Compared to significant amount of experimental efforts spent on catalytic BESR for surface reaction mechansim investigation, computational approach at molecular level still remains barely untouched in the past several decades probably due to its extreme complicacy and limited computation resources. However, recent years have witnessed the rapid development of computational technology, making the reaction simulation at catalyst surface technically feasible. For simplifying simulation work, many publications have purely focused on the ethanol or water alone adsorption and associated decomposition on single metal clusters [135-139].

Various methodologies have been developed to reasonably represent catalyst surface for obtaining more accurate simulation results. The slab geometry in contrast to cluster model is widely adopted to model the catalyst surface with certain thickness. In addition to the top atomic layer, several successive layers below are also included to simulate the bulk effect on the surface layer. The surface layer is thereafter allowed to be reconstructed in response to the constraint from bulk layers. Usually, a vacuum region with certain length is created right above the top layer of the slab model to prevent the interaction of adsorbed molecules with its periodic images [140]. The choice of supercell size comes from the compromise between computation accuracy and computation time span. "Nudged Elastic Band (NEB)" method [141, 142] is proven by many papers to be effective in transition state and associated energy barrier estimation and very useful in minimum energy pathway determination especially for complex chemical reactions. Most of recently published computational results are based on the self-consistent periodic density functional theory (DFT) calculation, which is more accurate than other commonly used computational methods such as ab initio, semi­empirical, and empirical methods.

According to the published papers, although there are some disagreements on the ethanol decomposition on model catalyst surface, the proposed pathways can still be generally classified into two routes. One is CH3CH2OH ^ CH3CH2O(a) ^ CH2CH2O(a) ^ CH2CHO(a) ^ CH2CO(a) ^ CH2(a)+CO(a)+4H(a) ^ CH4(g)+CO(g)+H2(g). In this route, ethanol molecule first prefers to adsorb at atop sites and binds to the surface through the oxygen atom after O-H bond cleavage, followed by a six-membered ring of an oxametallacyclic compound formation through the elimination of the hydrogen atom attached to the P-carbon. This six — membered ring is usually located at the interface of active metal and support, creating a bridge between them. The ethanol decomposition process then continues with two consecutive eliminations of hydrogen atom attached to a-carbon. Scission of C-C bond then occurs under the facilitation of active metal, resulting in the formations of a series of adsorbates which subsequently desorb from substrate at elevated temperature to yield final gas products such as CH4, CO, and H2 [142-144]. The other suggested route follows the track of CH3CH2OH ^ CHsCHOH(a) ^ CHsCHO(a) ^ CHsCO(a) ^ CH2CO(a) ^ CHCO(a) ^

CH(a)+CO(a) ^ CH4(g)+CO(g)+H2(g)+C(s) [145].

Unlike ethanol decomposition, water dissociation completes only in two steps (i. e., H2O ^ H(a)+OH(a) ^ 2H(a)+O(a)), which is obviously due to its rather simple formulation. Compared to the second O-H bond breakage, the first one can take place with much lower activation energy [146]. Therefore, it can be easily predicted that hydroxyl group will have much higher chance to participate in BESR for ethanol oxidation than O* after water complete dissociation.

After a careful literature review, it is worth noting that the role of catalyst support and co­adsorption of ethanol and water are barely considered, which is probably attributed to its awful computational complicacy. In order to give a clear picture of what is really happening on catalyst surface during BESR and provide a theoretical support to our experimental observations and proposed reaction mechanism, we launched a computational task in collaboration with the Chemistry Department at Ohio State University. We employed plane — wave periodic DFT method implemented in the Vienna ab initio simulation program (VASP) to investigate the ethanol steam reforming reactions [147-149]. The projector augmented wave (PAW) method [150, 151], combined with a plane-wave basis set, was utilized to describe the core and valence electrons. The generalized gradient approximation (GGA) [152] of Perdew and Wang (PW91) [153] was applied for the exchange-correlation functional. The convergence of the plane-wave expansion was obtained with moderate truncation energy of 500 eV, while the electronic relaxation was converged to a tolerance of 1 x 10-4 eV. The Monkhorst-Pack grid [154] served in the generation of the k-points, and a (4 x 4 x 1) k-point grid was used for Brillouin zone sampling for surface calculations. Spin polarization was applied in all calculations.

The relaxed bulk structure of CeO2 with a lattice parameter of 5.46 A was used to construct the slab model. The CeO2 (111) and Co/CeO2 (111) surfaces were modeled as 2 x 1 super cells. A three molecular CeO2 thick slab model was constructed, thus nine atomic layers in total. The super cell has dimensions: a = 7.72 A, b = 6.69 A, and c = 23.88 A, and a 16 A thick vacuum region is included to ensure that there is no interaction between the surface adsorbates of one layer and the next slab. To optimize the surface structure, the top three atomic layers of the slab with the adsorbates were allowed to relax. The bottom six atomic layers were fixed at the bulk positions of ceria. The NEB method [155-157] was employed to locate the transition states of various reactions over the catalyst surface. After numerical differentiation, each transition state was confirmed to have a single imaginary vibrational frequency.

Ethanol decomposition via steam reforming reaction was computationally studied on the CeO2(111) and Co/ CeO2(111) surfaces. From our results, the most likely reaction pathway is described below. The decomposition of ethanol starts with the breaking of the O-H bond on the catalyst surface. The produced ethoxide unit prefers to be adsorbed on the catalyst surface by the Oe-Co interaction. With the assistance of a surface-bound hydroxyl moiety, derived from water dissociation, the Ca-H bond breaking of the ethoxide unit could proceed to yield the thermodynamically stable product (adsorbed acetaldehyde and hydrogen atom). The surface-bound hydroxyl group could act as a better hydrogen acceptor to assist the Ca — H bond-breaking reaction as compared to the surface oxygen atom of ceria. In the subsequent step, the surface-bound hydroxyl addition to acetaldehyde produces the hydroxyl adduct, СНзСН(0)(0Н), as an intermediate. This СНзСН(0)(0Н) intermediate further undergoes the loss of H from the Ca position to generate acetic acid. Acetic acid can then lose the acidic hydrogen from the hydroxyl unit, yielding an adsorbed acetate and hydrogen. The acetate could be further converted to the CH2(OH)COO intermediate via H — atom abstraction and subsequent surface-bound hydroxyl addition reactions. As suggested by the calculations, the Ca-Cp bond rupture from the chemisorbed CH2(O)COO intermediate generates formaldehyde and CO2. Similar to acetaldehyde, the generated formaldehyde could react with a surface-bound hydroxyl group to produce the HCH(O)(OH) adduct that subsequently undergoes a H-atom abstraction reaction to yield formic acid. Then, formic acid loses the acidic hydrogen of the hydroxyl unit to generate surface-bound formate. Finally, formate could be converted to CO2. Throughout the favorable reaction pathway from ethanol to CO2, one of the most energetically costly steps on the potential energy surface is the Cp-H bond-breaking step of acetate for ethanol decomposition with the participation of surface-bound hydroxyl groups on the Co/CeO2(111) surface.

Our modeling indicates that surface-bound hydroxyl groups, which is formed from water dissociation, plays two critical roles in the ethanol steam reforming reaction. The first is to assist the hydrogen-abstraction reactions from carbon atoms. The second is their involvement in addition reactions to form the C=O or C=C double bond intermediates. Thus, a catalyst on which water could more effectively dissociate to form surface-bound hydroxyl and hydrogen might be a potentially better catalyst for steam reforming reactions. On the Co/CeO2(111) surface, our computational work elucidates the formation of acetaldehyde and acetate intermediates and is consistent with extant experimental observations [133]. The present computational studies do not account for the generation of acetone, carbon monoxide, and methane, which are byproducts observed in experimental studies. A model that includes larger Co particles with some surface-bound hydroxyl groups would be more realistic and may account for the formation of other byproducts.