On the Structure Sensitivity of Formic Acid Decomposition on Cu Catalysts
Catalytic decomposition of formic acid (HCOOH) has attracted substantial attention since HCOOH is a major by-product in biomass reforming, a promising hydrogen carrier, and also a potential low temperature fuel cell feed. Despite the abundance of experimental studies for vapor-phase HCOOH decomposition on Cu catalysts, the reaction mechanism and its structure sensitivity is still under debate. In this work, self-consistent, periodic density functional theory calculations were performed on three model surfaces of copper—Cu(111), Cu(100) and Cu(211), and both the HCOO (formate)-mediated and COOH (carboxyl)-mediated pathways were investigated for HCOOH decomposition. The energetics of both pathways suggest that the HCOO-mediated route is more favorable than the COOH-mediated route on all three surfaces, and that HCOOH decomposition proceeds through two consecutive dehydrogenation steps via the HCOO intermediate followed by the recombinative desorption of H2. On all three surfaces, HCOO dehydrogenation is the likely rate determining step since it has the highest transition state energy and also the highest activation energy among the three catalytic steps in the HCOO pathway. The reaction is structure sensitive on Cu catalysts since the examined three Cu facets have dramatically different binding strengths for the key intermediate HCOO and varied barriers for the likely rate determining step—HCOO dehydrogenation. Cu(100) and Cu(211) bind HCOO much more strongly than Cu(111), and they are also characterized by potential energy surfaces that are lower in energy than that for the Cu(111) facet. Coadsorbed HCOO and H represents the most stable state along the reaction coordinate, indicating that, under reaction conditions, there might be a substantial surface coverage of the HCOO intermediate, especially at under-coordinated step, corner or defect sites. Therefore, under reaction conditions, HCOOH decomposition is predicted to occur most readily on the terrace sites of Cu nanoparticles. As a result, we hereby present an example of a fundamentally structure-sensitive reaction, which may present itself as structure-insensitive in typical varied particle-size experiments.
KeywordsHCOOH decomposition Structure sensitivity DFT Catalysis Copper
Decomposition of formic acid (HCOOH) is an important process since HCOOH is a by-product produced in large quantities in the formation of levulinic acid from hydroxyl-methyl-furfural (HMF) during the production of fuels from biomass resources . The in situ catalytic dehydrogenation of HCOOH may provide hydrogen, required in hydrogenating levulinic acid to gamma-valerolactone (GVL) . GVL is a valuable platform molecule with potential of producing both renewable fuels and chemicals. Also, HCOOH may be used directly as a fuel in direct formic acid fuel cells [3, 4, 5, 6] or as a hydrogen carrier [7, 8, 9] with a closed carbon cycle for other hydrogenation reactions.
The vapor-phase decomposition of HCOOH has been widely used to test the catalytic properties of various metals [10, 11, 12, 13], metal oxides [14, 15, 16] and alloys since the 1960s as HCOOH is one of the simplest organic molecules and carboxylic acids. HCOOH can decompose on metal surfaces through dehydrogenation and/or dehydration processes: dehydrogenation leads to carbon dioxide (CO2) and hydrogen (H2) products, while dehydration leads to production of water (H2O) and carbon monoxide (CO), the latter being a poisoning species for noble metal catalysts such as Pt and Pd. The two decomposition pathways could be linked by a water gas shift (WGS) reaction. Cu catalysts have been reported to selectively decompose HCOOH via dehydrogenation to CO2 and H2 , accompanied by lack of CO in the products, which suggests that there is no WGS involvement in this chemistry. Further, temperature programmed reaction studies of HCOOH on various single crystal surfaces of Cu clearly rule out the occurrence of the WGS reaction during HCOOH decomposition on Cu catalysts [18, 19, 20, 21]. Spectroscopic studies of this reaction on Cu catalysts show the presence of a stable formate (HCOO) intermediate  on the catalyst surface that results from dissociation of the acidic hydrogen from HCOOH. Formate has also been proposed as a surface intermediate in several key catalytic reactions, including the WGS reaction [22, 23] and methanol synthesis .
Despite the large amount of experimental work concerning HCOOH decomposition on Cu catalysts, there is still an ongoing debate on the reaction mechanism and its structure sensitivity. Iglesia and Boudart  measured similar activation energies (94.5–100 kJ/mol) for HCOOH decomposition on Cu catalysts supported by various materials. These values are also similar to those measured on different preferentially oriented and polycrystalline Cu catalysts by other researchers [25, 26, 27]. Based on these findings, together with the observed insensitivity of the turnover rates to the particle size and to the nature of the catalyst support, they argued that HCOOH decomposition on Cu was structure insensitive. In contrast, Nakamura and coworkers  claimed that HCOOH decomposition on Cu catalysts was structure sensitive with dramatically different activation energies being observed for dehydrogenation of HCOO on Cu(111) and Cu(110), although formate formation from CO2 was structure insensitive when comparing these two Cu surfaces. Hu and Boyd  also found a strong dependence of the adsorption energy of HCOO on the surface structure of Cu; the adsorption energy of HCOO decreases in the order Cu(110) > Cu(100) > Cu(111). Bowker and other researchers [20, 30] reported that HCOOH deprotonates to HCOO on the clean Cu(110) and Cu(100) surfaces, but not on the clean Cu(111) surface where the presence of atomic oxygen is critical for formate to be formed.
A first-principles investigation of vapor-phase HCOOH decomposition on different Cu facets can help to elucidate the reaction mechanism as well as to obtain a better understanding of the behavior of Cu catalysts in terms of activity and structure sensitivity. In contrast to the abundant experimental studies, there is a lack of systematic theoretical studies for HCOOH decomposition on Cu surfaces. However, several elementary steps involved in the process are common to methanol synthesis and WGS reactions and have been investigated theoretically. In terms of shared reaction intermediates, prior work combining density functional theory (DFT) calculations, experimental kinetic studies, and microkinetic modeling showed that HCOO a formed on the surface of Cu  and Pt  catalysts during the WGS reaction but this HCOO intermediate is only a spectator species, whereas the reaction proceeds via a carboxyl (COOH)-mediated pathway. COOH is also a possible intermediate in the course of the HCOOH decomposition reaction through dissociation of the carbonic hydrogen from HCOOH. In a previous study of CO/CO2 hydrogenation to methanol on Cu, Grabow and Mavrikakis  found that HCOOH is a relevant reactive intermediate and it prefers to decompose through the HCOO-mediated route on Cu(111). In this work, density functional theory (DFT) calculations were performed on two additional Cu model surfaces [Cu(100) and Cu(211)], exploring both COOH- and HCOO-mediated pathways and comparing with the previous results on Cu(111) to elucidate the reaction mechanism and its structure sensitivity nature on these Cu surfaces.
2 Computational Methods
Periodic, self-consistent density functional theory calculations were performed with PW91-GGA [32, 33] exchange correlation functional using the DACAPO [34, 35] total energy code. Results on Cu(111) surface are taken from a previous work . The methods used are restated here for clarity. The Cu(111) surface was modeled by a three-layer slab using a (3 × 3) unit cell, repeated in a super cell geometry with five equivalent layers of vacuum (~10.6 Å) between two successive metal slabs. Since the surface relaxation effects have been shown to be negligible for similar systems [36, 37], Cu atoms on Cu(111) surface were fixed in their bulk truncated positions during the calculation. The Cu(100) surface was modeled by a four-layer slab using a (3 × 3) unit cell with 11.0 Å of vacuum spacing separating the periodic slabs and the optimization was carried out by relaxing the top two layers. The Cu(211) slab was constructed using a (1 × 3) unit cell and consisted of nine Cu layers (having a terrace three atoms deep and three atoms wide). Successive slabs were separated by a vacuum equivalent to 12 Cu layers (~12.1 Å) and the top four layers of the slab were allowed to relax during the calculation. The surface Brillouin zone was sampled at 54 Chadi-Cohen  k-points for Cu(111) and using a (4 × 4 × 1) Monkhorst–Pack grid  for Cu(100) and Cu(211) surfaces. Ultrasoft Vanderbilt pseudopotentials  were utilized to describe core electron interactions, and the Kohn–Sham one-electron valence states were expanded on the basis of plane waves with kinetic energy below 25 Ry. The electron density was determined by iterative diagonalization of the Kohn–Sham Hamiltonian, Fermi population of the Kohn–Sham states (kBT = 0.1 eV), and the Pulay mixing of the resulting electronic density.  All total energies were extrapolated to kBT = 0 eV. Adsorption was permitted on only one of the two exposed surfaces, and the electrostatic potential  was adjusted accordingly. Structures were fully relaxed until the Hellmann–Feynman forces acting on the atoms were smaller than 0.05 eV/Å. We calculated a lattice constant of 3.66 Å for bulk Cu, in good agreement with the experimental value 3.61 Å .
Binding energy (BE) of a surface intermediate was calculated with respect to the clean slab and the corresponding adsorbate (intermediate) in the gas phase. All the BE values, activation energies and heats of reactions in this work were reported with zero point energy (ZPE) corrections that were calculated by assuming a quantum harmonic oscillator possessing the calculated vibrational frequencies. The vibrational frequencies were calculated by numerical differentiation of forces using a second-order finite difference approach with a step-size of 0.015 Å.  Binding energies, activation energies and reaction energies without ZPE corrections can be found in the Supplementary Material. All activation energy barriers and reaction energies reported were relative to the reactant and product states at infinite separation. Minimum energy paths and activation energy barriers for all elementary steps were calculated using the climbing image nudged elastic band method (CI-NEB) . The minimum energy path for each elementary step was discretized by at least seven images, including the initial and final state. Convergence of the NEB calculations was reached when the magnitude of the force on all images was less than 0.1 eV/Å, except for the hydrogen recombination reaction, where a stricter convergence criterion (0.05 eV/Å) was used. The transition states (TS) were confirmed by vibrational frequency calculations yielding a single imaginary frequency along the reaction coordinate.
3 Results and Discussion
ZPE-corrected binding energies (BEs) on clean Cu(111), Cu(100) and Cu(211) facets
ZPE-corrected reaction energies (∆E) and activation energies (Ea) on clean Cu(111), Cu(100) and Cu(211) facets
HCOOH(g) + * ↔ HCOOH*
HCOOH* + 2* ↔ HCOO** + H*
HCOOH* + 2* ↔ COOH** + H*
HCOO** ↔ CO2* + H*
COOH** ↔ CO2* + H*
2H* ↔ H2(g) + 2*
CO2* ↔ CO2(g) + *
3.1 Structure and Adsorption Energetics of Reaction Intermediates
Carboxyl binds to two Cu atoms on the Cu(111) surface through its C and O atoms with the C–O bond almost parallel to the surface and O–H bond pointing towards the surface; the BE is calculated to be −1.44 eV. A similar adsorption configuration is found on the Cu(211) surface at the step edge. On Cu(100), however, COOH binds to two opposite bridge sites in a square encompassing a hollow site. COOH binds more strongly on Cu(100) and Cu(211) than on Cu(111) by 0.30 and 0.46 eV, respectively. On all three surfaces, formate binds in a bidentate configuration with both of its O atoms bound to adjacent top sites; step-edge top sites are preferred for HCOO on Cu(211). The calculated BEs of HCOO are −2.57, −2.95, and −3.17 eV on Cu(111), Cu(100), and Cu(211), respectively. The most stable configuration for formic acid binds to the top site on all three surfaces through its O atom with the C–H bond pointing away from the surface, and the O–H bond pointing towards the surface. On Cu(211), HCOOH binds to the top site on the step edge and the O–H bond is pointing towards the step foot atoms on the terrace. The binding strength of HCOOH increases in the order Cu(111) < Cu(100) < Cu(211), with BE values of −0.16, −0.32, and −0.47 eV on the three Cu surfaces respectively.
With the exception of CO2, which is physisorbed on all three surfaces, all intermediates prefer to bind to sites on the step edge of Cu(211). The magnitudes of the BEs of reaction intermediates on the three Cu surfaces are observed to decrease in the order Cu(211) > Cu(100) > Cu(111), in agreement with the general notion that adsorbates tend to exhibit stronger binding to more open facets.
3.2 Elementary Steps: Reaction Energies and Activation Energies
3.2.1 HCOOH* + 2* → HCOO** + H*
HCOOH decomposition to HCOO proceeds from the most stable configuration of HCOOH with O–H bond-breaking taking place over a bridge site on Cu(111) and Cu(100), and an off-top site on the step edge of the Cu(211) surface. This reaction is exothermic on all three Cu surfaces with reaction energies −0.33, −0.57 and −0.74 eV, increasing in magnitude in the order Cu(111) < Cu(100) < Cu(211), due to the much stronger binding of reaction product HCOO on the more open facets. Similar activation energies in the range of 0.41–0.48 eV, however, are found on all three Cu surfaces, indicating a rather weak structure-sensitivity of the O–H activation in HCOOH on Cu, further evidenced by the similar configurations of the transition states on the three surfaces.
3.2.2 HCOOH* + 2* → COOH** + H*
C–H bond-breaking in HCOOH, yielding adsorbed COOH and atomic hydrogen, involves a rotation of the reactant molecule such that the C-H bond is pointing towards the surface. At the transition state, C–H bond scission occurs over a top site where the product COOH readily adsorbs through its C atom after the reaction and the produced H atom adsorbs on the nearest fcc or bridge sites on Cu(111) and Cu(100), respectively. On the Cu(211) surface, the C-H bond is broken over the top site on the step edge; after the reaction, COOH adsorbs on two adjacent top sites of the step edge while H adsorbs on the hollow site at the step. In contrast to the HCOOH decomposition to HCOO, COOH formation from HCOOH is an endothermic reaction on all three surfaces, presumably because of the higher energy of adsorbed COOH than adsorbed HCOO. The activation energy of this reaction step varies slightly across the three surfaces, being 1.13, 0.94 and 1.06 eV on Cu(111), Cu(100) and Cu(211), respectively. As COOH formation has a much higher (more than 0.5 eV higher) activation energy than HCOO formation from HCOOH, Cu will preferentially break the O–H bond in HCOOH to form HCOO on all three surfaces.
3.2.3 HCOO** → CO2* + H*
Similar to the C–H scission in HCOOH, C-H bond scission in HCOO also involves a rotation of the molecule such that the C–H bond points towards the surface. One Cu–O bond is already broken at the transition state, with the C-H bond-breaking taking place over a bridge site on Cu(111) and Cu(100), and the bridge site on the step of Cu(211). This reaction is endothermic on all three surfaces with reaction energies being 0.15, 0.52 and 0.60 eV on Cu(111), Cu(100) and Cu(211), respectively. Both reaction energy (all, endothermic) and activation energy increase in the order Cu(111) < Cu(100) < Cu(211), indicating that this is a structure sensitive reaction on Cu surfaces. There is an increase of ~0.26 eV in the activation energy of HCOO decomposition each time we go from Cu(111) (0.95 eV) to Cu(100) (1.21 eV) and to Cu(211) (1.47 eV), presumably because of the stronger binding of HCOO, the reactant, on the more open facets. The structure-sensitive character found for this reaction step is in agreement with Nakamura’s findings that HCOO dehydrogenation has notably different activation energies on Cu(111) and Cu(110) . On Cu(111), the activation energy of HCOO decomposition measured by Nakamura is 108 kJ/mol  (~1.12 eV), comparable to our calculated value 0.95 eV. The activation energy of this step is much higher than that of the HCOO formation, indicating HCOO decomposition is probably the rate limiting step in the HCOO-mediated pathway for HCOOH decomposition.
3.2.4 COOH** → CO2* + H*
As the O–H bond in COOH points towards the surface in its lowest-energy configuration, COOH decomposition proceeds directly from the most stable configuration with the O–H bond-breaking taking place over the bridge site of all three Cu surfaces. This reaction is exothermic, and activation energy barriers are slightly different on the three surfaces, with values of 0.98, 1.12 and 1.08 eV on Cu(111), Cu(100) and Cu(211), respectively. The similar activation energies of this reaction on the three Cu surfaces are probably due to the similar transition states seen on these surfaces, as shown in Fig. 2. COOH decomposition has comparable barriers as HCOO decomposition on Cu(111) and Cu(100) surfaces while it is much easier than HCOO decomposition on Cu(211), 1.08 eV versus 1.47 eV, respectively.
3.2.5 H* + H* → H2(g) + 2*
H2 recombinative desorption starts with two H atoms adsorbed on adjacent fcc and hcp sites on the Cu(111) surface with H–H bond formation taking place over the bridge site, while the reaction starts with two H atoms adsorbed on two adjacent bridge sites on Cu(100) and Cu(211) with H–H bond formation over the off-top and top sites, respectively. Cu(111) and Cu(100) have similar activation energies, 0.74 and 0.72 eV, and reaction energies (0.23 and 0.25 eV on Cu(111) and Cu(100), respectively) for H2 recombination, while both energy values are higher on the Cu(211) surface (Ea = 0.97 eV; ∆E = 0.47 eV). This difference is entirely due to the difference in the binding strength of H, the reactant, on different surfaces, as the activation energy of the reverse reaction (H2 dissociation) is practically the same on all three surfaces at ca. 0.5 eV.
3.3 Potential Energy Surfaces (PES)
The PES of the HCOO-mediated pathway decreases in energy in the order Cu(111) > Cu(100) > Cu(211). For the HCOO-mediated pathway, Cu(100) and Cu(211) show similar TS energies (within ca. 0.2 eV) which are lower than that on Cu(111). Cu(211) binds intermediates more strongly than the Cu(100) surface, and accordingly shows lower energies for the adsorbed states and subsequently higher activation energy barriers than those on Cu(100). Since HCOO decomposition is likely the rate limiting step among the three catalytic steps, and Cu(211) has a similar TS energy for HCOO decomposition as on Cu(100) but much higher activation energy, it is likely that Cu(211) surface is less active than Cu(100) for HCOOH decomposition. On the PES, coadsorbed HCOO and H represents the most stable state, and the subsequent decomposition reaction of HCOO has the highest activation energy among the three catalytic steps. Therefore, these surfaces are likely to be partially covered by HCOO, especially on the Cu(100) and Cu(211) surfaces, which bind HCOO more strongly than Cu(111) by 0.38 and 0.60 eV, respectively. As a result, HCOOH decomposition may take place more readily on the terraces of the Cu catalysts, while the under-coordinated step sites and defect sites may be blocked by strongly adsorbed HCOO. Assuming that HCOO dehydrogenation is the rate limiting step and HCOO is the most abundant surface intermediate, a microkinetic analysis provided in the Supplementary Information rationalizes the rate expression which confirms the lower contributions from the less coordinated step/defect sites of Cu catalysts. The Cu catalysts used in Iglesia and Boudart’s study  contain relatively big nanoparticles with sizes 5–40 nm, within which range a small fraction of under-coordinated sites is expected and is potentially occupied by the HCOO intermediate; hence, the activity of Cu catalysts might come from the unoccupied terrace sites entirely. Accordingly, similar activation energies were observed for different particle-size catalysts. Their findings cannot differentiate the contributions from different types of sites (such as terrace, step, corner, defect sites) on Cu. To evaluate the contribution of under-coordinated sites to this chemistry, experiments should be conducted on Cu single crystals or smaller copper nanoparticles, where perimeter/corner/defect sites are present at a substantial percentage. It would also be necessary to substantiate these predictions from the DFT calculations using an experimentally validated full microkinetic model.
HCOOH decomposition was investigated using self-consistent, periodic DFT calculations on three Cu model surfaces—Cu(111), Cu(100) and Cu(211)—by exploring both HCOO-mediated and COOH-mediated routes. It is found that the HCOO-mediated route is energetically more favorable than the COOH-mediated route on all three Cu surfaces. For the HCOO-mediated route, HCOO decomposition is likely to be the rate determining step due to the fact that it has the highest TS energy as well as the highest activation energy barrier among the three catalytic steps. Cu(100) and Cu(211) show similar TS energies in the HCOO pathway, which are lower than those on Cu(111). While Cu(211) has similar TS energies as Cu(100), it binds the intermediates more strongly and has a higher activation energy for the likely rate limiting step (HCOO decomposition); hence, Cu(211) might be less active than the Cu(100) facet for HCOOH decomposition. Coadsorbed HCOO and H represents the most stable state on the PES, implying a surface partially covered with HCOO intermediate, especially at the under-coordinated step and defect sites. As a result, under realistic reaction conditions, HCOOH decomposition may be most facile on the terraces of the Cu catalysts and may show an invariance in turn-over-frequency as a function of nanoparticle size, as under-coordinated step/defect sites are likely poisoned by HCOO rendering them inactive. Yet, we showed here that HCOOH decomposition reaction on Cu surfaces is highly structure sensitive at the atomic scale level. We are currently exploring this reaction using a combined DFT, experiments and detailed microkinetic modeling approach to corroborate our tentative conclusions regarding the decreased, due to poisoning, importance of under-coordinated step or defect sites in HCOOH decomposition on Cu catalysts.
This work was supported by the U.S. Department of Energy (DOE)–Basic Energy Sciences (BES), Office of Chemical Sciences, grant DE-FG02-05ER15731. We thank Lars C. Grabow for performing the calculations on Cu(111), as reported originally in Ref.  and utilized here. Calculations were performed at supercomputing centers located at the Environmental Molecular Sciences Laboratory, which is sponsored by the DOE Office of Biological and Environmental Research at the Pacific Northwest National Laboratory; Center for Nanoscale Materials at Argonne National Laboratory, supported by DOE contract DE-AC02-06CH11357; and National Energy Research Scientific Computing Center, supported by DOE contract DE-AC02-05CH11231. We thank Anthony Plauck, Luke Roling and Dr. Srinivas Rangarajan for carefully proofreading this manuscript.
- 4.Liu C, Chen M, Du C, Zhang J, Yin G, Shi P, Sun Y (2012) Durability of ordered mesoporous carbon supported Pt particles as catalysts for direct formic acid fuel cells. Int J Electrochem Sci 7(11):10592–10606Google Scholar
- 11.Madix RJ (1980) Surface reaction modifiers—general overview. Abstracts of Papers of the American Chemical Society 180 (Aug): 26Google Scholar
- 25.Quinn DF, Taylor D (1965) Decomposition of formic acid and methanol on copper–nickel alloys. J Chem Soc 5248–5251Google Scholar