Topics in Catalysis

, Volume 55, Issue 5, pp 280–289

Density Functional Theory Study of Selectivity Considerations for C–C Versus C–O Bond Scission in Glycerol Decomposition on Pt(111)

Authors

  • Bin Liu
    • Center for Nanoscale MaterialsArgonne National Laboratory
    • Center for Nanoscale MaterialsArgonne National Laboratory
Original Paper

DOI: 10.1007/s11244-012-9806-2

Cite this article as:
Liu, B. & Greeley, J. Top Catal (2012) 55: 280. doi:10.1007/s11244-012-9806-2

Abstract

Glycerol decomposition via a combination of dehydrogenation, C–C bond scission, and C–O bond scission reactions is examined on Pt(111) with periodic Density Functional Theory (DFT) calculations. Building upon a previous study focused on C–C bond scission in glycerol, the current work presents a first analysis of the competition between C–O and C–C bond cleavage in this reaction network. The thermochemistry of various species produced from C–O bond breaking in glycerol dehydrogenation intermediates is estimated using an extension of a previously introduced empirical correlation scheme, with parameters fit to DFT calculations. Brønsted–Evans–Polanyi (BEP) relationships are then used to estimate the kinetics of C–O bond breaking. When combined with the previous results, the thermochemical and kinetic analyses imply that, while C–O bond scission may be competitive with C–C bond scission during the early stages of glycerol dehydrogenation, the overall rates are likely to be very low. Later in the dehydrogenation process, where rates will be much higher, transition states for C–C bond scission involving decarbonylation are much lower in energy than are the corresponding transition states for C–O bond breaking, implying that the selectivity for C–C scission will be high for glycerol decomposition on smooth platinum surfaces. It is anticipated that the correlation schemes described in this work will provide an efficient strategy for estimating thermochemical and kinetic energetics for a variety of elementary bond breaking processes on Pt(111) and may ultimately facilitate computational catalyst design for these and related catalytic processes.

Keywords

Density functional theoryScaling relationshipsBiomassGlycerolHydrogen productionPt(111)SelectivityReforming

1 Introduction

Catalytic transformations of biomass can be used to produce a variety of useful fuels and chemicals [16]. Hydrogen produced from oxygenated hydrocarbons, including alcohols and polyols, is one example of such a transformation and can be used either as a fuel or as a means to reduce the oxygen content of biomassic molecules [713]. Glycerol is a promising feedstock in this regard, as it can be a byproduct of biodiesel production [1417].

Reforming of oxygenated hydrocarbons involves a combination of dehydrogenation, C–C bond scission, and C–O bond scission reactions, as well as water–gas shift chemistry. The great complexity of these catalytic processes has inspired numerous experimental studies of the reforming of model alcohols and polyols [1825] in an effort to obtain fundamental mechanistic insights into these reaction networks. A variety of computational, Density Functional Theory (DFT)-based analyses have also been carried out and have focused largely on the competition between C–H, O–H, and C–C bond scission for these molecules on platinum [2632]; the studies have determined, among other conclusions, that the oxygenated hydrocarbons must be substantially dehydrogenated before it becomes favorable to cleave C–C bonds. Significantly, a combined experimental and computational analysis of ethylene glycol decomposition and reforming on platinum has shown that computational studies of decomposition reactions at gas–solid interfaces, including C–H, O–H, and C–C bond scission, can provide important insights into the related aqueous phase reforming processes, which include water–gas shift chemistry in addition to the decomposition reactions [29].

Fewer computational studies have focused explicitly on selectivity issues in biomolecule reforming, where cleavage of C–O bonds ultimately leads to the production of alkanes and reduces the hydrogen yield [7]. In a pioneering study, Ferrin et al. [33] performed a combined activity and selectivity analysis of ethanol decomposition on the close-packed surfaces of a series of transition metals; they determined that C–C scission was more energetically favorable than C–O scission on most studied metals, including platinum. This conclusion is broadly consistent with experimental results for smaller alcohols and polyols [7, 34], but selectivity to C–C scission is known to decrease under certain conditions and for more complex polyols, and with the exception of explicit transition state searches to probe the very early stages of glycerol decomposition on rhodium [35], the complexity of these processes has so far precluded any detailed elementary mechanistic analyses.

In this contribution, we take the first steps towards a computational analysis of the competition between C–C and C–O bond scission during glycerol decomposition on Pt(111). The work builds on a previous study of the C–H, O–H, and C–C bond breaking steps in glycerol and its direct dehydrogenation intermediates [32]. We introduce a correlation scheme and a modified Brønsted–Evans–Polanyi (BEP) relationship that greatly accelerate the analysis of the thermochemistry and kinetics, respectively, of C–O bond scission in intermediates resulting from glycerol dehydrogenation, and we describe the geometries of some key intermediates associated with these elementary reactions. We compare the resulting energetics with corresponding values obtained for C–C bond activation, and we close by discussing the competition between C–O and C–C bond scission during glycerol decomposition.

2 Computational Methods

The Vienna Ab initio Simulation Package (VASP) [3639], a periodic, plane wave-based code, is used for the DFT calculations. Ionic cores are described by the projector augmented wave (PAW) method [40, 41], and the exchange-correlation energy is described with the GGA–PW91 functional [42, 43].

A three layer, close-packed Pt(111) surface with at least five equivalent layers of vacuum between any successive metal slabs is used to determine the thermochemistry of the elementary reaction intermediates. The DFT-determined lattice constant is found to be 3.99 Å, which compares well with the experimental bulk lattice constant (3.92 Å) [44]; this lattice value is also consistent with other lattice constants reported in the literature for Pt [27]. A p(4 × 4) unit cell (surface coverage of 1/16 ML) is used in all calculations. The top layer is relaxed for all thermochemical geometry optimizations. The surface Brillouin zone is sampled with four special k points based on the Monkhorst–Pack sampling scheme [45], and the Kohn–Sham valence states are expanded in a plane wave basis set up to 25 Ry (or 340 eV). The self-consistent iterations are converged within 1 × 10-6 eV, and the ionic steps are converged to 0.02 eV/Å (the maximum force on each atom). A Methfessel–Paxton smearing of kBT = 0.2 eV is used [46], then the total energies are extrapolated to 0 K. Test calculations with different k-point sets and cutoff energies indicate that binding energies are converged to within better than 0.1 eV. Dipole corrections are included in the reported results [47], but no zero-point energy corrections are included. The energies and geometries of intermediates resulting from C–O bond breaking in glycerol dehydrogenation intermediates in the gas phase are calculated using a box with dimensions of 18 × 19 × 20 Å. The gamma-point k point sampling is used, together with a Gaussian smearing parameter of 0.01 eV, for these analyses. Spin polarization is applied for gas phase species due to the possible presence of unpaired electrons.

Transition states (TS) for C–O bond scission reactions are calculated using the Climbing Image Nudged Elastic Band (CI-NEB) [48, 49] method on static two layer slabs combined with the dimer method [50, 51]. Each transition state is confirmed to have a single imaginary frequency with vibrational frequency analysis.

3 Results

In this section, we introduce a simple correlation scheme to estimate the thermochemistry of adsorbed species resulting from the cleavage of a single C–O bond in glycerol dehydrogenation intermediates on Pt(111). This correlation is an extension of previously developed relationships in which the thermochemistry of glycerol dehydrogenation intermediates (all with equal numbers of carbon and oxygen atoms) was determined by fitting to the results of explicit DFT calculations [32]. We discuss this correlation procedure in detail, and we then describe the geometries, determined by explicit DFT calculations, of selected species resulting from C–O bond scission in glycerol dehydrogenation intermediates. We briefly introduce a BEP relationship for the determination of both C–C and C–O scission transition state energies, and we combine our thermochemical and kinetic results to produce a free energy diagram that compares the energetics of C–C and C–O bond scission on Pt(111).

3.1 Scaling Correlations for Binding Energies of Species Resulting from C–O Bond Scission in Glycerol Dehydrogenation Intermediates

A total of 84 intermediates, ranging from glycerol itself to CO–CO–CO, can result from cleavage of different combinations of C–H and O–H bonds in glycerol, and many additional species can be produced via cleavage of a single C–O bond in these intermediates. Performing explicit DFT calculations for all of these species on Pt(111) would be computationally intensive as well as unnecessary for practical catalytic investigations, as many such intermediates are likely to be highly metastable. Previously, a relatively simple correlation scheme was introduced to rapidly determine the thermochemistry of intermediates resulting from direct glycerol dehydrogenation with no C–O bond breaking [32]; each carbon atom in that group of intermediates was therefore bonded to one oxygen atom. The approach, which can be thought of as an empiricized version of well-known bond order conservation techniques [52, 53], is briefly reviewed below, and extensions of the approach to treat molecules with unequal numbers of carbon and oxygen atoms are then described.

Equation 1 is a definition of the binding energy of glycerol dehydrogenation intermediates (referenced to stoichiometrically appropriate amounts of gaseous glycerol and molecular hydrogen) with equal numbers of carbon and oxygen atoms in all intermediates; this expression is used to determine the binding energies for those intermediates for which explicit DFT calculations have been performed:
$$ {\text{BE}}_{\text{DFT}} = E_{{{\text{C}}_{3} {\text{H}}_{x} {\text{O}}_{3}^{*} }} - E_{*} - E_{{{\text{glyerol}}\left( {\text{g}} \right)}} + \frac{8 - x}{2}E_{{{\text{H}}_{ 2} \left( {\text{g}} \right)}} $$
(1)
BEDFT is the binding energy for the adsorbed C3HxO3* species, as determined from the DFT calculations. \( E_{{{\text{C}}_{3} {\text{H}}_{x} {\text{O}}_{3}^{*} }} \) is the DFT-calculated total energy of the adsorbed intermediate, \( E_{*} \) is the clean Pt(111) energy, \( E_{{{\text{glyerol}}\left( {\text{g}} \right)}} \) is the gas phase energy of glycerol, and \( E_{{{\text{H}}_{ 2} \left( {\text{g}} \right)}} \) is the corresponding energy for gas phase hydrogen. The scaling correlation, in turn, is defined as:
$$ {\text{BE}}_{{{\text{C}}_{3} {\text{H}}_{x} {\text{O}}_{3}^{*} }} = \sum\limits_{i} {p_{{{\text{C}}i}} v_{{{\text{C}}i}} } + \sum\limits_{i} {p_{{{\text{O}}i}} v_{{{\text{O}}i}} } + \sum\limits_{i,j} {p_{{{\text{C}}i{\text{O}}j}} v_{{{\text{C}}i{\text{O}}j}} } + \sum\limits_{i,j} {p_{{{\text{C}}i{\text{C}}j}} v_{{{\text{C}}i{\text{C}}j}} } + {\text{BE}}_{\text{glycerol}} . $$
(2)
\( {\text{BE}}_{{{\text{C}}_{3} {\text{H}}_{x} {\text{O}}_{3}^{*} }} \) is the estimated binding energy from the scheme, where x ranges from 0 to 8, depending on the level of dehydrogenation. \( {\text{BE}}_{{{\text{glycerol}}^{ *} }} \) is the binding energy (0.46 eV) of gas phase glycerol when adsorbing onto Pt(111) [32]. The variables vCi and vOi are a measure of the degree of undersaturation of the carbon or oxygen atoms:
$$ \nu_{i} = \frac{{n_{\max } - n_{\text{bond}} }}{{n_{\max } }}, $$
(3)
where \( n_{\max } \) is 4 for C and 2 for O, while \( n_{\text{bond}} \) is the total number of atoms, including H, C, and O, to which the atom in question is bonded (for adsorbed glycerol itself, all of the vCi and vOi terms are zero). pCi, pOi, pCiOj and pCiCj are parameters determined by fitting to DFT-determined binding energies (Eq. 1) with appropriate symmetry constraints (for example, the terminal carbon atoms in the glycerol backbone are assigned the same parameter). The first order sums are over all carbon and oxygen atoms in the dehydrogenation intermediates, while the second-order sums are restricted to nearest neighbor carbon–carbon or carbon–oxygen pairs. We found, previously, that only one second order parameter, representing the interaction between terminal carbon and oxygen atoms, was statistically significant in a data set of 47 DFT-determined binding energies, with pC1 = 0.90, pC2 = 0.65, pO1 = 1.95, pO2 = 1.56, and pC1O1 = pC3O3 = –2.16. These parameters described all binding energies with an average error of less than 0.15 eV.
The scaling approach described by Eqs. 2 and 3 is motivated by the principle, related to bond order conservation arguments, that the undersaturation of carbon and oxygen atoms is closely related to the strength of their interaction with the Pt(111) surface. We note, however, that the scheme is not intended to represent a rigorous application of bond order conservation theory, and its accuracy must be evaluated independently for different classes of molecules. Using exactly the same set of parameters, the scheme does, in fact, provide satisfactory descriptions of the binding energies of other simple alcohols or polyols with equal numbers of carbon and oxygen atoms. Figure. 1a is a parity plot showing the predictions of the scheme for intermediates resulting from dehydrogenation of methanol and ethylene glycol, together with selected glycerol dehydrogenation intermediates and mono-dehydrogenated erythritol intermediates (the associated DFT-calculated binding energies are reported in Table S1 of the Supporting Information). The average error is 0.11 eV, which is within the error bars typically associated with DFT calculations. However, Eq. 2 does not reliably describe the thermochemical properties of molecules with unequal numbers of carbon and oxygen atoms, such as ethanol. It appears that the presence of an oxygen atom, with or without an attached hydrogen atom, significantly alters the electronic properties of adjacent carbon atoms, and a modified expression must be introduced to account for these changes. In the somewhat heuristic spirit of our correlation approach, this can be accomplished by adding parameters that account for the presence or absence of oxygen adjacent to carbon atoms:
$$ {\text{BE}}_{{{\text{C}}_{x} {\text{H}}_{y} {\text{O}}_{x - 1} }} = \sum\limits_{i} {p_{{{\text{C}}i}} v_{{{\text{C}}i}} [1 + p_{{{\text{CO}}i}} (1 - n_{{{\text{CO}}i}} )]} + \sum\limits_{i} {p_{{{\text{O}}i}} v_{{{\text{O}}i}} } + \sum\limits_{i,j} {p_{{{\text{C}}i{\text{O}}j}} v_{{{\text{C}}i}} [1 + p_{{{\text{CO}}i}} (1 - n_{{{\text{CO}}i}} )]} \nu_{{{\text{O}}j}} + {\text{BE}}_{{{\text{C}}_{x} {\text{H}}_{2x + 2} {\text{O}}_{x - 1} }} $$
(4)
\( {\text{BE}}_{{{\text{C}}_{x} {\text{H}}_{y} {\text{O}}_{x - 1} }} \) is the estimated binding energy for the intermediate CxHyOx–1 from the scaling scheme, referenced to the relevant fully hydrogenated gas phase species and the appropriate numbers of gas phase H2 molecules. In the scope of this paper, x = 2 or 3, while y ranges from 0 to 8 depending on the number of carbon atoms and the level of dehydrogenation. \( {\text{BE}}_{{{\text{C}}_{x} {\text{H}}_{2x + 2} {\text{O}}_{x - 1} }} \) represents the binding energy of the gas phase reference molecules, C2H6O and C3H8O2 (ethanol, 1,2-propanediol, and 1,3-propanediol). The binding energies are 0.39, 0.48 and 0.62 eV, respectively. nCOi is a conditional parameter such that nCOi = 1 if the C atom is bonded to an O atom, and nCOi = 0 otherwise. pCO1 (0.79) and pCO2 (1.91) are the new parameters introduced to account for the differences in electronic structure of the C atoms not bonded to O atoms. pCO1 is assigned to C atoms at terminal positions and is fit to DFT data from ethanol dehydrogenation intermediates, while pCO2 is assigned to middle C atoms and is fit to 1,3-propanediol dehydrogenation intermediates. No additional re-fitting is needed for the parameters originally fit to the glycerol dataset.
https://static-content.springer.com/image/art%3A10.1007%2Fs11244-012-9806-2/MediaObjects/11244_2012_9806_Fig1_HTML.gif
Fig. 1

a Comparison of the predicted binding energies from an empirical scaling scheme and DFT calculations for methanol, ethylene glycol, glycerol and erythritol dehydrogenation intermediates on Pt(111). b Comparison of predicted binding energies from a modified scaling scheme and DFT calculations for intermediates with unequal numbers of carbon and oxygen atoms, including dehydrogenation intermediates of ethanol, 1,2-propanediol and 1,3-propanediol, on Pt(111)

In Fig. 1b, we illustrate the use of Eq. 4 to correlate the binding energies of 62 species, including a set of ethanol dehydrogenation intermediates (24 species), as well as intermediates resulting from varying degrees of dehydrogenation of 1,2-propanediol (21 species) and 1,3-propanediol (17 species). The particular dehydrogenation intermediates of 1,2- and 1,3-propanediol for which explicit DFT calculations are performed are chosen so that a reasonable variety of functional groups and molecular structures are represented in the parameter training set. The addition of the two extra parameters yields a mean absolute error, compared to the DFT-determined binding energies, of 0.16 eV for the entire set of 62 intermediates (see Table S2 in the Supporting Information for a complete listing of binding energies), and only one intermediate, CCOH, exhibits an error of >~0.4 eV. The fact that the agreement for the 1,2-propanediol intermediates is quite good, and no parameters were explicitly fit to the DFT-determined binding energies of its intermediates, is additional evidence for the predictive power of this approach.

3.2 Structures of Selected Species Resulting from Cleavage of a Single C–O Bond in Glycerol Dehydrogenation Intermediates

In Fig. 2, we show the most stable, DFT-determined adsorption configurations for selected species resulting from C–O bond scission in glycerol dehydrogenation intermediates. These most stable geometries are determined by explicitly calculating up to ten conformations for each intermediate, which are in turn inspired by binding motifs determined in previous studies of methanol, ethanol, and ethylene glycol adsorption on Pt(111) [2729]. These particular species correspond to 1,2- and 1,3-propanediol and to the products of the most kinetically favorable C–O bond breaking events at each level of glycerol dehydrogenation (the approach to identifying these specific events is described in detail in Sections 3.3 and 3.4 below).
https://static-content.springer.com/image/art%3A10.1007%2Fs11244-012-9806-2/MediaObjects/11244_2012_9806_Fig2_HTML.gif
Fig. 2

Schematic illustrations of the most stable configurations of 1,2- and 1,3-propanediol and intermediates resulting from the C–O bond breaking events corresponding to the lowest transition state free energies (based on a BEP analysis—see Fig. 3) of glycerol dehydrogenation intermediates on Pt(111). (i) CH2OH–CHOH–CH3; (ii) CH2OH–CH2–CH2OH; (iii) CH2OH–CH–CH2OH; (iv) CH2OH–CHOH–CH2; (v) CH2OH–C–CH2OH; (vi) CHOH–CHOH–CH; (vii) CH2OH–COH–C; (viii) CHOH–COH–C; (ix) COH–C–COH; (x) COH–COH–C; (xi) CO–C–CO; (xii) CO–CO–C

3.2.1 1,2- and 1,3-Propanediol

Figure 2i illustrates the adsorption of 1,2-propanediol (CH2OH–CHOH–CH3) binding to Pt(111) via the hydroxyl group attached to the middle C atom. The Pt–O distance is 2.20 Å. This middle hydroxyl group is also pointing towards the terminal hydroxyl group, which is in turn pointing down towards the surface at a distance of 3.46 Å from the Pt atom below. 1,3-propanediol (CH2OH–CH2–CH2OH) binds via one of the terminal hydroxyl groups, as illustrated in Fig. 1ii. The Pt–O distance is 2.19 Å. This hydroxyl group is pointing at the other terminal hydroxyl group, forming a hydrogen bond-like structure. The second hydroxyl group is pointing toward the surface with a Pt–O bond length of 3.36 Å.

3.2.2 Removal of 1H

CH2OH–CH–CH2OH (Fig. 2iii) binds to the surface at a top site via the middle C atom. The Pt–C bond length is 2.11 Å. Both C–C bonds are 1.52 Å, similar to the length of a C–C single bond in the gas phase, indicating that the gas phase valency of the middle C atom (four) is approximately maintained in the adsorbed configuration. Both hydroxyl groups are located above the Pt top sites, with one hydroxyl group pointing at the other (again resembling a hydrogen bond-like structure). The Pt–O distances are 2.22 and 3.14 Å, respectively.

CH2OH–CHOH–CH2 (Fig. 2iv) binds via the terminal –CH2 group at a top site. The Pt–C distance is 2.09 Å. The C–C bond lengths are 1.51 and 1.54 Å, respectively, between –CH2 and –CHOH– and between –CHOH– and –CH2OH groups, again suggesting that the gas phase valency of the binding terminal C atom is approximately satisfied in its adsorbed configuration. The hydroxyl group at the terminal C is above the Pt top site at a distance of 2.21 Å; it is also pointing at the middle (downward-pointing) hydroxyl group, thus forming a hydrogen bond-like structure.

3.2.3 Removal of 2H

CH2OH–C–CH2OH (Fig. 2v) binds at the bridge site via the middle C atom. Both Pt–C bond lengths for this central C atom are 2.09 Å. The C–C bond lengths are 1.53 Å (upper bond in figure) and 1.52 Å (lower bond in figure), respectively. The lower hydroxyl group is above the Pt atom with a Pt–O distance of 2.24 Å. The upper hydroxyl group is above a threefold site, pointing downward to the surface. No intramolecular hydrogen bonds are formed in this intermediate.

3.2.4 Removal of 3H

The terminal –CH group in CHOH–CHOH–CH (Fig. 2vi) adsorbs at a bridge site. The Pt–C bond lengths are 2.07 Å (upper Pt atom in figure) and 2.06 Å (lower Pt atom in figure), respectively. The C–C bond length between –CH and –CHOH– is 1.51 Å. In addition, CHOH–CHOH–CH also binds via the terminal –CHOH group at a top site. The Pt–C bond length is 2.13 Å. The C–C bond length between the two –CHOH– groups is 1.54 Å, showing that the gas phase valencies for both unsaturated C atoms are approximately maintained on the surface. The terminal hydroxyl group is pointing at the middle hydroxyl group (pointing downward), forming a hydrogen bond-like structure.

3.2.5 Removal of 4H

CH2OH–COH–C (Fig. 2vii) binds mainly in a tilted position via the terminal C at an fcc site. The Pt–C bond lengths are 1.99 Å (between adsorbed C and the Pt atom to the upper right), 2.00 Å (between adsorbed C and the lower Pt atom), and 2.10 Å (between adsorbed C and the Pt atom to the upper left), respectively. The C–C bond length between terminal C and the –COH– group is 1.42 Å, suggesting that the bond has partial double-bond character, and the distance between the carbon atoms in –COH and –CH2OH is 1.52 Å. The distance between the C in the –COH– group and Pt is 2.36 Å. The hydroxyl group in –COH– is pointing downward, while the hydroxyl group in –CH2OH is above the top site of Pt at a distance of 2.53 Å. No intramolecular hydrogen bond-like structure is observed.

3.2.6 Removal of 5H

CHOH–COH–C (Fig. 2viii) binds via the terminal C at an fcc site. The Pt–C bond lengths are 2.03 Å (with the Pt atom to the upper left), 2.02 Å (with the lower Pt atom), and 2.06 Å (with the Pt atom to the upper right), respectively. The C–C bond lengths are 1.43 and 1.38 Å between the terminal C and –COH– and between the C atoms in –COH– and –CHOH, respectively. The shortened bond lengths indicate that these bonds have at least partial double bond character, consistent with the planar geometry. The middle hydroxyl group in –COH– is pointing downward, while the hydroxyl in the terminal –CHOH group is pointing towards the middle hydroxyl group.

3.2.7 Removal of 6H

COH–C–COH (Fig. 2ix) binds via the middle C at a bridge site, while the C atoms in the terminal –COH groups also adsorb near bridge sites. COH–C–COH shows a slightly bent structure. The Pt–C bond lengths are 2.16 and 2.02 Å between the middle C and the Pt atoms above and below this C atom in the figure, respectively. The Pt–C bond lengths are 2.55 Å (left Pt) and 1.98 Å (right Pt) for the C atom in the rightmost –COH group. The Pt–C bond lengths are 2.46 Å (upper Pt) and 1.99 Å (lower Pt) for the C atom in the leftmost –COH group. Both C–C bond lengths are 1.43 Å. Both hydroxyl group are pointing downward, and no internal hydrogen bond-like structures are observed.

COH–COH–C (Fig. 2x) binds via the terminal C atom in COH–COH–C at an fcc site. The molecule has a tilted planar geometry, with the Pt–C bond lengths being 1.99 Å (with the Pt atom to the upper left), 2.12 Å (with the Pt atom to the upper right) and 1.98 Å (with the lower Pt atom), respectively. The other terminal C in –COH adsorbs near a bridge site with associated Pt–C bond lengths of 2.59 Å (with the upper Pt atom) and 1.97 Å (with the lower Pt atom), respectively. The middle C is near an off-top site with a C–Pt bond length of 2.34 Å. The C–C bond lengths are 1.44 and 1.47 Å, respectively. The middle hydroxyl group is pointing towards the terminal hydroxyl group.

3.2.8 Removal of 8H

No intermediates corresponding to the removal of seven hydrogen atoms from glycerol are predicted to be relevant to the C–O bond scission events described below in our free energy diagram. Therefore, their geometries are not reported in this section. Two relevant intermediates correspond to removal of eight hydrogen atoms, however. CO–C–CO (Fig. 2xi) binds via the middle C at a bridge site with two terminal C atoms above nearby threefold sites. The molecule has a planar geometry perpendicular to the surface. The Pt–C bond lengths of the middle C atom are 2.06 Å (with the Pt atom to the left) and 2.07 Å (with the Pt atom to the right), respectively. Both terminal C atoms are located at the off-top sites with Pt–C distances of 2.09 Å. The C–C bond lengths are 1.48 Å (upper) and 1.47 Å (lower), respectively.

CO–CO-C (Fig. 2xii) binds via the terminal C atom adsorbed at an fcc site. This molecule shows a planar geometry perpendicular to the surface. The Pt–C bond lengths are 2.03 Å (with the Pt atom to the lower right), 2.01 Å (with the Pt atom to the lower left), and 2.01 Å (with the upper Pt atom). The terminal C in the carbonyl group is near the off-top site with a Pt–C distance of 2.05 Å. The C–C bond lengths are 1.48 Å (between C and –CO–) and 1.64 Å (between carbonyl groups), respectively.

As a general rule, the configurations of these intermediates seem to follow the rules established in many previous computational studies for the adsorption of oxygenated hydrocarbons [28, 29, 32, 54, 55] on Pt(111). On Pt(111), it is found that molecules prefer to bind in such a way that both C and O atoms can approximately maintain their gas phase valencies. This adsorption behavior, which roughly conserves gas phase bond orders, is one reason that our bond order-motivated empirical correlations (Eqs. 2, 4) are able to reliably predict the binding energies of such species. We note, however, that other factors, such as intramolecular strain or highly delocalized bonding, may dictate that other configurations are more favorable in certain cases. Finally, intramolecular hydrogen bonding plays an important role in many oxygenated intermediates by providing additional stabilization of the molecules.

3.3 Energy Barriers of Elementary Reactions

While it would be prohibitively expensive to perform explicit transition state searches for all elementary reactions in the glycerol decomposition network, BEP plots are effective means of estimating reaction barriers and transition state energies from the knowledge of reaction thermodynamics alone. These BEP relationships are very efficient although it is important to keep in mind that their usage can often be associated with error bars of a few tenths of an eV in transition state energies, particularly where chemistries involving highly complex reaction intermediates are involved. We have previously introduced a BEP relationship for C–C bond scission in glycerol dehydrogenation intermediates on Pt(111) [32]. In this work (see Supporting Information Fig. S1), we demonstrate that a very similar relationship describes combined C–C and C–O bond scission kinetics. Such a result is fully consistent with previous literature that introduced common BEP relationships for C–C and C–O scission in ethanol dehydrogenation intermediates on platinum [27]. The energies of final states and transition states are referenced to the energies of the gas phase reactants for the elementary step in question where the surface reactions are written in the exothermic direction. The C–O bond scission activation barriers used in developing this BEP (referenced to the best configuration of the adsorbed reactant of the appropriate elementary step) are reported in Table S3 of the Supporting Information. The slope and intercept are 1.03 and 1.45 (in eV), respectively. The statistical uncertainty associated with the slope is 0.03, and the value is 0.14 for the intercept. The standard error is 0.28 eV.

3.4 Free Energy Diagram

Free energy changes for dehydrogenation and C–C bond breaking during glycerol decomposition have been previously determined at conditions of 483 K (which is on the lower end of the temperature range for biomass reforming experiments) and standard pressure [7, 18, 32, 56]. In the present work, we extend these results to include the possibility of a single C–O bond scission event at any point in the glycerol dehydrogenation pathway; this analysis thus gives a first estimate of selectivity between C–C and C–O bond scission under reforming conditions. The thermochemistry of glycerol dehydrogenation is determined with a couple of assumptions, with the enthalpy changes approximated by the values determined directly from DFT calculations; the neglect of zero-point energy corrections in this analysis will have only very modest effects on the calculated energetics [28, 29]. For entropy changes, we assume that gas phase glycerol loses its translational entropy when adsorbing and that H2 gains its translational entropy when desorbing (effectively, we assume that adsorbed hydrogen atoms are in a quasi-equilibrated state with the gas phase H2 molecules). We do not directly include desorption of products resulting from C–C or C–O bond scission in this entropy analysis; since these bond breaking events are surface-localized reactions, they are assumed to have no entropy change. The entropy correction of H2 in the gas phase at the above conditions is 0.64 eV. The corresponding value for glycerol is 0.88 eV.

The resulting free energy diagram is presented in Fig. 3. The most thermochemically stable intermediates at each level of dehydrogenation are shown with black squares. The blue diamonds indicate the lowest energy transition states for dehydrogenation, including either C–H or O–H bond scission, referenced to gas phase glycerol, a clean Pt(111) slab, and a stoichiometrically appropriate amount of gas phase H2 molecules, at each stage in the decomposition process (we note that these transition states do not necessarily correspond to bond scission in the most thermochemically stable dehydrogenation intermediates). The red triangles denote the corresponding lowest energy transition states for C–C bond scission at each stage of decomposition, and the purple circles show the lowest transition state energies for C–O scission reactions. The lowest energy transition states are estimated from the BEP relationships (in some cases, two transition states estimated by the BEP relationship have extremely close energies—within two decimal places in eV—in which case both are reported and indicated. Otherwise, we only show the lowest energy transition state). The thermochemistry of the adsorbed intermediates involved in the C–O bond scissions shown in the free energy diagram has been explicitly calculated with DFT calculations, and the DFT-determined values (not the values determined from our correlation scheme) are used as the input to the BEP relationships. For the lowest C–O bond scission transition state energy at a given level of dehydrogenation, if the lowest energy BEP transition state has also been determined with full DFT–NEB calculations, the corresponding NEB value is reported (represented by filled symbols in Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs11244-012-9806-2/MediaObjects/11244_2012_9806_Fig3_HTML.gif
Fig. 3

Free energy diagram for glycerol decomposition on Pt(111) at 483 K and standard pressure. Black squares represent the adsorption thermochemistry of the most stable glycerol dehydrogenation intermediates, blue diamonds represent the most stable dehydrogenation transition states at each level of dehydrogenation, red triangles represent the corresponding energetics for C–C bond breaking transition states, and purple circles represent the lowest energy transition states for C–O bond scissions. Hollowsymbols represent BEP estimates, while solid symbols represent DFT–NEB calculations where available (see text for further details). Gas phase glycerol and the clean platinum surface are used as the reference state. Bold letters indicate the particular bonds that are broken

4 Discussion

The free energy diagram, summarized in Fig. 3 and described briefly above, provides a simple and concise representation of several important aspects of the glycerol decomposition thermochemistry and kinetics. The diagram shows the most stable dehydrogenation intermediates in the reaction network at each level of glycerol dehydrogenation, together with the corresponding lowest energy C–H/O–H, C–C, and C–O scission transition states. These energetic data provide useful insights into the reaction network and are a practical alternative to a complete microkinetic analysis.

The low energies of the dehydrogenation transition states, depicted by blue diamonds in the figure, suggest that dehydrogenation may be quasi-equilibrated on Pt(111) at moderate to high temperatures [27]. This result further implies that the effective activation barriers and relative rates of C–C and C–O bond scission will be determined by the lowest transition state energies shown in Fig. 3. Early in the dehydrogenation network, the transition state energies for both C–C and C–O scission are quite high, at least 0.5 eV higher than the corresponding transition state free energies for dehydrogenation. The C–O transition state energies are somewhat lower than the corresponding C–C transition state values. After three or four hydrogen atoms have been removed from glycerol, however, the effective barriers for C–C scission decrease considerably compared to the corresponding values earlier in the reaction network, while the C–O bond scission energies have decreased only modestly, as shown in Fig. 3. Since this is the point in the reaction network where the lowest C–C/C–O bond breaking transition state energies are found, it has an important impact on the overall rate and selectivity patterns on Pt(111). The lower transition state energies associated with C–C bond breaking at this point, in turn, suggest that this surface will be selective to C–C scission, although we note that some alkanes could still form due to C–O bond scission in more highly decomposed (C1 or C2) products. These conclusions are in broad agreement with the experimentally determined tendency of platinum surfaces to be selective to C–C bond breaking under reforming conditions [7, 20], and they provide a compact explanation for this observed selectivity.

The divergence between the C–C and C–O scission transition state energies at intermediate levels of glycerol dehydrogenation, described above, has an important impact on the selectivity of glycerol decomposition on Pt(111). It is believed that this divergence, in turn, can be related to the stability of carbon monoxide on Pt(111). As shown in Fig. 3, most C–C bond breaking reactions after the removal of at least three hydrogen atoms involve the production of adsorbed CO, a species that is known to be highly stable on Pt(111) [26, 28], via decarbonylation. This stability of products, in turn, translates into lower transition state energies for C–C bond breaking. These highly stable product states involving carbon monoxide are not available for C–O bond breaking reactions on Pt(111), thus resulting in less favorable transition state energies for this class of reactions. Based on this analysis, we tentatively suggest that metals that promote decarbonylation may, in general, be more selective to C–C than to C–O bond activation.

It is important to emphasize that our analysis of the C–C/C–O bond scission selectivity is based only on the thermodynamics and kinetics obtained on an idealized pure metal surface. A plethora of studies, however, have suggested that, on technical catalysts, C–O bond scission may be catalyzed by a second metal [19, 5659], by non-metal co-catalysts or supports [56, 60, 61], and by partially oxidized base metals [56]. It is therefore likely that observed selectivity patterns on such technical catalysts would not be identical to what is predicted here. Even on pure platinum catalysts, however, it is possible that defects or adsorbed oxidic moieties in aqueous phase environments might promote C–O scission over C–C scission; the detailed analysis of such effects requires further study. Finally, we note that an explicit microkinetic analysis, involving the hundreds of elementary reaction steps in this reaction network and the impact of site blocking by adsorbed carbon monoxide products, would provide a more precise prediction of the selectivity patterns of glycerol decomposition on Pt(111) than the free energy analysis that we have presented above. However, such a detailed analysis would be unlikely to change the important conclusions of this work, which constitutes a first step in the elucidation of selectivity patterns for complex biomolecule decomposition and reforming processes on smooth platinum surfaces. Further, the methodological approaches introduced in this analysis are, as described above, applicable to other alcohols and polyols, opening up the possibility that similar analyses could be used to understand the chemistry of other biomolecules on platinum catalysts. Finally, by extending these analyses to other transition metals, it might ultimately be possible to design catalysts with enhanced activity and selectivity for biomolecule reforming.

5 Conclusions

Glycerol decomposition via a combination of dehydrogenation, C–C bond scission, and C–O bond scission reactions has been examined on Pt(111) with periodic DFT calculations. Building upon a previous study that analyzed C–C bond breaking in glycerol on this surface, we introduce a bond order conservation-motivated correlation for reaction intermediates resulting from C–O bond breaking in glycerol. Combining these results with a BEP relationship, we present a comprehensive free energy diagram for glycerol decomposition. The results suggest that high activity will only occur after several hydrogen atoms have been removed from glycerol, and at this point in the reaction network, C–C scission transition state energies are considerably lower than are the corresponding C–O transition state energies. This result is traced to the possibility of decarbonylation via C–C scission at intermediate levels of dehydrogenation in the reaction network; this process is energetically favored on Pt(111) due to the stability of carbon monoxide. By extending these correlations and principles to other transition metal surfaces, it may ultimately be possible to suggest improved catalysts for hydrogen production from glycerol.

Acknowledgments

This study was supported as part of the Institute for Atom-efficient Chemical Transformations (IACT), an Energy Frontier Research Center funded by the US Department of Energy, Office of Science, Office of Basic Energy Sciences. Use of the Center for Nanoscale Materials (CNM) is supported by the Office of Science of the US Department of Energy under contract no. DE-AC02-06CH11357. We acknowledge grants of computer time from EMSL, a national scientific user facility located at Pacific Northwest National Laboratory, and the Argonne Laboratory Computing Resource Center (LCRC). This research used resources of the National Energy Research Scientific Computing Center, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231.

Supplementary material

11244_2012_9806_MOESM1_ESM.docx (435 kb)
Supplementary material 1 (DOCX 435 kb)

Copyright information

© Springer Science+Business Media (outside USA) 2012