decomposition of furan on pd(111)

10
ORIGINAL PAPER Decomposition of Furan on Pd(111) Ye Xu Published online: 24 April 2012 Ó Springer Science+Business Media, LLC 2012 Abstract Periodic density functional theory calculations (GGA-PBE) have been performed to investigate the mechanism for the decomposition of furan up to CO for- mation on the Pd(111) surface. At 1/9 ML coverage, furan adsorbs with its molecular plane parallel to the surface in several states with nearly identical adsorption energies of -1.0 eV. The decomposition of furan begins with the opening of the ring at the C–O position with an activation barrier of E a = 0.82 eV, which yields a C 4 H 4 O aldehyde species that rapidly loses the a H to form C 4 H 3 O (E a = 0.40 eV). C 4 H 3 O further dehydrogenates at the d position to form C 4 H 2 O(E a = 0.83 eV), before the ab C– C bond dissociates (E a = 1.08 eV) to form CO. Each step is the lowest-barrier dissociation step in the respective species. A simple kinetic analysis suggests that furan decomposition begins at 240–270 K and is mostly com- plete by 320 K, in close agreement with previous experi- ments. It is suggested that the C 4 H 2 O intermediate delays the decarbonylation step up to 350 K. Keywords Furan Palladium Decomposition Kinetics Hetoregenous catalysis Density functional theory 1 Introduction Furan compounds have gained significant attention in recent years in the effort to convert biomass, a renewable resource, into fuels and value-added chemicals. For instance, 5-hydroxymethylfurfural has been proposed as a versatile platform that can be converted into many industrial com- pounds including additional furan compounds and linear oxygenates such as levulinic acid [1]. It has also been shown that furan, furfural, and HMF can be converted to fuel compounds including butanol [2], methyl-tetrahydrofuran [3], and large alkanes [1, 4]. 2,5-dimethylfuran has been proposed as a suitable fuel that has a number of desirable characteristics compared to ethanol [5]. These furan com- pounds can be synthesized from sugars, cellulose, hemi- cellulose, and agricultural and forestry byproducts [58]. The conversions of furan compounds involve modifying or removing side groups, saturating the furan ring, and opening the furan ring for making linear compounds, all of which can be accomplished through metal-catalyzed hydrogenation and hydrogenolysis, among other catalytic processes. Knowing what catalyst and reaction conditions favor ring opening versus hydrogenation and vice versa will contribute to enhancing the versatility of the furan platform, and is also relevant to the removal of heteroatoms from heterocyclic compounds, an important step in the processing of heavy feed [9]. Finite pressure experiments have shown, for instance, that at 373 K furan hydrogenation on Pd/zir- conia produces little tetrahydrofuran (THF) at low H 2 :furan ratio but solely THF at high H 2 :furan ratio [10]. Furan hydrogenation on Pt(111) and (100) produces primarily tet- rahydrofuran and butanol, and the selectivity for one comes at the expense of the other [2]. The hydrogenation of furfural, a singly substituted furan compound, can produce different distributions of furfuryl alcohol, furan, and ring-opening products on SiO 2 -supported Cu, Pd, and Ni catalysts, and the product selectivity for hydrogenation decreases and that for decarbonylation and ring opening increases, with increasing temperature [11]. Y. Xu (&) Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, 1 Bethel Valley Road, P.O. Box 2008, MS-6493, Oak Ridge, TN 37831, USA e-mail: [email protected] 123 Top Catal (2012) 55:290–299 DOI 10.1007/s11244-012-9797-z

Upload: ye

Post on 26-Aug-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

ORIGINAL PAPER

Decomposition of Furan on Pd(111)

Ye Xu

Published online: 24 April 2012

� Springer Science+Business Media, LLC 2012

Abstract Periodic density functional theory calculations

(GGA-PBE) have been performed to investigate the

mechanism for the decomposition of furan up to CO for-

mation on the Pd(111) surface. At 1/9 ML coverage, furan

adsorbs with its molecular plane parallel to the surface in

several states with nearly identical adsorption energies of

-1.0 eV. The decomposition of furan begins with the

opening of the ring at the C–O position with an activation

barrier of Ea = 0.82 eV, which yields a C4H4O aldehyde

species that rapidly loses the a H to form C4H3O

(Ea = 0.40 eV). C4H3O further dehydrogenates at the dposition to form C4H2O (Ea = 0.83 eV), before the a–b C–

C bond dissociates (Ea = 1.08 eV) to form CO. Each step

is the lowest-barrier dissociation step in the respective

species. A simple kinetic analysis suggests that furan

decomposition begins at 240–270 K and is mostly com-

plete by 320 K, in close agreement with previous experi-

ments. It is suggested that the C4H2O intermediate delays

the decarbonylation step up to 350 K.

Keywords Furan � Palladium � Decomposition �Kinetics � Hetoregenous catalysis �Density functional theory

1 Introduction

Furan compounds have gained significant attention in recent

years in the effort to convert biomass, a renewable resource,

into fuels and value-added chemicals. For instance,

5-hydroxymethylfurfural has been proposed as a versatile

platform that can be converted into many industrial com-

pounds including additional furan compounds and linear

oxygenates such as levulinic acid [1]. It has also been shown

that furan, furfural, and HMF can be converted to fuel

compounds including butanol [2], methyl-tetrahydrofuran

[3], and large alkanes [1, 4]. 2,5-dimethylfuran has been

proposed as a suitable fuel that has a number of desirable

characteristics compared to ethanol [5]. These furan com-

pounds can be synthesized from sugars, cellulose, hemi-

cellulose, and agricultural and forestry byproducts [5–8].

The conversions of furan compounds involve modifying

or removing side groups, saturating the furan ring, and

opening the furan ring for making linear compounds, all

of which can be accomplished through metal-catalyzed

hydrogenation and hydrogenolysis, among other catalytic

processes. Knowing what catalyst and reaction conditions

favor ring opening versus hydrogenation and vice versa will

contribute to enhancing the versatility of the furan platform,

and is also relevant to the removal of heteroatoms from

heterocyclic compounds, an important step in the processing

of heavy feed [9]. Finite pressure experiments have shown,

for instance, that at 373 K furan hydrogenation on Pd/zir-

conia produces little tetrahydrofuran (THF) at low H2:furan

ratio but solely THF at high H2:furan ratio [10]. Furan

hydrogenation on Pt(111) and (100) produces primarily tet-

rahydrofuran and butanol, and the selectivity for one comes

at the expense of the other [2]. The hydrogenation of furfural,

a singly substituted furan compound, can produce different

distributions of furfuryl alcohol, furan, and ring-opening

products on SiO2-supported Cu, Pd, and Ni catalysts, and the

product selectivity for hydrogenation decreases and that for

decarbonylation and ring opening increases, with increasing

temperature [11].

Y. Xu (&)

Center for Nanophase Materials Sciences, Oak Ridge National

Laboratory, 1 Bethel Valley Road, P.O. Box 2008, MS-6493,

Oak Ridge, TN 37831, USA

e-mail: [email protected]

123

Top Catal (2012) 55:290–299

DOI 10.1007/s11244-012-9797-z

To better understand the surface chemistry of furan

compounds on Pd surfaces, I begin by conducting a

density functional theory (DFT) study of the decomposi-

tion of furan, as the simplest of furan compounds, on

Pd(111) in the low coverage limit. Several surface science

studies of furan on single-crystal Pd(111) surfaces have

been reported in the literature. At sub-saturation coverage

furan lies flat on the surface with the ring above fcc or

hcp threefold hollow site, as determined by scanning

tunneling microscopy (STM) [12], near-edge X-ray

absorption fine structure (NEXAFS), and photoelectron

diffraction (PhD) [13]. The flat geometry is corroborated

by DFT calculations [14]. The molecular desorption of

furan is observed for dosages above 0.25 L and starts at

200–250 K and peaks at up to 360 K [15]. At lower

dosage furan decomposition starts at 230–270 K and is

essentially complete by 320 K as determined by X-ray

photoelectron spectroscopy (XPS), high-resolution elec-

tron energy loss spectroscopy (HREELS), ultraviolet

photoelectron spectroscopy (UPS) [15], PhD [13], and

laser-induced thermal desorption (LITD) [16]. Tempera-

ture programmed desorption (TPD) shows desorption-

limited H2 and CO evolution from the surface starting at

*300 and 450 K respectively, indicating that H and CO

form at these temperatures or lower [15]. LITD identifies

concomitant increase in CO coverage with the disap-

pearance of furan starting at 270 K and the formation of

a small amount of benzene at 350 K [16]. Based on the

formation of CO and the appearance of benzene, the

carbon fragment has been concluded to be C3H3 [13, 15,

16] and studied in detail using NEXAFS, PhD as well as

DFT calculations [17, 18].

While the basic feature of furan decomposition seems

plain enough, subtle questions remain, including how the

process is initiated, and whether the a C–H bond scission

precedes or follows ring opening. As Caldwell et al. have

reasoned, it is not safe even to rule out C–C bond scission

as the first step, and they have pointed out the lack of

information on the activation barrier for the breaking of the

C–O bond in an aromatic ring to be an obstacle for better

understanding the decomposition mechanism of furan on

Pd(111) [19]. The loss of the a H is common in five-

membered heterocycles, including on reactive surfaces

such as Ru(0001), to form a-furyl species [20]. Yet, anal-

ysis of kinetic isotope effects suggests that a C–H bond

scission cannot be the sole rate-determining step [19]. In

this study all the C–H, C–O, and C–C bond scission steps

up to the formation of CO are considered. C–O bond

scission is determined to be the first step in the decompo-

sition process, followed by the loss of up to two H atoms

before decarbonylation occurs to yield CO. Kinetic anal-

ysis based on the DFT results predicts that the decompo-

sition of furan on Pd(111) begins at 240–270 K and is

largely completely 320 K, in close agreement with avail-

able experimental evidence.

2 Methods

Periodic DFT calculations have been performed using the

Vienna Ab initio Simulation Package (VASP) [21, 22] in

the generalized gradient approximation (GGA-PBE [23]).

The core electrons are described by the projector aug-

mented wave (PAW) method [24], and the Kohn–Sham

valence states are expanded in plane wave basis sets up to

400 eV. A first-order Methfessel-Paxton scheme [25] is

used to smear the electronic states with a temperature of

0.1 eV. All total energies are extrapolated back to 0 K. The

equilibrium lattice constant of bulk fcc Pd is calculated to

be 3.953 A, in close agreement with the experimental value

(3.89 A) [26].

The Pd(111) surface is modeled by a four-layer slab

with a (3 9 3) surface unit cell, giving a coverage of 1/9

monolayer (ML) per adsorbate. The top two layers of the

slab are relaxed. Neighboring slabs are separated in the z

direction by a 20 A thick vacuum region. The surface

Brillouin zone is sampled with a 3 9 391 Monkhorst–

Pack k-point grid. Increasing the k-point grid density

to 7 9 791 changes the total energies of surfaces with

adsorbed furan by less than 0.1 eV and so the lower grid

density is used. A (3 9 4) surface unit cell is used to

achieve a high coverage of 1/6 ML, with all the other

parameters being identical to those mentioned above.

The adsorption energy (DE) for an atom or molecule

adsorbed on the Pd surface is calculated as:

DE ¼ EA=Pd þ ZPEA=Pd

� �� EA þ ZPEAð Þ � EPd

where EA/Pd is the total energy of a Pd(111) slab with an

adsorbate; ZPEA/Pd is the zero point energy (ZPE) of the

adsorbate; EA and ZPEA are the total energy and zero-point

energy of the isolated neutral adsorbate in the gas phase,

respectively; and EPd is the energy of the clean Pd(111)

slab. A more negative adsorption energy therefore corre-

sponds to stronger adsorption. Geometry optimization for

adsorbates and the top two layers of the Pd(111) slab is

converged to 0.03 eV/A for each relaxed atomic degree of

freedom.

The transition states (TSs) of elementary steps are cal-

culated using the dimer method [27], with geometry opti-

mization converged to 0.01 eV/A for each relaxed atomic

degree of freedom. Each TS is verified to possess a single

imaginary vibrational frequency that corresponds to the

reaction coordinate of the elementary step and is verified to

converge back to the initial state (IS) when perturbed in

the backward direction. The activation barrier for an ele-

mentary step is defined to be the difference between the

Top Catal (2012) 55:290–299 291

123

energies of the TS and IS: Ea = (ETS ? ZPETS) - (EIS ?

ZPEIS). The reaction energy is defined to be the difference

between the energies of the product or final state (FS) and the

IS: DErxn = (EFS ? ZPEFS) - (EIS ? ZPEIS). If the IS or the

FS involves more than one adsorbate, its energy is calculated

with each adsorbate in their most stable configurations at 1/9

ML coverage.

The ZPE values are listed in the Appendix.

3 Results

3.1 Adsorption of Furan

Furan prefers to adsorb with its molecular plane lying

parallel to the Pd(111) surface so that either the a–b or the

b–b C–C bond is located on the top of a Pd atom, and the

center of the ring is located either over a threefold site or a

bridge site. Thus there are several such adsorption states

available for furan on Pd(111) (see Fig. 1 for illustrations

of some of the adsorption states; their associated adsorption

energies; and the site and C atom labeling). Those states

with the ring located above the fcc or hcp threefold sites

are energetically degenerate, with an adsorption energy of

*-1.0 eV at the coverage of 1/9 ML. The states with the

ring located over the bridge site have adsorption energies

that are *0.4 eV more positive but are also stable local

minima. Furan with its ring located on top of a Pd atom or

perpendicular to the surface is not stable at this coverage.

The TPD study of Ormerod et al. [15] has found that while

furan dosed at\0.25 L on Pd(111) can desorb only as CO

and H2, exposures C0.25 L permit it to desorb molecularly.

The maximum molecular desorption temperature observed

is 360 K at 0.25 L, which suggests that the desorption

barrier for furan is *0.9 eV or greater [28]. LITD exper-

iments have enabled Caldwell et al. to calculate the

desorption barrier to be 0.98 eV [19]. The GGA-PBE

adsorption energy of *-1.0 eV at 1/9 ML is therefore

numerically in line with the experimental findings. Bradley

et al. [14] reported the adsorption energies of furan to be

-1.1 eV in GGA-PW91 and -0.45 eV in GGA-RPBE.

The saturation surface coverage of furan has been esti-

mated at 15 % of a ML [19]. It can be seen in Fig. 1 that

furan, whether adsorbed in the more stable fcc/hcp states or

the less stable br states, occupies 6 Pd atoms. The maxi-

mum coverage in UHV conditions is therefore limited to

*1/6 ML without physical overlap or too small distances

between neighboring furan molecules. I have calculated the

differential adsorption energy of furan in the hcp/bbPd

state as the coverage increases from 1/9 ML to 1/6 ML to

be -0.11 eV, which suggests that the saturation coverage

is indeed around 1/6 ML.

In the gas phase, the two a–b C–C bonds in furan are

1.36 A, the b–b C–C bond is 1.43 A, and the two C–O bonds

are 1.37 A. In comparison to gas-phase ethanol, acetalde-

hyde, benzene, and ethylene molecules (Table 1), it can be

seen that C–C bonds in furan are intermediate between

single and double bonds, and the C–O bonds are

αα α

β β

f

h br

hcp/ββPd-1.00 eV

fcc/ββPd-0.99 eV

br/ββPd-0.61 eV

hcp/αβPd-0.98 eV fcc/αβPd

-0.98 eV

α1β2

β3α4 t

α α

β β

f

h br

hcp/ββPd-1.00 eV

fcc/ββPd-0.99 eV

br/ββPd-0.61 eV

hcp/αβPd-0.98 eV fcc/αβPd

-0.98 eV

α1β2

β3α4 t

Fig. 1 Several stable high-symmetry states for furan adsorbed on

Pd(111). The adsorption energy for each state is as labeled. The

labeling of the high-symmetry surface sites on Pd(111) (t top, brbridge, h hcp, f fcc) and the different C atoms in furan (a1, b2, b3, a4)

is indicated. Small black, red, white; and large white spheresrepresent C, O, H, and Pd atoms, respectively. Molecules are placed

close to each other for illustration purposes

Table 1 Comparison of C–C and C–O bond lengths (in A) in several

gas-phase molecules with furan in the gas phase and on Pd(111)

C–C C–O C–Pd

Ethanol 1.51 1.43 –

Acetaldehyde 1.50 1.22 –

Benzene 1.40 – –

Ethylene 1.33 – –

Furan (this work)

Gas-phase 1.36 (a–b) 1.37 –

1.43 (b–b)

hcp/bbPd 1.46 (a–b) 1.41 2.11 (a-Pd)

1.43 (b–b) 2.22 (b-Pd)

hcp/abPd 1.42, 1.47 (a–b) 1.39, 1.43 2.18, 2.10 (a-Pd)

1.46 (b–b) 2.27, 2.14 (b-Pd)

Furan (Bradley et al. from Ref. [14])

Gas-phase 1.35 (a–b) 1.36 –

1.42 (b–b)

Hollow-h 1.44 (a–b) 1.40 2.11 (a-Pd)

1.41 (b–b) 2.23 (b-Pd)

Off-hollow-h 1.40, 1.46 (a–b) 1.37, 1.41 2.21, 2.11 (a-Pd)

1.44 (b–b) 2.25, 2.13 (b-Pd)

292 Top Catal (2012) 55:290–299

123

intermediate between C–O and C=O, consistent with the

aromaticity of the molecule. On Pd(111), the C–O and a–bC–C bonds are both lengthened compared to the gas-phase

furan molecule, which suggests that the interaction with the

surface reduces the net bonding character of the p orbitals.

The calculated bond lengths of the adsorbed furan agree with

those calculated by Bradley et al. [14] in GGA-RPBE to

within 0.02 A (see Table 1 for comparison). The molecular

plane of furan is located at *2.1 A above the relaxed top

layer of Pd atoms. For comparison, the adsorption of thi-

opehene, the sulfur analog of furan, on MoS2 is completely

dominated by van der Waals (vdW) contributions at *-0.4 eV with the adsorption height being *3.5 A [29]. The

adsorption height of furan on Pd(111) as determined

by GGA functionals is much smaller than the likely vdW

equilibrium distance for furan, which suggests that vdW

contributions are not significant in furan adsorption on

Pd(111).

3.2 Dissociation of Furan

To investigate the decomposition of furan, I begin by

calculating C–H, C–O, and C–C bond scission in furan.

This is done for two adsorbed states, hcp/bbPd and hcp/

abPd. The hcp/bbPd state has five unique such steps: (1)

C–H bond scission at the a position; (2) C–H bond scission

at the b position; (3) C–O bond scission; (4) C–C bond

scission at the a–b position; (5) C–C bond scission at the

b–b position. The structure of the TS of each of the five

steps and the associated activation energies (Ea) relative to

the hcp/bbPd state are shown in Fig. 2.

As can be seen in Fig. 2, the TSs for the scission of the

C–H bonds always involve the dissociating H atom, the C

atom, and an underlying Pd atom forming a three-centered

bonding structure, a characteristic common of C–H bond

dissociation steps. The C–O and C–C bonds dissociate over

a Pd atom with the C atoms located in either a near-top or

bridge position at the TSs. The Ea for dissociating the

bC–H bond is 1.16 eV, whereas the Ea for dissociating the

aC–H bond is higher at 1.53 eV because the a H is closer to

the electron-withdrawing O atom. The C–O bond is the

weakest bond in furan, with Ea = 0.82 eV for its dissoci-

ation. At the TS, the dissociating C–O bond lengthens to

1.93 A over a Pd atom, and the C atom is stabilized in part

by two Pd atoms, while the remaining C–O bond contracts

to 1.33 A. The two different C–C bonds are also quite

difficult to break: The Ea for dissociating the a–b and the

b–b C–C bonds are calculated to be 1.62 and 1.49 eV,

respectively. Again I attribute the difference to the prox-

imity of the a–b C–C bonds to the O atom.

The hcp/abPd state has lower symmetry than the hcp/

bbPd state, so all the bonds are nonequivalent. The struc-

ture and associated Ea for the TS of each of the steps for

the hcp/bbPd state are shown in Fig. 3. Despite the non-

Fig. 2 Structure of the TSs of

the bond dissociation steps in

furan in the hcp/bbPd state on

Pd(111): C–H bond scission at

the (a) a and (b) b positions;

(c) C–O bond scission; C–C

bond scission at the (d) a–b and

(e) b–b positions. In each panel,

top view is shown on the left and

a bonding scheme with bond

lengths indicated is shown on

the right (length of a C–H bond

is labeled only if it is

dissociating). The value

corresponding to the

dissociating bond is

underscored. The activation

energy (Ea) associated with

each TS is also indicated. Smallblack, red, white; and largewhite spheres represent C, O, H,

and Pd atoms, respectively

Top Catal (2012) 55:290–299 293

123

equivalence, it remains more difficult to dissociate the aC–H

bonds than the bC–H bonds, and more difficult to dissociate

the a–b C–C bonds than the b–b C–C bond. This suggests

that the relative strengths of the bonds in furan on Pd(111) is

mainly influenced by the atomic arrangement in the mole-

cule. C–O bond scission has Ea = 0.74 eV relative to the

hcp/abPd state, but in terms of the ZPE-corrected total

energy, this TS is only 0.06 eV more stable than the corre-

sponding TS for the hcp/bbPd state (Fig. 2c).

3.3 Dissociation of Ring-Opening Products

Since the opening of the furan ring preferentially occurs at

the C–O position with nearly identical TS energies for the

two different initial states, I use the higher-symmetry hcp/

bbPd state in the subsequent narrative for brevity. The

opening of the furan ring at the C–O position produces

what may be considered a dehydrogenated C4 aldehyde

species, C4H4O, the optimized structure for which is shown

in Fig. 4a. The dissociation of this linear C4H4O interme-

diate via four C–H bond scission steps at the a–d positions

and three C–C bond scission steps at the a–b, b–c, c–dpositions are calculated (see Fig. 4a for the labeling of the

C atoms). The structure and associated Ea for the TS of

each of these steps are depicted in Figs. 4b, h. I find that

C4H4O most readily undergoes C–H bond scission at the aposition, with Ea = 0.40 eV. C–H bond scission at the dand c positions is also facile, with Ea = 0.85 and 0.55 eV.

The dissociation of all three C–C bonds is difficult, with

each Ea exceeding 1 eV. The dissociation of the carbonyl

C=O bond is presumed to have a much higher barrier and

ignored in the subsequent calculations. This is in accord

with the fact that no oxygen-containing species is observed

to desorb in the TPD experiment besides CO [15].

Next I investigate the decomposition of the C4H3O

species that results from the loss of the a H in C4H4O

(Fig. 5a), and the C4H2O species that results from the C–H

bond dissociation at the d position in C4H3O (Fig. 6a).

Fig. 3 Structure of the TSs of

the bond dissociation steps in

furan in the hcp/abPd state on

Pd(111): C–H bond scission at

the (a) a1; (b) b2; (c) b3, and

(d) a4 positions; (e) C–O bond

scission at a4 position; C–C

bond scission at the (f) a1–b2;

(g) b2–b3; (h) b3–a4 positions.

In each panel, top view is shown

on the left and a bonding

scheme with bond lengths

indicated is shown on the right(length of a C–H bond is labeled

only if it is dissociating). The

value corresponding to the

dissociating bond is

underscored. The activation

energy (Ea) associated with

each TS is also indicated. Smallblack, red, white; and largewhite spheres represent C, O, H,

and Pd atoms, respectively

294 Top Catal (2012) 55:290–299

123

Three C–H bond scission steps, at the b, c, and d positions,

and three C–C bond scission steps at the a–b, b–c, c–dpositions are calculated for C4H3O (Figs. 5b, g). The most

facile dissociation step is C–H bond scission, occurring at

the d position, with Ea = 0.83 eV. C–C bond scission

at the a–-b position follows closely behind, with

Ea = 0.93 eV. For C4H2O, two C–H bond scission steps, at

the b and c positions, and three C–C bond scission steps at

the a–b, b–c, c–d positions are calculated (Figs. 6b, f).

Here C–C bond scission at the a–b position has the lowest

activation barrier (Ea = 1.08 eV), yielding CO.

3.4 Adsorption of Dissociated Products

Dissociated products including H, CO, CHO, C3H2, C3H3,

ring-opened benzene, and benzene have also been

calculated (Fig. 7). H, CO, and benzene are calculated in

previously reported preferred sites: fcc for H and CO

[30–32], and bridge prpr for benzene [33]. The adsorption

energies are calculated to be -2.81, -2.13, and -1.28 eV,

in close agreement with the previous DFT results. It should

be noted that the GGA-PBE predicted adsorption of CO is

known to be too strong [30]. CHO prefers to bind through

the C atom in a bridge site with the carbonyl oxygen located

in a near-top position. The C–O bond length is 1.27 A.

C3H2 prefers to be located with its C end in a threefold site

and the CH end bonded to the top of an adjacent Pd atom. Its

CH–CH bond is 1.36 A and CH–C bond is 1.45 A. C3H3

adsorbs with both ends in a near-top position and the

molecular plane tilted toward an adjacent Pd atom. Both of

the C–C bonds in C3H3 are 1.44 A, which therefore still has

a conjugated character.

(a) (b) (c) (d)

(e) (f) (g) (h)

Ea = 0.40 eV Ea = 1.32 eV Ea = 0.85 eV

Ea = 0.55 eV Ea = 1.03 eV Ea = 1.48 eV Ea = 1.32 eV

αα

βγ

δ

Ea = 0.40 eV Ea = 1.32 eV Ea = 0.85 eV

Ea = 0.55 eV Ea = 1.03 eV Ea = 1.48 eV Ea = 1.32 eV

α

βγ

δ

Fig. 4 Top views of a the

C4H4O intermediate on Pd(111)

and b–h the TSs of the bond

dissociation steps: C–H bond

scission at the b a; c b; d c; and

e d positions; and C–C bond

scission at the f a–b; g b–c; and

h c–d positions. The activation

energy (Ea) associated with

each TS is indicated. Smallblack, red, white; and largewhite spheres represent C, O, H,

and Pd atoms, respectively. The

labeling of the four C atoms is

shown in (a)

(a) (b) (c) (d)

(e) (f) (g)

Ea = 1.40 eV Ea = 1.16 eV Ea = 0.83 eV

Ea = 0.93 eV Ea = 1.17 eV Ea = 1.68 eV

αα

βγ

δ

Ea = 1.40 eV Ea = 1.16 eV Ea = 0.83 eV

Ea = 0.93 eV Ea = 1.17 eV Ea = 1.68 eV

α

βγ

δ

Fig. 5 Top views of a the

C4H3O intermediate on Pd(111)

and b–g the TSs of the bond

dissociation steps: C–H bond

scission at the b b; c c; and d dpositions; and C–C bond

scission at the e a–b; f b-c; and

g c–d positions. The activation

energy (Ea) associated with

each TS is indicated. Smallblack, red, white; and largewhite spheres represent C, O, H,

and Pd atoms, respectively. The

labeling of the four C atoms is

shown in (a)

Top Catal (2012) 55:290–299 295

123

4 Discussion

Based on the TSs that have been located for the dissocia-

tion of the various bonds in furan and its dissociated

derivatives, a reaction energy profile for the decomposition

of furan leading up CO formation is constructed (Fig. 8).

The highest-energy TS on the reaction energy landscape is

that of C–O bond scission in furan and is located at 0.18 eV

below gas-phase furan, indicating that furan adsorbed on

Pd(111) prefers decomposition over desorption, which

agrees with the TPD results of Ormerod et al. [15]. No a-

furyl species is formed in contrast to Ru(0001) [20]. The

resulting C4H4O species undergoes rapid C–H bond dis-

sociation to produce C4H3O, whereas C–C bond scission in

C4H4O yielding CHO has a much higher activation barrier

(Path C in Fig. 8). C4H3O can dehydrogenate to produce

C4H2O with an activation barrier that is nearly identical to

the initial C–O bond scission (Path A in Fig. 8), or it can

undergo a–b C–C scission by overcoming an activation

barrier that is 0.1 eV higher to directly produce CO and

C3H3 surface species (Path B in Fig. 8). The ratio of

the dehydrogenation rate versus the decarbonylation rate,

calculated as

rA

rB¼ mA

mBe�EA

a �EAa

jT ;

with Ea= 0.83 eV; Ea= 0.93 eV (see Fig. 5); mA = 1.6 9

1014 and mB = 8.2 9 1013 s-1 (see Appendix), indicates that

only 1–2 % of C4H3O undergoes decarbonylation to yield CO

and C3H3 between 300 and 350 K. The majority of C4H3O

dehydrogenates further to C4H2O particularly since the rec-

ombinative desorption of H2 commences around 300 K to

eliminate H from the Pd(111) surface [34]. The loss of up to two

H atoms occurs after ring opening, which accounts for the

Fig. 6 Top views of a the C4H2O intermediate on Pd(111) and

b–f the TSs of the bond dissociation steps: C–H bond scission at the

b b; and c c positions; and C–C bond scission at the d a–b; e b–c; and

f c–d positions. The activation energy (Ea) associated with each TS is

indicated. Small black, red, white; and large white spheres represent

C, O, H, and Pd atoms, respectively. The labeling of the four C atoms

is shown in (a)

Fig. 7 Top views of a H; b CO; c CHO; d C3H2; e C3H3; f ring-

opened benzene; and g benzene in their respective preferred

adsorption states; h TS of 2C3H3 ? ring-opened benzene; i TS of

ring-opened benzene ? benzene. The activation energy (Ea) associ-

ated with each TS is indicated. Small black, red, white; and largewhite spheres represent C, O, H, and Pd atoms, respectively

Fig. 8 Reaction energy profile for furan decomposition on Pd(111).

All species are surface-adsorbed except for gas-phase furan (‘‘(g)’’).

The balance of H atoms is not shown for brevity. Points marked by

circles are TSs. The difference in energy between a pair of points

separated by a TS is the DErxn for that step as defined in ‘‘Methods’’.

The three alternate reaction paths (A, B, and C) are referenced to in

the text

296 Top Catal (2012) 55:290–299

123

desorption-limited H2 evolution observed by Ormerod et al.

[15], before decarbonylation. Therefore I conclude that on a

clean Pd(111) surface, the majority of furan undergoes ring

opening, dehydrogenation, and decarbonylation in that order.

Since C–C bond scission at the b–c and c–d positions consis-

tently has higher activation barriers than at the a–b position

(decarbonylation), CO2 is far less likely to form than CO on

Pd(111) [15, 35].

Previous experimental studies have identified the tem-

perature at which furan decomposition commences on

Pd(111) to be between 230 and 270 K as evidence by a

decrease of the furan vibrational signatures in HREELS

[15]; changes in the C signal in XPS [15] and PhD [13];

and a decrease in the furan desorption signal in LITD [16].

I perform a Redhead-like analysis to estimate the temper-

ature in TPD at which furan begins to dissociate on

Pd(111) using the DFT results. It is assumed that the ring-

opening step is a uni-molecular, first-order surface process,

and that the dissociation is not limited by the availability of

free sites, assumptions that are reasonable at the surface

coverage of 1/9 ML. As such, the classic Redhead equation

[28] applies:

r ¼ � dhdt¼ vhne�

EakT with T ¼ T0 þ bt:

For furan in the hcp/bbPd state undergoing C–O bond

dissociation, Ea is 0.82 eV (Fig. 2c); m = 5.6 9 1013 s-1

(see ‘‘Appendix’’); initial coverage h0 = 0.11; n = 1; and

the heating rate b is set to 10 K/s. The calculated coverage is

plotted as a function of temperature in Fig. 9, which shows

that the coverage of furan on Pd(111) begins to decrease

appreciably at 270 K and mostly reaches zero by 320 K. I

note briefly that a co-adsorbed H atom located in the 2nd

nearest-neighbor hcp site increases the Ea of the ring-

opening step slightly from 0.82 to 0.84 eV. For furan in the

hcp/abPd state, dissociation begins appreciably at 240 K.

The reaction pathway that I have constructed identifies

C4H2O as a stable surface intermediate prior to CO

formation. It is worth noting that in the HREELS experi-

ment by Ormerod et al. [15], a peak is initially seen at

1,750 cm-1 at 300 K and below, and it blue-shifts to

1,800 cm-1 and increases in intensity at 400 K. The

authors have interpreted this blue shift and increase in

intensity as due to increasing coverage of CO. Vibrational

spectroscopies of CO adsorbed on Pd(111) generally locate

the m(C=O) mode at 1,820–1,850 cm-1 for coverage up to

1/3 ML [36, 37]. On the other hand, C4H2O has a m(C=O)

mode calculated to be 1,700 cm-1 with a dipole compo-

nent perpendicular to the surface, and the m(C=O) fre-

quency of CO on Pd(111) is calculated to be 1,776 cm-1 at

1/9 ML. Therefore the observed C=O signature and its blue

shift may instead be due to the C4H2O intermediate and its

decarbonylation to yield CO.

The same kind of kinetic analysis as performed above

indicates that the decarbonylation of C4H2O becomes signif-

icant at 350 K and is mostly complete by 410 K (Fig. 9). This

does not directly agree with the LITD results of Caldwell et al.

[16], which show a simultaneous rise of the CO signal with the

decrease in the furan signal beginning at 270 K. I note that

after furan undergoes ring opening and subsequent dehydro-

genation, the surface C4 intermediate(s) has no ready desorp-

tion channel except undergoing further decomposition to

release CO as the first desorbing species. Laser-induced dis-

sociation of surface species is uncommon but can occur in

LITD experiments [38]. This may explain why CO formation

and furan disappearance appear to be correlated between 270

and 310 K and share similar kinetics in LITD experiments

[19]. I propose that the existing experimental evidence can

accommodate a ring-opened surface intermediate between 300

and 350 K, but whether this is true requires further experi-

mental investigation (e.g. using variable-temperature STM)

and kinetic analysis.

The exact fate of the C3 fragment is outside the scope of

this study. In the TPD experiments furan, CO, and H2 are

the only gaseous products observed [15]. No CxHy surface

species has been unambiguously characterized [18]. It

stands to reason that as furan cracks in the absence of

sufficient hydrogen in TPD, carbon residuals are left on the

surface that eventually become coke. In the LITD experi-

ments benzene is detected as an additional gaseous species

because this technique favors high-entropy pathways such

as desorption [16, 19]. I have calculated benzene on

Pd(111) [33] and its formation from two C3H3 fragments

(see Fig. 7f–i). It is a two-step process, with Ea = 1.21 and

0.29 eV for the 1st and 2nd C–C coupling steps, respec-

tively, and a large overall driving force of DErxn =

-1.82 eV (Fig. 10). This is not a major pathway even in

LITD experiments (2 % of furan is estimated to end up as

benzene [16]). This can be rationalized by our finding that

most of the C3 fragments that are produced in the

decarbonylation step are already hydrogen-poor (C3H2).

Fig. 9 Calculated coverages of furan (left solid curve) and the C4H2O

intermediate (right dashed curve) on Pd(111) both as functions of

temperature. See text for more detail

Top Catal (2012) 55:290–299 297

123

5 Conclusions

Periodic density functional theory calculations (GGA-PBE)

have been performed to investigate the mechanism for the

decomposition of furan up to CO formation on Pd(111). At

1/9 ML, furan adsorbs with its molecular plane parallel to

the surface in several configurations with nearly identical

adsorption energies of -1.0 eV, in close agreement with

experimental estimates. The TSs for the dissociation of all

the bonds in furan and several ring-opened C4 intermedi-

ates have been determined. The decomposition of furan in

the hcp/bbPd states preferentially begins with the opening

of the ring at the C–O bond (with an activation barrier

Ea = 0.82 eV). The Ea for the dissociation of the b H is

higher at 1.16 eV, while those for dissociating the a H and

the a–b and b-b C–C bonds are all *1.5 eV or higher. The

activation barriers are nearly identical for furan adsorbed

in the alternate, energetically degenerate hcp/abPd state,

suggesting that the atomic arrangement in furan largely

determines the relative strengths of its bonds.

The C4H4O intermediate that is produced from the ring-

opening of furan is calculated to preferentially and rapidly

lose the a H to produce C4H3O, with Ea = 0.40 eV. The

C4H3O intermediate preferentially undergoes further

dehydrogenation at the d position to yield C4H2O

(Ea = 0.83 eV), or it can decarbonylate to yield CO and

C3H3 (Ea = 0.93 eV). C3H3 and subsequent benzene

formation should only account for 1–2 % of furan. The

dehydrogenation channel is likely to prevail in a TPD

experiment particularly because H2 desorption from

Pd(111) commences at 300 K. The C4H2O intermediate

then preferentially decarbonylates to produce CO

(Ea = 1.08 eV).

Simple kinetic analysis indicates that the initial

decomposition of the furan ring should begin between 240

and 270 K, in close agreement with experimental evidence,

and that the C4H2O intermediate should persist on the

surface until the temperature is ramped up to 350 K. I

suggest that existing experimental evidence can accom-

modate such a stable intermediate prior to CO formation,

although further confirmation will be needed. The finding

that after ring opening, the loss of up to two H atoms

occurs before the fracture of the molecular backbone sug-

gests that factors that favor ring opening but disfavor

dehydrogenation should make the formation of linear C4

oxygenates from furan possible on metallic Pd.

Acknowledgments This work was performed at the Center for

Nanophase Materials Sciences, which is sponsored at Oak Ridge

National Laboratory by the Division of Scientific User Facilities, U.S.

Department of Energy (DOE), and used resources of the National

Energy Research Scientific Computing Center, which is supported by

DOE Office of Science under Contract DE-AC02-05CH11231, and

resources of the Oak Ridge Leadership Computing Facility, which is

supported by DOE Office of Science under Contract DE-AC05-

00OR22725. I thank Dr. Aditya Ashi Savara for helpful discussions.

Appendix

See Table 2.

Fig. 10 Reaction energy profile for benzene formation from two

C3H3 fragments on Pd(111). All species are surface-adsorbed. Points

marked by circles are TSs. The difference in energy between a pair of

points separated by a TS is the DErxn for that step as defined in

‘‘Methods’’

Table 2 Zero point energy (ZPE, in eV) for the adsorbed species,

and ZPE corrections (equal to the ZPE of IS minus the ZPE of TS)

and prefactors (in s-1) for the dissociation steps, as mentioned in the

figures in the main text

ZPE m

Furan(g) 1.85

CO(g) 0.13

Benzene(g) 2.67

Fig. 2

Furan

R/hcp/bbPd 1.84

a 0.24 1.5 9 1014

b 0.16 3.8 9 1014

c 0.09 5.6 9 1013

d 0.08 1.4 9 1013

e 0.12 2.3 9 1013

Fig. 3

Furan

R/hcp/abPd 1.84

a 0.24 2.6 9 1014

b 0.18 6.8 9 1013

c 0.16 5.0 9 1014

d 0.24 4.1 9 1014

e 0.08 1.1 9 1014

f 0.08 1.6 9 1013

g 0.11 2.7 9 1013

h 0.08 1.5 9 1013

Fig. 4

298 Top Catal (2012) 55:290–299

123

References

1. Chheda JN, Huber GW, Dumesic JA (2007) Angew Chem Int Ed

46:7164

2. Kliewer CJ, Aliaga C, Bieri M, Huang WY, Tsung CK, Wood JB,

Komvopoulos K, Somorjai GA (2010) J Am Chem Soc

132:13088

3. Huber GW, Iborra S, Corma A (2006) Chem Rev 106:4044

4. Huber GW, Dumesic JA (2006) Catal Today 111:119

5. Roman-Leshkov Y, Barrett CJ, Liu ZY, Dumesic JA (2007)

Nature 447:982

6. Roman-Leshkov Y, Chheda JN, Dumesic JA (2006) Science

312:1933

7. Carlson TR, Vispute TP, Huber GW (2008) ChemSusChem 1:397

8. Yu S, Brown HM, Huang XW, Zhou XD, Amonette JE, Zhang

ZC (2009) App Catal A Gen 361:117

9. Furimsky E (2000) App Catal A Gen 199:147

10. Jackson SD, Canning AS, Vass EM, Watson SR (2003) Ind Eng

Chem Res 42:5489

11. Sitthisa S, Resasco DE (2011) Catal Lett 141:784

12. Loui A, Chiang S (2004) Appl Surf Sci 237:555

13. Knight MJ, Allegretti F, Kroger EA, Polcik M, Lamont CLA,

Woodruff DP (2008) Surf Sci 602:2524

14. Bradley MK, Robinson J, Woodruff DP (2010) Surf Sci 604:920

15. Ormerod RM, Baddeley CJ, Hardacre C, Lambert RM (1996)

Surf Sci 360:1

16. Caldwell TE, Abdelrehim IM, Land DP (1996) J Am Chem Soc

118:907

17. Knight MJ, Allegretti F, Kroger EA, Polcik M, Lamont CLA,

Woodruff DP (2008) Surf Sci 602:2743

18. Bradley MK, Duncan DA, Robinson J, Woodruff DP (2011) Phys

Chem Chem Phys 13:7975

19. Caldwell TE, Land DP (1999) J Phys Chem B 103:7869

20. Qiao MH, Yan FQ, Sim WS, Deng JF, Xu GQ (2000) Surf Sci

460:67

21. Kresse G, Furthmuller J (1996) Comput Mater Sci 6:15

22. Kresse G, Hafner J (1994) Phys Rev B 49:14251

23. Perdew JJ, Burke K, Ernzerhof M (1996) Phys Rev Lett 77:3865

24. Kresse G, Joubert D (1999) Phys Rev B 59:1758

25. Methfessel M, Paxton AT (1989) Phys Rev B 40:3616

26. Ashcroft NW, Mermin ND (1976) Solid state physics. Saunders

College, Orlando

27. Henkelman G, Jonsson H (1999) J Chem Phys 111:7010

28. Masel RI (1996) Principles of adsorption and reaction on solid

surfaces. Wiley, New York

29. Moses PG, Mortensen JJ, Lundqvist BI, Nørskov JK (2009) J

Chem Phys 130:104709

30. Gajdos M, Eichler A, Hafner J (2004) J Phys Condens Matter

16:1141

31. Lim KH, Chen ZX, Neyman KM, Rosch N (2006) J Phys Chem B

110:14890

32. Xu L, Xu Y (2011) Catal Today 165:96

33. Morin C, Simon D, Sautet P (2006) Surf Sci 600:1339

34. Conrad H, Ertl G, Latta EE (1974) Surf Sci 41:435

35. Medlin J, Horiuchi C, Rangan M (2010) Top Catal 53:1179

36. Bradshaw AM, Hoffmann FM (1978) Surf Sci 72:513

37. Surnev S, Sock M, Ramsey MG, Netzer FP, Wiklund M, Borg M,

Andersen JN (2000) Surf Sci 470:171

38. Hall RB (1987) J Phys Chem 91:1007

Table 2 continued

ZPE m

C4H4O

a 1.76

b 0.16 1.2 9 1014

c 0.20 6.6 9 1014

d 0.16 1.4 9 1014

e 0.12 4.2 9 1013

f 0.09 7.0 9 1013

g 0.09 3.2 9 1013

h 0.09 3.5 9 1013

Fig. 5

C4H3O

a 1.47

b 0.22 2.8 9 1014

c 0.24 2.6 9 1014

d 0.15 1.6 9 1014

e 0.07 8.2 9 1013

f 0.09 3.1 9 1013

g 0.08 3.8 9 1013

Fig. 6

C4H2O

a 1.19

b 0.19 7.5 9 1013

c 0.23 1.3 9 1014

d 0.09 6.4 9 1013

e 0.12 3.4 9 1013

f 0.05 7.2 9 1012

Fig. 7

H

a 0.17

CO

b 0.19

CHO

c 0.47

C3H2

d 0.92

C3H3

e 1.19

C6H6 half-ring

f 2.54

Benzene

g 2.63

h -0.06 2.9 9 1011

i 0.01 1.4 9 1013

Top Catal (2012) 55:290–299 299

123