−2524.

(5) Albahri, T. A.; Riasi, M. R.; Alqattan, A. A. Fuel Chemistry Division

Preprints 2002, 47, 710−711.

(6) (a) Yoon, Y.; Rousseau, R.; Weber, R. S.; Mei, D.; Lercher, J. A. J.

Am. Chem. Soc. 2014, 136, 10287−10298. (b) Bu, Q.; Lei, H.; Zacher,

A. H.; Wang, L.; Ren, S.; Liang, J.; Wei, Y.; Liu, Y.; Tang, J.; Zhang, Q.;

Ruan, R. Bioresour. Technol. 2012, 124, 470−477. (c) Parsell, T. H.;

Owen, B. C.; Klein, I.; Jarrell, T. M.; Marcum, C. L.; Haupert, L. J.;

Amundson, L. M.; Kenttamaa, H. I.; Ribeiro, F.; Miller, J. T.; Abu-

Omar, M. M. Chem. Sci. 2013, 4, 806−813. (d) Pucher, H.; Schwaiger,

N.; Feiner, R.; Pucher, P.; Ellmaier, L.; Siebenhofer, M. Int. J. Energy

Res. 2014, 38, 1964−1974.

(7) Moses, P. G.; Grabow, L. C.; Fernandez, E. M.; Hinnemann, B.;

Topsøe, H.; Knudsen, K. G.; Nørskov, J. K. Catal. Lett. 2014, 144,

1425−1432.

(8) (a) Zhao, F.; He, L. Prep. Pap.Am. Chem. Soc., Div. Fuel Chem.

2012, 57, 766. (b) Stoch, J.; Dao, H. Q.; Czeppe, T. Bull. Polym. Acad.

Sci. 1987, 35, 387−391. (c) He, L.; Wu, C.; Cheng, H.; Yu, Y.; Zhao,

F. Catal. Sci. Technol. 2012, 2, 1328−1331. (d) Mendes, M. J.; Santos,

O. A. A.; Jordao, E.; Silva, A. M. Appl. Catal., A 2001, 217, 253−262.

(e) Lugo-Jose, Y. K.; Monnier, J. R.; Williams, C. T. Appl. Catal., A

2014, 469, 410−418.

(9) (a) Wildschut, J.; Mahfud, F. H.; Venderbosch, R. H.; Heeres, H.

J. Ind. Eng. Chem. Res. 2009, 48, 10324−10334. (b) Wan, S.; Pham, T.

T.; Zhang, S.; Lobban, L. L.; Resasco, D. E.; Mallinson, R. AIChE J.

2013, 59, 2275−2285. (c) Omotoso, T.; Boonyasuwat, S.; Crossley, S.

P. Green Chem. 2014, 16, 645−655.

AUTHOR INFORMATION

■

Corresponding Authors

*E-mail: grabow@uh.edu.

*E-mail: raustin@barnard.edu.

Present Address

^⊥(R.N.A.) Department of Chemistry, Barnard College,

Columbia University, New York, NY 10027.

Notes

The authors declare no competing ﬁnancial interest.

ACKNOWLEDGMENTS

■

Thanks to Dr. Yong Zhang for assistance with the high-

resolution TEM. This work was supported by the U.S.

Department of Energy, Oﬃce of Science, Oﬃce of Basic

Energy Sciences under Award Nos. DE-SC0011983 and DOE

EPSCoR Grant DE-FG02-07ER46373. This research used

resources of the National Energy Research Scientiﬁc Comput-

ing (NERSC) Center, a DOE Oﬃce of Science User Facility

supported by the Oﬃce of Science of the U.S. Department of

Energy under Contract No. DE-AC02-05CH11231; and of the

Center for Nanoscale Materials (CNM), supported by the U.S.

DOE, Oﬃce of Science, Oﬃce of Basic Energy Sciences, under

Contract No. DE-AC02-06CH11357. Additional computational

resources were provided through the Extreme Science and

Engineering Discovery Environment (XSEDE), which is

supported by National Science Foundation (No.

ACI5681053575). We also acknowledge the use of the

Maxwell/Opuntia Cluster and the advanced support from the

Center of Advanced Computing and Data Systems (CACDS)

at the University of Houston to carry out the research

presented here. This report was prepared as an account of work

sponsored by an agency of the United States Government.

Neither the United States Government nor any agency thereof,

nor any of their employees, makes any warranty, express or

implied, or assumes any legal liability or responsibility for the

accuracy, completeness, or usefulness of any information,

apparatus, product, or process disclosed, or represents that its

use would not infringe privately owned rights. Reference herein

to any speciﬁc commercial product, process, or service by trade

name, trademark, manufacturer, or otherwise does not

necessarily constitute or imply its endorsement, recommenda-

tion, or favoring by the United States Government or any

agency thereof.

(10) (a) Hensley, A. J. R.; Hong, Y.; Zhang, R.; Zhang, H.; Sun, J.;

Wang, Y.; McEwen, J.-S. ACS Catal. 2014, 4 (10), 3381−3392.

(b) Hong, Y.; Zhang, H.; Sun, J.; Ayman, K. M.; Hensley, A. J. R.; Gu,

M.; Engelhard, M. H.; McEwen, J.-S.; Wang, Y. ACS Catal. 2014, 4

(10), 3335−3345.

REFERENCES

■

(11) Furimsky, E. Appl. Catal., A 2000, 199, 147.

(1) (a) Huber, G. W.; Iborra, S.; Corma, A. Chem. Rev. 2006, 106,

4044−4098. (b) Garcia-Perez, M.; Chaala, A.; Pakdel, H.; Kretschmer,

D.; Roy, C. Biomass Bioenergy 2007, 31, 222−242.

(2) Mukkamala, S.; Wheeler, M. C.; van Heiningen, A. R. P.; DeSisto,

W. J. Energy Fuels 2012, 26, 1380−1384.

(12) (a) Lu, J.; Heyden, A. J. Catal. 2015, 321, 39−50. (b) Chiu, C.-

C.; Genest, A.; Borgna, A.; Rosch, N. ACS Catal. 2014, 4, 4178−4188.

(13) Chiu, C.-C.; Genest, A.; Borgna, A.; Rosch, N. Phys. Chem.

̈

Chem. Phys. 2015, 17, 15324−15330.

(14) Hensley, A. J. R.; Wang, Y.; McEwen, J.-S. ACS Catal. 2015, 5,

523−536.

(15) Zhao, C.; Kou, Y.; Lemonidou, A. A.; Li, X.; Lercher, J. A.

Angew. Chem., Int. Ed. 2009, 48, 3987−3990.

(3) (a) Newman, C.; Zhou, X.; Goundie, B.; Ghampson, I. T.;

Pollock, R. A.; Ross, Z.; Wheeler, M. C.; Meulenberg, R.; Austin, R.

N.; Frederick, B. G. Appl. Catal., A 2014, 477, 64−74. (b) Badawi, M.;

Paul, J.-F.; Cristol, S.; Payen, E. Catal. Commun. 2011, 12, 901−905.

(c) Chang, J.; Danuthai, T.; Dewiyanti, S.; Wang, C.; Borgna, A.

ChemCatChem 2013, 5, 3041−3049. (d) Boullosa-Eiras, S.; Løpdeng,

R.; Bergem, H.; Stocker, M.; Hannevold, L.; Blekkan, E. A. Catal.

Today 2014, 223, 44−53. (e) Echeandia, S.; Arias, P. L.; Barrio, V. L.;

Pawelec, B.; Fierro, J. L. G. Appl. Catal., B 2010, 101, 1−12.

(f) Huynh, T. M.; Armbruster, U.; Pohl, M.-M.; Schneider, M.;

Radnik, J.; Hoang, D.-L.; Phan, B. M. Q.; Nguyen, D. A.; Martin, A.

ChemCatChem 2014, 6, 1940−1951. (g) Leiva, K.; Sepulveda, C.;

Garcia, R.; Fierro, J. L. G.; Escalona, N. Catal. Commun. 2014, 53, 33−

37. (h) Mortensen, P. M.; Grunwaldt, J.-D.; Jensen, P. A.; Jensen, A. D.

(16) Stein, S. E. Mass Spectra. NIST Chemistry WebBook, NIST

Standard Reference Database Number 69; National Institute of

Standards and Technology: Gaithersburg, MD, 20899, http://

webbook.nist.gov; accessed July 17, 2015.

(17) Mohler, F. L.; Dibeler, V. H.; Williamson, L.; Dean, H. Journal of

Research of the National Bureau of Standards 1952, 48, 188−192.

(18) (a) Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater.

Phys. 1993, 47, 558−561. (b) Kresse, G.; Hafner, J. Phys. Rev. B:

Condens. Matter Mater. Phys. 1994, 49, 14251−14269. (c) Kresse, G.;

Furthmuller, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54,

̈

6522

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

ACS Catalysis

Research Article

11169−11186. (d) Bahn, S. R.; Jacobsen, K. W. Comput. Sci. Eng.

2002, 4, 56−66.

(43) An, K.; Alayoglu, S.; Musselwhite, N.; Plamthottam, S.; Melaet,

G.; Lindeman, A. E.; Somorjai, G. A. J. Am. Chem. Soc. 2013, 135 (44),

16689−16696.

(19) Blochl, P. E. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 50,

̈

(44) Hensley, A. J. R.; Wang, Y.; McEwen, J.-S. Surf. Sci. 2014, 630,

244−253.

17953−17979.

(20) (a) Hammer, B.; Hansen, L. B.; Nørskov, J. K. Phys. Rev. B:

Condens. Matter Mater. Phys. 1999, 59, 7413−7421. (b) Perdew, J. P.;

Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865−3868.

(21) Stevens, E. D.; Rys, J.; Coppens, P. J. Am. Chem. Soc. 1978, 100,

2324−2328.

(45) Hammer, B.; Wendt, S.; Besenbacher, F. Top. Catal. 2010, 53,

423−430.

(46) Rekoske, J. E.; Barteau, M. A. J. Phys. Chem. B 1997, 101, 1113−

1124.

(47) Xi, Y.; Zhang, Q.; Cheng, H. J. Phys. Chem. C 2014, 118, 494−

501.

(48) (a) Zhu, X.; Lobban, L. L.; Mallinson, R. G.; Resasco, D. E. J.

Catal. 2011, 281, 21−29. (b) Zhu, X.; Nie, L.; Lobban, L. L.;

Mallinson, R. G.; Resasco, D. E. Energy Fuels 2014, 28, 4104−4111.

(49) Behtash, S.; Lu, J.; Faheem, M.; Heyden, A. Green Chem. 2014,

16, 605−616.

(50) Green, D. W.; Perry, R. H. Perry’s Chemical Engineers’ Handbook;

McGraw-Hill: New York, 2008.

(51) Conley, B. L.; Pennington-Boggio, M. K.; Boz, E.; Williams, T. J.

(22) Dudarev, S. L.; Botton, G. A.; Savrasov, S. Y.; Humphreys, C. J.;

Sutton, A. P. Phys. Rev. B: Condens. Matter Mater. Phys. 1998, 57,

1505−1509.

(23) Hu, Z.; Metiu, H. J. Phys. Chem. C 2011, 115, 5841−5845.

(24) Saavedra, J.; Doan, H. A.; Pursell, C. J.; Grabow, L. C.;

Chandler, B. D. Science 2014, 345, 1599−1602.

(25) Campbell, C. T. Acc. Chem. Res. 2013, 46, 1712−1719.

(26) (a) Morgan, B. J.; Watson, G. W. J. Phys. Chem. C 2009, 113,

7322−7328. (b) Chueh, Y. L.; Hsieh, C. H.; Chang, M. T.; Chou, L. J.;

Lao, C. S.; Song, J. H.; Gan, J. Y.; Wang, Z. L. Adv. Mater. 2007, 19,

143−149.

(27) Bengtsson, L. Phys. Rev. B: Condens. Matter Mater. Phys. 1999,

59, 12301−12304.

(28) Monkhorst, H. J.; Pack, J. D. Phys. Rev. B 1976, 13, 5188−5192.

(29) Sheppard, D.; Terrell, R.; Henkelman, G. J. Chem. Phys. 2008,

128, 13410601−13410610.

Chem. Rev. 2010, 110 (4), 2294−2312.

(52) Arce-Ramos, J. M.; Grabow, L. C.; Handy, B. E.; Cardenas-

Galindo, M.-G. J. Phys. Chem. C 2015, 119, 15150−15159.

(53) Merte, L. R.; Peng, G.; Bechstein, R.; Rieboldt, F.; Farberow, C.

A.; Grabow, L. C.; Kudernatsch, W.; Wendt, S.; Laegsgaard, E.;

Mavrikakis, M.; Besenbacher, F. Science 2012, 336, 889−893.

(30) (a) Farberow, C. A.; Dumesic, J. A.; Mavrikakis, M. ACS Catal.

2014, 4, 3307−3319. (b) Deng, J.; Li, H.; Xiao, J.; Tu, Y.; Deng, D.;

Yang, H.; Tian, H.; Li, J.; Ren, P.; Bao, X. Energy Environ. Sci. 2015, 8,

1594−1601. (c) Sun, X. Y.; Li, B.; Metiu, H. J. Phys. Chem. C 2013,

117, 7114−7122. (d) Wang, S.; Vorotnikov, V.; Vlachos, D. G. ACS

Catal. 2015, 5, 104−112.

(31) Hong, D.-Y.; Miller, S. J.; Agrawal, P. K.; Jones, C. W. Chem.

Commun. 2010, 46, 1038−1040.

(32) (a) Romero, Y.; Richard, F.; Reneme, Y.; Brunet, S. Appl. Catal.,

A 2009, 353, 46−53. (b) Elliott, D. C.; Hart, T. R. Energy Fuels 2009,

23, 631−637. (c) Gutierrez, A.; Kaila, R. K.; Honkela, M. L.; Slioor, R.;

Krause, A. O. I. Catal. Today 2009, 147, 239−246. (d) Bunch, A. Y.;

Wang, X.; Ozkan, U. S. Appl. Catal., A 2008, 346, 96−103. (e) Senol,

O. I.; Viljava, T.-R.; Krause, A. O. I. Catal. Today 2005, 100, 331−335.

(f) Bunch, A. Y.; Ozkan, U. S. J. Catal. 2002, 206, 177−187.

(g) Prochazkova, D.; Zamostny, P.; Bejblova, M.; Cerveney, L.; Cejka,

J. Appl. Catal., A 2007, 332, 56−64. (h) Wang, W.; Yang, Y. Q.; Luo,

H.; Hu, T.; Liu, W. Catal. Commun. 2011, 12, 436−440.

(33) Zhao, H. Y.; Li, D.; Bui, P.; Oyama, S. T. Appl. Catal., A 2011,

391, 305−310.

(34) Massoth, F. E.; Politzer, P.; Concha, M. C.; Murray, S.;

Jakowski, J.; Simons, J. J. Phys. Chem. B 2006, 110, 14283−14291.

(35) In the previous study, the catalysts were activated by heating in

an atmospheric pressure ﬂow of 3:1 N₂/H₂gas for several hours. In

this study, a sample of unactivated catalyst was placed in a high-

pressure Parr reactor and heated (575 K) in a static atmosphere (550

psig) of pure H₂gas for 2 h. Upon cooling, the activated catalyst was

weighed and immediately returned to the Parr reactor along with the

phenol necessary for the reaction.

(36) El-Sharkawy, E.-S. A. Monatsh. Chem. 2006, 137, 1487−1498.

(37) (a) Martin, D.; Duprez, D. J. Mol. Catal. A: Chem. 1997, 118,

113−128. (b) Datka, J.; Gil, B.; Vog, O.; Rakoczy, J. Stud. Surf. Sci.

Catal. 1999, 125, 409−415.

(38) Sajiki, H.; Ito, N.; Esaki, H.; Maesawa, T.; Maegawa, T.; Hirota,

K. Tetrahedron Lett. 2005, 46, 6995−6998.

(39) Ito, N.; Watahiki, T.; Maesawa, T.; Maegawa, T.; Sajiki, H. Adv.

Synth. Catal. 2006, 348, 1025−1028.

(40) Kawazoe, Y.; Ohnishi, M. Chem. Pharm. Bull. 1966, 14, 1413−

1418.

(41) Prins, R. Chem. Rev. 2012, 112, 2714−2738.

(42) Moreau, C.; Aubert, C.; Durand, R.; Zmimita, N.; Geneste, P.

Catal. Today 1988, 4, 117−131.

6523

DOI: 10.1021/acscatal.5b01554

ACS Catal. 2015, 5, 6509−6523

Article Doi

Experimental and Theoretical Insights into the Hydrogen-Efficient Direct Hydrodeoxygenation Mechanism of Phenol over Ru/TiO2

DOI: 10.1021/acscatal.5b01554

Source and publish data:

Authors:

Article abstract of DOI:10.1021/acscatal.5b01554

Full text of DOI:10.1021/acscatal.5b01554

Products guided by the article

R&D Labs maybe for 1076-43-3

Relevant to this article

Hot Product