Anda di halaman 1dari 11

Journal of Catalysis 324 (2015) 1424

Contents lists available at ScienceDirect

Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat

Theoretical investigation of the decarboxylation and decarbonylation


mechanism of propanoic acid over a Ru(0 0 0 1) model surface
Jianmin Lu, Muhammad Faheem 1, Sina Behtash, Andreas Heyden
Department of Chemical Engineering, University of South Carolina, 301 S. Main St., Columbia, SC 29208, USA

a r t i c l e i n f o a b s t r a c t

Article history: The hydrodeoxygenation of organic acids is often found to be a rate-controlling process during upgrading
Received 5 September 2014 of biomass feedstocks into fuels. We developed a microkinetic model based on data obtained from den-
Revised 3 January 2015 sity functional theory calculations for the decarboxylation and decarbonylation mechanisms of propanoic
Accepted 6 January 2015
acid (CH3CH2COOH) over a Ru(0 0 0 1) model surface. The model predicts that the decarbonylation mech-
anism is two orders of magnitude faster than the decarboxylation mechanism. The most favorable
decarbonylation pathway proceeds via removal of the acid OH group to produce propanoyl (CH3CH2CO)
Keywords:
followed by CCO bond scission of propanoyl to produce CH3CH2 and CO. Finally, CH3CH2 is hydrogenated
Ruthenium
Density functional theory
to CH3CH3. Dehydrogenation reactions that have been observed to be important over Pd catalysts play no
Microkinetic model role over Ru(0 0 0 1), and a sensitivity analysis indicates that removal of the acid OH group is the
Deoxygenation rate-controlling step in the deoxygenation. Overall, our results suggest that to improve the Ru catalyst
Hydrodeoxygenation performance for the decarbonylation of organic acids, the free site coverage needs to be increased by,
Organic acid for example, adding a catalyst promoter that decreases the hydrogen and CO adsorption strength (with-
out signicantly affecting the COH bond scission rate), or by raising the reaction temperature and
operating at relatively low CO and H2 partial pressures.
2015 Elsevier Inc. All rights reserved.

1. Introduction In our recent papers [11,12], we have studied the DCN and DCX
mechanisms of propanoic acid (CH3CH2COOH) to C2 hydrocarbons
Increased consumption of non-renewable fossil fuels in the last on Pd(1 1 1) surface models and found that the DCN is favored over
couple of decades has led to growing concerns over the environ- the DCX. To gain more insights into the DCN and DCX mechanisms
mental impact of fossil fuel utilization. Liquid fuels derived from of organic acids (such as propanoic acid) on transition metal sur-
renewable raw materials such as biomass have thus gained more faces and to possibly design novel metal catalysts, it is necessary
and more attention and are increasingly being developed. Intensive to study multiple metal surfaces. Ruthenium is more oxophilic
research is currently being conducted on the catalytic hydrodeox- than palladium and has experimentally been found to be an active
ygenation (HDO) of triglycerides (a major constituent of vegetable catalyst for the conversion of organic acids to hydrocarbon mole-
oil and animal fats) and fatty/carboxylic acids over transition metal cules [2,1315]. Experiments on the HDO of propanoic acid over
surfaces [19]. During the HDO of triglycerides, carboxylic acids Ru/C by Chen et al. [13] and the HDO of acetic acid over Ru/C by
have often been found to be reaction intermediates and the HDO Olcay et al. [14] show that at high H2 partial pressure (>20 bar)
of organic acids is often the rate-controlling process [10]. Three and at temperatures of around 400 K, Ru/C has a high selectivity
reaction mechanismsdecarbonylation (DCN), decarboxylation to propanol and ethanol, respectively. On the other hand, it has
(DCX), and reductive deoxygenation without CC bond cleavage been shown that when owing acetic acid over Ru/C at a low H2
(RDO)have been proposed to be the dominant deoxygenation partial pressure (<1 bar) [14], mostly methane (70% selectivity)
pathways for organic acids. Considering that the rst two reaction and only small amounts of ethanol are produced (6% selectivity).
mechanisms require less hydrogen, which is likely constrained in a Experiments on propanoic acid on Ru/SiO2 at low H2 partial pres-
future biorenery, there is a desire to design more active and selec- sure (<1 bar) by Lugo-Jose et al. [15] showed that the selectivity
tive decarbonylation/decarboxylation catalysts. to C2 hydrocarbons increases with increasing temperature in the
range of 473573 K and that the deoxygenation mechanism pro-
Corresponding author. ceeds by decarbonylation.
1
Current address: Department of Chemical Engineering, University of Engineering In this paper, we report a microkinetic modeling study based on
& Technology, Lahore 54890, Pakistan. data obtained from density functional theory (DFT) calculations for

http://dx.doi.org/10.1016/j.jcat.2015.01.005
0021-9517/ 2015 Elsevier Inc. All rights reserved.
J. Lu et al. / Journal of Catalysis 324 (2015) 1424 15

the DCN and DCX mechanisms of propanoic acid over Ru(0 0 0 1) 100 cm1 during the partition function calculations to minimize
terrace sites. We aim at determining the preferred reaction the errors of the harmonic approximation for small frequencies
mechanism and identifying possible descriptors relevant for the [12]. All reverse reaction rate constants are determined by the
selectivity of the DCN and DCX at low H2 partial pressure where equilibrium and forward rate constants. The nonlinear steady state
the alcohol production selectivity is low and can to a rst approx- surface species equations have nally been solved using the Bzz-
imation be neglected. Also, we will compare our Ru(0 0 0 1) results Math library developed by Buzzi-Ferraris [26].
with those obtained from Pd(1 1 1) [12].

3. Results and discussion


2. Methods
Fig. 1 illustrates the reaction pathways investigated for the DCN
We have employed the Vienna Ab Initio Simulation Package and DCX of propanoic acid to C2 alkanes over Ru(0 0 0 1). This
(VASP) [16,17], a periodic plane-wave-based density functional scheme does not include reactions that lead to C1 or C3 products.
theory program, to calculate all adsorption energies and vibra- C1 hydrocarbons are experimentally not observed; while C3 prod-
tional properties. The electronion interactions are described by ucts are possibly important at high hydrogen partial pressure, we
the projector-augmented wave method (PAW), which is a frozen- decided to focus this study on the relevant mechanisms at low
core all-electron method that uses the exact shape of the valence hydrogen partial pressure. The reductive deoxygenation (RDO)
wave functions instead of pseudo-wave functions [18]. The mechanism leading to C3 products is currently under investigation
exchange correlation energy has been calculated within the gener- in the context of a high hydrogen partial pressure environment.
alized gradient approximation by the Perdew and Wang 1991 for- In the following, we will discuss the adsorbed intermediates in
mulation (GGA-PW91) [19,20]. The kinetic energy cutoff of the Section 3.1. Then, we will investigate the potential energy surface
plane-wave basis set was xed to 400 eV in all calculations. To of various reaction pathways in Sections 3.2. Finally, we will
approximately consider dispersion interactions for adsorption develop a microkinetic model, compute turnover frequencies, and
and desorption processes of hydrocarbon species, we also performed compare our results to experimental observations in Section 3.3.
selected calculations using the PBE-D3 method [30]. Further details
about these calculations can be found in Section 3.3.2.
3.1. Adsorbed intermediates
Using the PW91 functional, the total energy of hcp-Ru bulk
approaches a minimum when its lattice constants are
Fig. 2 shows adsorption geometries of the reactants, products,
a = 2.7264 and c = 4.3134 , which are in reasonable agreement
and possible reaction intermediates involved in both DCN and
with the experimental values (a = 2.7059 and c = 4.2815
DCX mechanisms. There are 10 surface intermediates on
[21]). The Ru(0 0 0 1) surface was built as a periodic slab with four
Ru(0 0 0 1) which have a different (most stable) adsorption cong-
Ru layers separated by a vacuum layer of 15 in order to diminish
uration from their counterparts on Pd(1 1 1) [11]. The propanoic
lateral interactions between the slab and its periodic images. Each
p acid molecule adsorbs in an orientation in which both the carbonyl
Ru layer has 12 Ru atoms with a (3  2 3) periodicity, allowing for
O and the hydroxyl H bind to Ru atoms while on Pd(1 1 1) only the
adsorbate coverages as low as 1/12 ML. For all calculations, the
carbonyl O binds strongly to the Pd atop site. Similarly, both the
bottom two Ru layers were xed to their bulk positions to repre-
hydroxyl O of CH3CHCOOH and the carbonyl O of CH3CHCO bind
sent the semi-innite substrate while the top two layers were
to Ru atoms in atop position, while they do not bind to Pd atoms
allowed to relax in all directions. The adsorbates were free to relax
on Pd(1 1 1). Also, CO and the carbonyl C of CH3CH2CO bind to an
in all directions, too. Coordinates of the adsorbates and the Ru
atop site on Ru(0 0 0 1) while they bind to fcc hollow and bridge
atoms in the relaxed layers were optimized to a force less than
sites on Pd(1 1 1), respectively. Next, one C atom in CH2CH2 and
0.03 eV/ on each atom. All self-consistent eld (SCF) calculations
the C in COOH bind to bridge sites on Ru(0 0 0 1) while they bind
were converged to 1  103 kJ/mol. Brillouin-zone integration was
to atop sites on Pd(1 1 1). Finally, CHCH, CH3C, and OH bind to
performed using a 4  4  1 MonkhorstPack grid and a Methfes-
hcp hollow sites on Ru(0 0 0 1) while they prefer fcc hollow sites
selPaxton smearing of 0.2 eV.
on Pd(1 1 1).
Adsorption energies, Eads, of all surface intermediates were cal-
Table 1 lists the binding modes, adsorption energies, and zero-
culated in their most stable adsorption modes using equation 1
point energy corrections of all intermediates. We have employed
Eads Eslabadsorbate  Eslab  Eadsorbategas 1 the nomenclature gilj to designate that i atoms of the adsorbate
are binding to j atoms of the Ru surface. We observe that saturated
where Eslab+adsorbate is the total energy of the Ru slab with an adsor- intermediates, such as propanoic acid (Eads = 0.71 eV), ethane
bate binding on to it, Eslab is the total energy of the clean slab, and (0.07 eV), water (0.43 eV), and carbon dioxide (0.20 eV), are
Eadsorbate(gas) is the total energy of the adsorbate in the gas phase. relatively weakly adsorbed to the surface. These relatively weak
Transition states of elementary reactions were located by a bonds are formed either by interactions between the lone-pair
combination of the nudged elastic band (NEB) method [22] and electrons on the hydroxyl oxygen with the surface or by weak p-
the dimer method [2325]. In the NEB method, an interpolated bonded interactions of the C@O group with the surface [27]. Mean-
chain of congurations/images between the initial and nal states while, open-shell intermediates strongly adsorb onto the
was relaxed simultaneously. The image of the NEB calculation Ru(0 0 0 1) surface with adsorption energies ranging from
which is closest to a likely transition state structure was then 1.02 eV to 5.95 eV. Especially, the CH3C species binds very
employed as an initial guess structure for the dimer method. All strongly to the surface with an adsorption energy of 5.95 eV. Car-
adsorption energies and activation barriers reported in this study bon atoms tend to satisfy their tetrahedral bonding geometry on
have been zero-point energy corrected (DZPE). the surface [8], and a deeper dehydrogenation results in stronger
In our microkinetic model development, we used the same adsorption. For example, CH3CH3 adsorbs very weakly on
methodology as described in our latest publication [12]. For Ru(0 0 0 1). The rst dehydrogenation product CH3CH2 forms a sig-
adsorption processes, we used collision theory with a sticking coef- nicantly stronger bond with the surface (Eads = 1.76 eV), CH3CH
cient of 1 to calculate the reaction rate constant. For the calcula- adsorbs about twice as strong as CH3CH2 (Eads = 4.23 eV), and the
tion of surface reaction rate constants, harmonic transition state third dehydrogenation product CH3C adsorbs about three times as
theory has been used. Frequencies below 100 cm1 are shifted to strong as CH3CH2 (Eads = 5.95 eV) such that to a good approxima-
16 J. Lu et al. / Journal of Catalysis 324 (2015) 1424

Fig. 1. Reaction scheme of various decarbonylation and decarboxylation pathways of propanoic acid to C2 hydrocarbons over Ru(0 0 0 1).

tion each dehydrogenation leads to an increase in adsorption The former group might be viable since step 5 (dehydrogena-
strength of 1.62.0 eV. All intermediates, except ethane, have a tion of CH3CH2CO) has a low activation barrier of 0.32 eV and
stronger adsorption energy on Ru(0 0 0 1) than on Pd(1 1 1) which a reaction energy of 0.37 eV. The later pathway might be
can be explained by the higher oxophilicity of Ru. viable based on an activation barrier of 0.56 eV which is only
slightly larger than the barrier for the direct COH bond
3.2. Potential energy surface of various reaction pathways cleavage (step 1).
In the CH3CHCO-derived DCN pathways, CH3CHCO can be
The pathways shown in Fig. 1 can be classied into decarbonyla- decomposed by decarbonylation (step 15), a-carbon dehydrogena-
tion (DCN) or decarboxylation (DCX) pathways. Both DCN and DCX tion (step 16), or b-carbon dehydrogenation (step 17). Among
pathways can be further distinguished between direct decomposition these pathways, step 16 is preferred with an activation barrier of
pathways and pathways involving dehydrogenation steps. Snapshots 0.66 eV and a reaction energy of 0.43 eV. Step 17 has a similar
of all transition states involved are shown in Figs. 3 and 4, and the acti- reaction energy of 0.44 eV, but possesses a slightly larger activa-
vation barriers, reaction energies, and rate parameters at four charac- tion barrier of 0.78 eV. Finally, decarbonylation of CH3CHCO to
teristic reaction temperatures are listed in Table 2. CH3CH (step 15) is the most difcult step with an activation barrier
of 0.85 eV and a reaction energy of 0.66 eV.
3.2.1. Reaction pathways for the DCN mechanism The CH3CHCOOH-derived DCN pathways can proceed either by
The direct decomposition DCN pathway starts with OH initial OH removal to produce CH3CHCO (step 6) or by b-carbon
removal of propanoic acid (CH3CH2COOH) to produce CH3CH2CO dehydrogenation to CH2CHCOOH (step 7). We nd step 7 to be
and OH surface intermediates (step 1). This reaction is rather facile slightly preferred (Ea = 0.52 eV, DE0 = 0.51 eV) over step 6
with an activation barrier of only 0.49 eV, and a reaction energy of (Ea = 0.61 eV, DE0 = 0.44 eV) that merges here with the CH3-
0.36 eV. CH3CH2CO can decarbonylate into CH3CH2 and CO (step CHCO-derived pathways. Adsorbed CH2CHCOOH can be decom-
4) which has an activation barrier of 0.84 eV and a reaction energy posed either by initial OH removal to CH2CHCO (step 18)
of 0.41 eV. followed by decarbonylation to CH2CH (step 26), or by further
DCN pathways involving dehydrogenation steps can be b-carbon dehydrogenation to CHCHCOOH (step 20) followed by
divided into two groups. One group starts with a-carbon OH removal (step 28) and decarbonylation to CHCH (step 32).
dehydrogenation of CH3CH2CO to yield CH3CHCO (step 5); in Note that we also considered COOH removal steps, that is, steps
the following, we call these pathways CH3CHCO-derived reac- 10 (CH3CHCOOH ? CH3CH + COOH) and 27 (CH2CHCOOH ?
tion pathways. The other group starts with a-carbon dehydro- CH2CH + COOH); however, both steps have activation barriers of
genation of the acid to yield CH3CHCOOH (step 2) and we will 0.89 eV (step 10) and 0.98 eV (step 26) that are much higher than
call these pathways CH3CHCOOH-derived reaction pathways. other competitive reaction steps.
J. Lu et al. / Journal of Catalysis 324 (2015) 1424 17

Fig. 2. Side (upper panel) and top view (lower panel) of preferred adsorption structure of various surface intermediates in the reaction network of decarbonylation and
decarboxylation of propanoic acid over Ru(0 0 0 1).

3.2.2. Reaction pathways for the DCX mechanism a-carbon dehydrogenation steps of CH3CH2COO (step 11) and
The DCX mechanisms investigated can also be classied into CH3CHCOOH-derived pathways that start with dehydrogenation
direct decomposition pathways and pathways involving dehydro- step 2. CH3CHCOO can either be decarboxylated to CH3CH (step
genation steps. The direct decomposition pathway starts with ini- 23) or go through further dehydrogenation steps to CH3CCOO
tial OH bond scission of propanoic acid (CH3CH2COOH) to (step 22), followed by decarboxylation to CH3C (step 29). Both
produce CH3CH2COO (step 3), which is a facile and exothermic step steps 22 and 23 possess large barriers of 0.94 and 1.18 eV,
(Ea = 0.21 eV, DE0 = 1.02 eV), followed by decarboxylation of CH3- respectively.
CH2COO to CH3CH2 and CO2 (step 12). Step 12 has a high activation CH3CHCOOH-derived pathways follow either an OH bond scis-
barrier of 1.71 eV and is endothermic by 0.95 eV. Considering the sion of CH3CHCOOH to produce CH3CHCOO (step 9) or a full a-car-
high reaction barrier, it becomes apparent that the DCN pathways bon dehydrogenation to CH3CCOOH (step 8), followed by OH
are likely preferred to this DCX mechanism. bond scission to CH3CCOO (step 21). Step 9 is facile with an activa-
Next, other DCX pathways involving dehydrogenation steps tion barrier of 0.39 eV and a reaction energy of 0.68 eV. Also, step
can be divided into CH3CHCOO-derived pathways that start with 8 is relatively facile with an activation barrier of 0.62 eV and a
18 J. Lu et al. / Journal of Catalysis 324 (2015) 1424

Table 1
Binding modes, zero-point energy-corrected adsorption energies (Eads, in eV), and zero-point energy corrections (DZPE, in eV) of reaction intermediates calculated from DFT.

Species Stoichiometry Binding mode Eads (eV) DZPE (eV)


Propanoic acid CH3CH2COOH g1l1 (O) 0.64 0.04
Ethylidene-1-ol-1-olate CH3CHCOOH g3l3 (C,O,O) 1.94 0.05
Ethylidyne-1-ol-1-olate CH3CCOOH g2l3 (C,O) 4.13 0.11
Propanoyl CH3CH2CO g2l2 (C,O) 2.74 0.08
Carbonylethylidyne CH3CHCO g2l3 (C,O) 1.80 0.04
Carbonylethylidyne CH3CCO g2l4 (C,C) 3.61 0.00
Vinyl-1-ol-1-olate CH2CHCOOH g3l3 (C,C,O) 1.56 0.02
Carbonylvinyl CH2CHCO g4l4 (C,C,C,O) 2.99 0.07
Ethyne-1-ol-1-olate CHCHCOOH g3l3 (C,C,O) 3.88 0.08
Carbonylethyne CHCHCO g4l4 (C,C,C,O) 4.65 0.05
Propanoate CH3CH2COO g2l2 (O,O) 3.57 0.08
Carboxylethylidene CH3CHCOO g3l3 (C,O,O) 2.62 0.01
Carboxylethylidyne CH3CCOO g2l3 (C,O) 2.66 0.04
Carboxylic COOH g2l3 (C,O) 2.62 0.07
Ethyne CHCH g2l3 (C,C) 2.59 0.07
Vinyl CH2CH g2l3 (C,C) 3.38 0.10
Ethene CH2CH2 g2l2 (C,C) 1.00 0.10
Ethylidyne CH3C g1l3 (C) 5.92 0.18
Ethylidene CH3CH g1l2 (C) 4.20 0.08
Ethyl CH3CH2 g1l2 (C) 1.76 0.08
Ethane CH3CH3 g1l1 (H) 0.03 0.01
Hydrogen H g1l3 (H) 2.74 0.17
Hydroxyl OH g1l3 (O) 3.53 0.11
Water H2O g1l1 (O) 0.40 0.08
Carbon monoxide CO g1l1 (C) 1.88 0.08
Carbon dioxide CO2 g2l2 (C,O) 0.17 0.01

reaction energy of 0.48 eV. In contrast, the thermoneutral step 21 For the differential adsorption energy of CH3C, ECH3C(hCH3C), we
is slow with an activation barrier of 1.17 eV. found
Overall, it is difcult to identify the preferred reaction mecha-
ECH3 C hCH3 C 6:101 6:062  hCH3 C  0:192 4
nism from the activation and reaction energies under relevant
reaction conditions. As a result, we developed a microkinetic From coadsorption of H and CO at various coverages, we get
model based on parameters obtained from rst principles and har- EH(hH, hCO) and ECO(hH, hCO) as follows:
monic transition state theory.
EH hH ; hCO 2:906 0:168  hH  0:139 0:622  hCO
3.3. Microkinetic modeling  0:609  hCO hH hCO 0:5 5

3.3.1. Lateral interaction effects ECO hH ; hCO 1:955 1:748  hCO  0:087 0:622  hH
Mean-eld microkinetic models that do not consider lateral
 0:609  hH hH hCO 0:5 6
interactions at least approximately often predict non-physically
low turnover frequencies due to an exceedingly low free site cov- Similarly, from coadsorption of H and CH3C at various cover-
erage. This is a particular challenge for reactions involving gas- ages, we get EH(hH, hCH3C) and ECH3C(hH, hCH3C) as follows:
phase CO whose lateral interaction can be very strong. As a result,
we used a method similar to the one proposed by Grabow et al. EH hH ; hCH3 C 2:906 0:168  hH  0:139 1:049  hCH3 C
[28] and previously applied by us [12] to take lateral interaction  2:096  hCH3 C hH hCH3 C 0:5 7
effects for those surface intermediates into account that a preli-
minary study identied to have signicant surface coverages, that ECH3 C hH ; hCH3 C 6:101 6:062  hCH3 C  0:192
is, H, CO, and CH3C (linear parameterization of differential adsorp-
tion energies). We note that this approximate procedure for con- 1:049  hH  2:096  hH hH hCH3 C 0:5 8
sidering lateral interactions does not change surface reaction rate Next, from coadsorption of CO and CH3C at various coverages,
constants, but only changes (reduces) the surface coverages of we get ECO(hCO, hCH3C) and ECH3C(hCO, hCH3C) as follows:
the most dominant surface species. While lateral interactions can
also change elementary surface reaction rate constants, the deter- ECO hCO ; hCH3 C 1:955 1:748  hH  0:087  0:800
mination of coverage-dependent rate constants is beyond the  hCH3 C 8:105  hCH3 C hCO hCH3 C 0:5 9
scope of this work. The data used to t the coverage-dependent
adsorption energies can be found in the Supporting information. ECH3 C hH ; hCH3 C 6:101 6:062  hCH3 C  0:192  0:800
From adsorption of H atoms at various coverages (hH = 1/4 ML,
2/4 ML, 3/4 ML, 4/4 ML), the differential adsorption energy of H  hCO 8:105  hCO hCO hCH3 C 0:5 10
as a function of its coverage hH, EH(hH), is calculated to be
Finally, we get the coverage-dependent adsorption energy of H
EH hH 2:906 0:168  hH  0:139 2 and CO, EH(hH, hCO, hCH3C), and ECO(hH, hCO, hCH3C) as follows:

Similarly, the differential adsorption energy of CO as a function EH hH ; hCO ; hCH3 C 2:906 0:168  hH  0:139 0:622
of its coverage hCO, ECO(hCO), is calculated to be
 hCO  0:609  hCO hH hCO 0:5 1:049
ECO hCO 1:955 1:748  hCO  0:087 3
 hCH3 C  2:096  hCH3 C hH hCH3 C 0:5 11
J. Lu et al. / Journal of Catalysis 324 (2015) 1424 19

Fig. 3. Snapshots of transition states of steps 118 on the Ru(0 0 0 1) surface. Upper panels are for side views and lower ones for top views.

ECO hH ; hCO ; hCH3 C 1:955 1:748  hCO  0:087 0:622 to, the experimental value. To investigate the effect of dispersion
0:5 interactions on our microkinetic model, we report in the next sub-
 hH  0:609  hH hH hCO  0:800
section both microkinetic modeling results with parameters
 hCH3 C 8:105  hCH3 C hCO hCH3 C 0:5 12 obtained with PW91 functional and microkinetic modeling results
in which the PBE-D3 functional is used to compute the adsorption
Eqs. (11) and (12) have been implemented into our microkinetic
and desorption energetics of CH3CH2COOH, H2, CO, CO2, H2O, CH2-
model to include lateral interaction effects on the hydrogen and CO
CH2, CH3CH3. The PBE-D3 results are reported in [] brackets and
surface coverages.
Table 3 lists the zero-point energy-corrected adsorption energies
computed with both PW91 and PBE-D3. The computational setup
3.3.2. Dispersion interaction corrections
for the PBE-D3 calculations is identical to the one for the PW91 cal-
Dispersion interactions are potentially relevant for all adsorp-
culations except that the equilibrium lattice constants of the Ru
tion and desorption processes of hydrocarbon species in the reac-
bulk unit cell are calculated to be a = 2.7022 and c = 4.2736
tion mechanism of the HDO of propanoic acid over Ru(0 0 0 1).
with PBE-D3 functional which are slightly smaller than the PW91
Unfortunately, these interactions are not included in the PW91
result and very close to the experimental values.
functional such that adsorption energies computed by PW91 are
likely lower estimates. For example, it has experimentally been
shown that methanol with 1/9 ML coverage on Pt(1 1 1) has an 3.3.3. Turnover frequencies
adsorption energy of 58.6 0.8 kJ/mol [29]. But the PW91 func- The reaction temperature in our microkinetic model is 473 K,
tional predicts an adsorption energy of only 27.2 kJ/mol from and the reactant gas-phase pressures of propanoic acid and H2
our calculations with a 1/12 ML coverage. Empirical dispersion cor- are 0.01 bar and 1 bar, respectively, which are typical values in lab-
rections to DFT can to a large extent resolve this problem. For oratory-scale experiments. Assuming a small conversion of
instance, we have used the PBE-D3 method [30] to calculate the approximately 10%, we set the product partial pressures of CO,
adsorption energy of methanol with 1/16 ML coverage and CO2, H2O, CH2CH2, CH3CH3 to 103 bar. We note that our model
obtained 66.2 kJ/mol, which is slightly larger than, but quite close does not contain a watergas shift model such that we had to
20 J. Lu et al. / Journal of Catalysis 324 (2015) 1424

Fig. 4. Snapshots of transition states of steps 1936 on the Ru(0 0 0 1) surface. Upper panels are for side views and lower ones for top views.

approximate the product/CO partial pressures. Fortunately, we nd [0.04%]), and considering the approximate nature of our lateral
that our results are not very sensitive to the product/CO partial interaction model, an increase in the free site surface coverage of
pressures (see below). Under these conditions, we nd that ethane 12 orders of magnitude is conceivable which would increase the
is the preferred reaction product. TOF by 24 orders of magnitude (since we will show below that
Our model predicts the dominant surface intermediates to be H the rate-controlling step involves 2 free surface sites). In other
(57.2%, [33.3%]), CO (41.1%, [50.9%]), CH3C (1.2%, [2.2%]), and CH3- words, very realistic TOFs are conceivable on the Ru(0 0 0 1) surface.
CH2COO (0.14%, [6.8%]). The free site coverage is 0.16% [0.04%], that Next, the preference of the DCN over the DCX has previously
is, the catalyst surface is crowded with intermediates. We note that also been reported by us for the HDO of propanoic acid over
dispersion interactions signicantly increase the CH3CH2COO cov- Pd(1 1 1). The noticeable difference between these two surface
erage and decrease the free site coverage. Fig. 5 illustrates the turn- models is that on Ru(0 0 0 1) none of the key reaction pathways
over frequencies of various reaction pathways in our microkinetic involves dehydrogenation steps that we found to be essential over
model. Pd(1 1 1), particularly at low hydrogen partial pressures. For exam-
It is apparent that the direct DCN pathway is dominant (inde- ple, the pathways starting with a-carbon dehydrogenation of the
pendent of inclusion of dispersion interactions). It proceeds by ini- acid (step 2) have a low TOF of 3.0  1012 s1 [1.9  1011 s1]
tial OH removal of propanoic acid to CH3CH2CO (step 1), followed and are 5 orders of magnitude slower the than the direct decompo-
by decarbonylation to CH3CH2 (step 4) and hydrogenation to eth- sition DCN pathway. The pathways starting with a-carbon dehy-
ane (step 13). It has a TOF of 2.4  107 s1 [6.5  106 s1] which drogenation of CH3CH2CO and CH3CH2COO are also much slower.
is 2 orders of magnitude faster than the direct DCX pathway This observation can be explained with Ru(0 0 0 1) being a good
(2.8  109 s1, [9.8  108 s1]), steps 3, 12, and 13, such that CO bond cleavage catalyst and a poor dehydrogenation catalyst,
we conclude that the DCN mechanism is the dominant reaction while Pd(1 1 1) is worse at cleaving CO bonds but quite good at
pathway. We note that the microkinetic model with dispersion- dehydrogenation [12]. In other words, Pd catalyzes the HDO of
corrected adsorption/desorption energies predicts one order of propanoic acid by starting with a- and b-carbon dehydrogenation
magnitude higher TOF than the model without dispersion correc- of the acid to activate the acid group carbon before the COH bond
tions. Nevertheless, the predicted TOF is quite small which could can be cleaved and propanoic acid can be decarbonylated. These
mean that the Ru(0 0 0 1) surface is not active. However, it is noted dehydrogenation steps are not necessary on the more oxophilic
that the free site coverage is very small in our models (0.16% Ru surface that is primarily limited by its low free site coverage.
J. Lu et al. / Journal of Catalysis 324 (2015) 1424 21

Table 2
ZPE-corrected reaction energies (DE), activation barriers (E), equilibrium constants (K), and forward rate constants (k+) at various temperatures in the decarbonylation and
decarboxylation of propanoic acid to ethane over Ru(0 0 0 1).

Step Reaction Constant 473 K 523 K 573 K 623 K


0 CH3CH2COOH + ? CH3CH2COOH K0 2.11  104 5.24  105 1.73  105 6.83  106
DE0 = 0.63 eV k+0, s1 bar1 9.06  107 8.61  107 8.23  107 7.89  107
1 CH3CH2COOH + 2 ? CH3CH2CO + OH K1 4.09  103 1.89  103 1.07  102 5.99  102
DE1 = 0.36 eV, E1 = 0.49 eV k+1, s1 1.45  107 4.95  107 1.38  108 3.28  108
2 CH3CH2COOH + 3 ? CH3CHCOOH + H K2 3.48  102 1.77  102 1.03  102 6.54  101
DE2 = 0.30 eV, E2 = 0.56 eV k+2, s1 3.74  106 1.54  107 5.01  107 1.36  108
3 CH3CH2COOH + 2 ? CH3CH2COO + H K3 1.56  1010 1.40  109 1.92  108 3.64  107
DE3 = 1.02 eV, E3 = 0.21 eV k+3, s1 1.12  1010 2.00  1010 3.26  1010 4.95  1010
4 CH3CH2CO ? CH3CH2 + CO K4 5.84  104 2.45  104 1.20  104 6.66  103
DE4 = 0.41 eV, E4 = 0.84 eV k+4, s1 9.82  103 7.95  104 4.50  105 1.94  106
5 CH3CH2CO + 2 ? CH3CHCO + H K5 4.66  103 1.98  103 9.82  102 5.50  102
DE5 = 0.37 eV, E5 = 0.32 eV k+5, s1 2.29  109 5.27  109 1.06  1010 1.92  1010
6 CH3CHCOOH + ? CH3CHCO + OH K6 5.48  104 2.11  104 9.64  103 5.03  103
DE6 = 0.44 eV, E6 = 0.61 eV k+6, s1 1.38  106 6.00  106 2.02  107 5.57  107
7 CH3CHCOOH + ? CH2CHCOOH + H K7 1.03  105 2.98  104 1.08  104 4.61  103
DE7 = 0.51 eV, E7 = 0.52 eV k+7, s1 2.99  107 1.12  108 3.34  108 8.45  108
8 CH3CHCOOH + ? CH3CCOOH + H K8 3.07  104 9.73  103 3.79  103 1.72  103
DE8 = 0.48 eV, E8 = 0.62 eV k+8, s1 2.39  106 1.14  107 4.15  107 1.24  108
9 CH3CHCOOH + ? CH3CHCOO + H K9 5.14  106 9.93  105 2.56  105 8.25  104
DE9 = 0.68 eV, E9 = 0.39 eV k+9, s1 2.67  108 7.01  108 1.57  109 3.10  109
10 CH3CHCOOH + 2 ? CH3CH + COOH K10 2.54  102 1.63  102 1.13  102 8.34  101
DE10 = 0.21 eV, E10 = 0.89 eV k+10, s1 1.75  103 1.54  104 9.26  104 4.20  105
11 CH3CH2COO + 2 ? CH3CHCOO + H K11 3.97  102 4.36  102 4.75  102 5.15  102
DE11 = 0.05 eV, E11 = 0.75 eV k+11, s1 2.09  104 1.31  105 6.00  105 2.18  106
12 CH3CH2COO ? CH3CH2 + CO2 K12 1.83  1010 1.90  109 1.32  108 6.76  108
+

DE12 = 0.95 eV, E12 = 1.71 eV k12 , s1 1.44  106 8.63  105 2.56  103 4.45  102
13 CH3CH2 + H ? CH3CH3 + K13 1.79  102 3.41  102 5.76  102 8.89  102
DE13 = 0.26 eV, E13 = 0.84 eV k+13, s1 4.48  104 3.60  105 2.01  106 8.53  106
14 CH3CH + H ? CH3CH2 + 3 K14 3.27  107 1.42  106 4.73  106 1.29  105
DE14 = 0.63 eV, E14 = 0.70 eV k+14, s1 2.82  105 1.54  106 6.28  106 2.04  107
15 CH3CHCO + ? CH3CH + CO K15 3.83  107 8.73  106 2.59  106 9.36  105
DE15 = 0.66 eV, E15 = 0.85 eV k+15, s1 2.34  104 1.94  105 1.12  105 4.88  106
16 CH3CHCO + 2 ? CH3CCO + H K16 7.07  104 2.71  104 1.24  104 6.43  103
DE16 = 0.43 eV, E16 = 0.66 eV k+16, s1 1.94  106 1.02  107 4.06  107 1.30  108
17 CH3CHCO + 2 ? CH2CHCO + H K17 8.87  103 3.04  103 1.26  103 6.07  102
DE17 = 0.44 eV, E17 = 0.78 eV k+17, s1 4.82  104 3.35  105 1.67  106 6.51  106
18 CH2CHCOOH + 2 ? CH2CHCO + OH K18 4.73  103 2.15  103 1.13  103 6.63  102
DE18 = 0.37 eV, E11 = 0.99 eV k+18, s1 3.85  102 4.53  103 3.51  104 1.97  105
19 CH2CHCOOH + 2 ? CH2CH + COOH K19 8.48 7.34 6.56 5.99
+
DE37 = 0.09 eV, E37
= 0.94 eV k19 , s1 4.31  102 4.26  103 2.84  104 1.40  105
20 CH2CHCOOH + ? CHCHCOOH + H K20 3.08  103 1.35  103 6.86  102 3.90  102
DE20 = 0.35 eV, E20 = 0.27 eV k+20, s1 1.09  1010 2.25  1010 4.10  1010 6.81  1010
21 CH3CCOOH + ? CH3CCOO + H K21 2.01 2.13 2.24 2.34
DE21 = 0.01 eV, E21 = 1.17 eV k+21, s1 5.64 1.02  102 1.12  103 8.48  103
22 CH3CHCOO + ? CH3CCOO + H K22 1.20  102 2.08  102 3.31  102 4.91  102
DE22 = 0.22 eV, E22 = 0.94 eV k+22, s1 1.74  103 1.83  104 1.29  105 6.66  105
23 CH3CHCOO + ? CH3CH + CO2 K23 2.76  102 5.33  102 9.26  102 1.49  101
DE23 = 0.22 eV, E23 = 1.18 eV k+23, s1 1.43  101 2.61  102 2.88  103 2.18  104
24 CH3C + H + ? CH3CH K24 1.01  108 5.79  108 2.43  107 8.09  107
DE24 = 0.75 eV, E24 = 0.90 eV k+24, s1 2.45  103 2.22  104 1.37  105 6.31  105
25 CH3CCO ? CH3C + CO + 2 K25 5.35  1010 5.57  109 8.61  108 1.80  108
DE25 = 0.98 eV, E25 = 0.36 eV k+25, s1 1.98  109 5.08  109 1.11  1010 2.13  1010
26 CH2CHCO ? CH2CH + CO K26 1.48  107 3.87  106 1.28  106 5.10  105
+
DE26 = 0.61 eV, E26
= 0.98 eV k26 , s1 9.32  102 1.08  104 8.20  104 4.52  105
27 CH2CHCO + ? CHCHCO + H K27 1.02  105 3.34  104 1.34  104 6.27  103
DE27 = 0.48 eV, E27 = 0.51 eV k+27, s1 2.78  107 1.02  108 3.02  108 7.56  108
28 CHCHCOOH + 2 ? CHCHCO + OH K28 4.24  103 2.04  103 1.13  103 6.89  102
DE28 = 0.35 eV, E28 = 1.43 eV k+28, s1 9.86  103 3.37  101 6.31  100 7.45  101
29 CH3CCOO ? CH3C + CO2 + K29 1.32  108 2.89  107 8.28  106 2.90  106
DE29 = 0.69 eV, E29 = 0.44 eV k+29, s1 2.08  108 6.36  108 1.61  109 3.51  109
30 CH2CH + H ? CH2CH 2 +2

K30 4.30  105 1.15  104 2.56  104 5.00  104
DE30 = 0.42 eV, E30 = 0.75 eV k+30, s1 1.31  105 8.33  105 3.84  106 1.38  107
31 CHCH + H ? CH2CH + K30 9.44  107 3.64  106 1.10  105 2.77  105
DE31 = 0.58 eV, E31 = 0.95 eV k+30, s1 8.89  102 8.69  103 5.70  104 2.76  105
32 CHCHCO ? CHCH + CO K32 1.54  108 3.18  107 8.70  106 2.94  106
DE32 = 0.71 eV, E32 = 0.99 eV k+32, s1 5.09  102 5.91  103 4.50  104 2.49  105
33 CH3CH2 + 2 ? CH2CH 2 +H

K33 4.51  105 3.28  104 9.45  103 3.32  103
DE33 = 0.59 eV, E33 = 0.06 eV k+33, s1 9.05  1011 1.10  1012 1.30  1012 1.51  1012
34 COOH ? CO2 + H K34 3.22  102 2.12  102 1.51  102 1.14  102
DE34 = 0.20 eV, E34 = 0.68 eV k+34, s1 7.09  105 3.92  106 1.63  107 5.42  107
35 COOH ? CO + OH K35 8.25  109 1.13  109 2.21  108 5.64  107
DE35 = 0.89 eV, E35 = 0.34 eV k+35, s1 5.46  109 1.39  1010 3.02  1010 5.82  1010
36 OH + H ? H2O + K36 2.47  106 1.08  105 3.66  105 1.01  104

(continued on next page)


22 J. Lu et al. / Journal of Catalysis 324 (2015) 1424

Table 2 (continued)

Step Reaction Constant 473 K 523 K 573 K 623 K


DE36 = 0.61 eV, E36 = 1.22 eV k+36, s1 2.12 4.13  101 4.81  102 3.80  103
37 H2O ? H2O + K37 4.92  102 1.26  103 2.74  103 5.10  103
DE37 = 0.41 eV k+37, s1 9.03  1010 2.20  1011 4.57  1011 8.16  1011
38 CH3CH3 ? CH3CH3 + K38 4.04  106 4.24  106 4.33  106 4.38  106
DE38 = 0.07 eV k+38, s1 5.75  1014 5.74  1014 5.60  1014 5.43  1014
39 CH2CH2 ? CH2CH2 + K39 3.10  102 3.56  101 2.67  100 1.46  101
DE39 = 1.01 eV k+39, s1 4.56  106 4.99  107 3.57  108 1.88  109
40 CO2 ? CO2 + K40 3.84  105 5.56  105 7.40  105 9.30  105
DE40 = 0.20 eV k+40, s1 4.74  1013 6.52  1013 8.29  1013 9.99  1013
41 CO ? CO + K41 9.43  1012 1.14  109 6.24  108 1.86  106
DE41 = 1.88 eV k+41, s1 1.39  103 1.60  101 8.35  100 2.39  102
42 0.5H2(g) + ? H K42 2.20  103 4.57  102 1.23  102 4.03  101
DE42 = 0.60 eV k+42, s1 7.80  108 7.41  108 7.08  108 6.79  108

Table 3 COOH ? CH3CO + OH step is also rate controlling [10,27]. Consider-


ZPE-corrected adsorption energies (Eads, in eV) in eV of all stable gas-phase species in ing furthermore that two free sites are involved in step 1 and no
the decarbonylation and decarboxylation of propanoic acid to ethane on Ru(0 0 0 1)
free sites are involved in step 12, we conclude that increasing the
computed using the PW91 functional and the PBE-D3 method.
free site coverage will increase the DCN rate while it will have no
Eads (eV) PW91 PBE-D3 effect on the rate-controlling DCX step.
H2(g) + 2 ? 2H 1.26 1.27
CH3CH2COOH 0.64 0.93
3.3.5. Degree of thermodynamic rate control
CH3CH3 0.03 0.38
CH2CH2 1.00 1.37 We have used the degree of thermodynamic rate control [33
H2O 0.40 0.56 35], XTRC, to perform a sensitivity analysis and to determine the
CO 1.88 2.05 key surface intermediates at 473 K
CO2 0.17 0.40 0 1
1 @ @r A
X TRC;n   14
r @ G0n
Our predictions from this microkinetic model agree well with RT
G0mn ;G0;TS
i
experimental observations by Olcay et al. [14] and Lugo-Jose
et al. [15]. Olcay et al. studied the hydrodeoxygenation of acetic where Gn0 is the free energy of adsorbate n. Rate-controlling inter-
acid (CH3COOH) over Ru/C catalysts and found that at 458 K and mediates are identied by nonzero XTRC values. In the following, we
low H2 partial pressure, methane production has a selectivity of only report thermodynamic rate control values that deviate from
70%, following a decarbonylation mechanism. With increasing zero. Negative values signify that reducing the adsorption strength
H2 partial pressure, the selectivity to methane decreases linearly will increase the overall rate, while positive values signify that
while the selectivity to ethanol increases linearly. At the highest increasing the adsorption strength will increase the overall reaction
H2 partial pressure (corresponding to 1.4 mol/L in water), the rate. Our model predicts that adsorbed CH3CH2COO, CO, H, and the
selectivity to methane decreases to 40% while it is 55% to ethanol. free site are the surface intermediates with a signicant degree of
We will discuss the change in mechanism to a reductive deoxygen- thermodynamic control. Our model predicts that CH3CH2COO has
ation at high hydrogen partial pressure in a separate paper. Next, an XTRC of 0.10 [0.82], CO has an XTRC of 0.37 [0.14], H has
Lugo-Jose et al. investigated the hydrodeoxygenation of propanoic an XTRC of 2.40 [0.03], and the free site has an XTRC of 1.88
acid over Ru/SiO2 catalysts at low H2 partial pressures and found [0.65]. In other words, our model without considering dispersion
that in a temperature range of 473573 K the selectivity to C2 interactions for adsorption predicts that destabilizing the adsorp-
hydrocarbons increases with increasing temperature and generally tion of H and to a lesser degree destabilizing the adsorption of CO
follows a decarbonylation mechanism. will increase the overall rate, while our model that considers (and
possibly overestimates) dispersion interactions for adsorption nds
3.3.4. Degree of rate control that destabilizing CH3CH2COO and to a smaller degree CO will
We have used Campbells degree of rate control [31,32], XRC, to increase the overall rate. Finally, the large positive XTRC value for
perform a sensitivity analysis and to determine the rate-control- the free site suggests that increasing the free site coverage of the
ling steps in our mechanism at 473 K. The degree of rate control Ru surface will dramatically increase the TOF.
is dened as
0 1 3.3.6. Apparent activation energy and reaction orders
1 @ @r A To understand the sensitivity of our modeling results with
X RC;n   13 regard to temperature and partial pressure, we have calculated
r @ GTS
n
RT
GTS 0;ads
mn ;Gi the apparent activation energy (Eapp) and reaction orders (ai)
 
where r is the overall rate of reaction and GnTS is the free energy of @ lnr
Eapp RT 2 15
the transition state for elementary step n. Rate-controlling steps are @T Pi
identied by nonzero XRC values. In the following, we only report
 
rate control values that deviate from zero. Our microkinetic model @ lnr
predicts that the reaction CH3CH2COOH + 2 ? CH3CH2CO + OH ai 16
@ lnpi T;pji
(step 1) has an XRC of 0.99 [0.98] and CH3CH2COO ? CH3CH2 + CO2-

(step 12) has an XRC of 0.01 [0.02], that is, the COH bond scission Our model predicts a large apparent activation energy (Eapp) of
is rate controlling. These results agree with observations on the 2.27 eV [1.67 eV] in the temperature range of 433473 K which can
HDO of acetic acid where it has been reported that the CH3- be explained by the free energy barrier of COH bond cleavage
J. Lu et al. / Journal of Catalysis 324 (2015) 1424 23

Fig. 5. Turnover frequencies (s1) of all elementary reactions steps at 473 K. Numbers in brackets are from the microkinetic model with dispersion-corrected adsorption/
desorption energies. The red and green numbered pathways correspond to the preferred decarbonylation and decarboxylation pathways, respectively.

relative to the clean surface and the low free site coverage in the cient number of free sites since the Ru(0 0 0 1) surface is almost
low temperature range. As for the reaction orders at 473 K, the fully covered by H, CO, CH3C, and CH3CH2COO. To improve the
model predicts that the order with respect to propanoic acid is Ru(0 0 0 1) catalyst for the decarbonylation of organic acids, the free
0.35 [0.06] in the acid partial pressure range of 0.010.33 bar, illus- site coverage needs to be increased by, for example, adding a cat-
trating that increasing the propanoic acid chemical potential does alyst promoter that decreases the hydrogen and CO adsorption
not increase the TOF signicantly due to a lack of free sites. The strength without signicantly affecting the COH bond scission
order with respect to H2 is 0.97 [0.25] in the H2 partial pressure rate, or by raising the reaction temperature and operating at rela-
range of 0.52.0 bar, indicating again the high hydrogen surface tively low CO and H2 partial pressures. Finally, we note that this
coverage and the need for more free sites. Finally, the reaction investigation did not include reactions that lead to C3 products
order with respect to CO is 0.22 [0.16] in the CO partial pressure such as propanol that are relevant at high hydrogen partial pres-
range of 105103 bar, illustrating that CO only slightly dimin- sures. The reductive deoxygenation (RDO) mechanism leading to
ishes the reaction rate and that our results and conclusions are C3 products is currently under investigation in the context of a high
likely valid over a large CO pressure range. hydrogen partial pressure environment.

4. Conclusions Acknowledgments

The decarbonylation and decarboxylation mechanisms of prop- We appreciate the support from the National Science Founda-
anoic acid (CH3CH2COOH) to C2 hydrocarbons have been studied tion (NSF grant number CHE-1153012) and the U.S. Department
over Ru(0 0 0 1) by mean-eld microkinetic modeling based on data of Energy, Ofce of Basic Energy Sciences, Chemical Sciences Divi-
obtained from density functional theory and transition state the- sion under Contract DE-FG02-11ER16268 (DE-SC0007167). Com-
ory. Our model predicts that at low H2 partial pressures the decarb- putational resources have been provided by the National Energy
onylation mechanism is preferred over the decarboxylation Research Scientic Computing Center (NERSC) which is supported
mechanism. The most favorable decarbonylation pathway pro- by the Ofce of Science of the U.S. Department of Energy and in
ceeds by direct removal of the acid OH group to produce propa- part by the Extreme Science and Engineering Discovery Environ-
noyl (CH3CH2CO), followed by CC bond scission of CH3CH2CO to ment (XSEDE) provided by the Texas Advanced Computing Center
CH3CH2 and CO, and nally hydrogenation to ethane. Dehydroge- (TACC) at the University of Texas at Austin under grant number TG-
nation pathways that have been found to be relevant over Pd cat- CTS090100. Finally, computing resources from USC NanoCenter
alysts do not play any role over Ru(0 0 0 1). The rate-controlling and USCs High Performance Computing Group are gratefully
step is the removal of the acid OH group and the creation of a suf- acknowledged.
24 J. Lu et al. / Journal of Catalysis 324 (2015) 1424

Appendix A. Supplementary material [15] Y.K. Lugo-Jose, J.R. Monnier, C.T. Williams, Appl. Catal. A-Gen. 469 (2014) 410
418.
[16] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558561.
Supplementary data associated with this article can be found, in [17] G. Kresse, J. Furthmuller, Comput. Mater. Sci. 6 (1996) 1550.
the online version, at http://dx.doi.org/10.1016/j.jcat.2015.01.005. [18] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 17581775.
[19] J.P. Perdew, Y. Wang, Phys. Rev. B 33 (1986) 88008802.
[20] J.P. Perdew, Y. Wang, Phys. Rev. B 45 (1992) 1324413249.
References [21] V.A. Finkel, G.P. Kovtun, M.I. Palatnik, Phys. Met. Metallogr. 32 (1971) 231
235.
[1] B. Donnis, R.G. Egeberg, P. Blom, K.G. Knudsen, Top. Catal. 52 (2009) 229240. [22] G. Henkelman, B.P. Uberuaga, H. Jonsson, J. Chem. Phys. 113 (2000) 9901
[2] M. Snare, I. Kubickova, P. Maki-Arvela, K. Eranen, D.Y. Murzin, Ind. Eng. Chem. 9904.
Res. 45 (2006) 57085715. [23] G. Henkelman, H. Jonsson, J. Chem. Phys. 111 (1999) 70107022.
[3] I.L. Simakova, O.A. Simakova, P. Maki-Arvela, A. Simakov, M. Estrada, D.Y. [24] R.A. Olsen, G.J. Kroes, G. Henkelman, A. Arnaldsson, H. Jonsson, J. Chem. Phys.
Murzin, Appl. Catal. A-Gen. 355 (2009) 100108. 121 (2004) 97769792.
[4] J.G. Immer, H.H. Lamb, Energy Fuels 24 (2010) 52915299. [25] A. Heyden, A.T. Bell, F.J. Keil, J. Chem. Phys. 123 (2005) 224101.
[5] J. Fu, X. Lu, P.E. Savage, Energy Environ. Sci. 3 (2010) 311317. [26] G. Buzzi-Ferraris, BzzMath: Numerical libraries in C++, Politecnico di Milano:
[6] E.W. Ping, R. Wallace, J. Pierson, T.F. Fuller, C.W. Jones, Microporous <http://www.chem.polimi.it/homes/gbuzzi>.
Mesoporous Mater. 132 (2010) 174180. [27] V. Pallassana, M. Neurock, J. Catal. 209 (2002) 289305.
[7] L. Boda, G. Onyestyak, H. Solt, F. Lonyi, J. Valyon, A. Thernesz, Appl. Catal. A- [28] L.C. Grabow, B. Hovlbak, J.K. Norskov, Top. Catal. 53 (2010) 298310.
Gen. 374 (2010) 158169. [29] E.M. Karp, T.L. Silbaugh, M.C. Crowe, C.T. Campbell, J. Am. Chem. Soc. 134
[8] L. Xu, Y. Xu, Surf. Sci. 604 (2010) 887892. (2012) 2038820395.
[9] B. Rozmyslowicz, P. Maki-Arvela, A. Tokarev, A.-R. Leino, K. Eranen, D.Y. [30] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, J. Chem. Phys. 132 (2010) 154104.
Murzin, Ind. Eng. Chem. Res. 51 (2012) 89228927. [31] C.T. Campbell, Top. Catal. 1 (1994) 353366.
[10] H. Olcay, L. Xu, Y. Xu, G.W. Huber, ChemCatChem 2 (2010) 14201424. [32] C.T. Campbell, J. Catal. 204 (2001) 520524.
[11] J.M. Lu, S. Behtash, A. Heyden, J. Phys. Chem. C 116 (2011) 1432814341. [33] S. Kozuch, S. Shaik, J. Am. Chem. Soc. 128 (2006) 33553365.
[12] J.M. Lu, S. Behtash, M. Faheem, A. Heyden, J. Catal. 305 (2013) 5666. [34] S. Kozuch, S. Shaik, J. Phys. Chem. A 112 (2008) 60326041.
[13] Y. Chen, D.J. Miller, J.E. Jackson, Ind. Eng. Chem. Res. 46 (2007) 33343340. [35] C. Stegelmann, A. Andreasen, C.T. Campbell, J. Am. Chem. Soc. 131 (2009)
[14] H. Olcay, Y. Xu, G.W. Huber, Green Chem. 16 (2014) 911924. 80778082.

Anda mungkin juga menyukai