Following the fabrication of the first practical single-atom catalyst (SAC), Pt1/FeOx [1], the concept of SACs has been successfully extended to other FeOx supported metals, including Ir1/FeOx [2, 3], Au1/FeOx [2, 4], and Ni1/FeOx [5]. Recently, considerable progress has also been made in the investigation of other single-atom catalysts [2, 6-31]. These SACs possess well-defined surface-supported single-atom active sites, and exhibit high activity and stability in various heterogeneous reactions, including the water-gas shift (WGS), CO oxidation or preferential oxidation (PROX), and hydrogenation. Among these various reactions, CO oxidation is a simple yet important reaction, which has been extensively investigated from the perspective of environmental protection, exhaust purification for motor vehicles, gas purification for closed-cycle CO2 lasers, and for CO detectors [32, 33]. Moreover, this reaction is also an elementary step in the WGS reaction [2, 34], and the preferential oxidation of CO in H2 (i.e., the PROX reaction) serves as a key step for removing CO from reforming gas in fuel cell applications [35]. Over the past few decades, a number of oxide-supported noble-metals (e.g., Pt, Au, Ir, and Pd) [1, 19, 36-38] have been identified as active catalysts for CO oxidation. In particular, as for FeOx based SACs, the supported non-noble 3d-transition metal has been even found to be effective for the oxidation of CO, that is, our recent theoretical investigation indicated that FeOx supported 3d-transition metals such as Ni1/FeOx, can catalyze CO oxidation with high activity, even at room temperature [5].
While Fe is a typical 3d-transition metal and the Fe atoms in the FeOx support can also be considered to be a monodisperse metal center on the surface of the support. To be consistent with the categorization of Pt1/FeOx SAC, here we denote the FeOx substrate with an oxygen vacancy as Fe1/FeOx. However, this base material has been found to have a much lower activity for CO oxidation than that of Pt1/FeOx [1]. Our previous density functional theory (DFT) calculations showed that the high catalytic activity of Pt1/FeOx correlates with the partially vacant 5d orbitals of the positively charged, high-valence single-atom Pt1. This effect reduces the probability of CO poisoning and facilitates adsorption of oxygen. However, it remains unclear what factors give Pt1/FeOx such a high catalytic activity [1] and limit the activity of the FeOx support as a catalyst in CO oxidation. To understand the differences between these two catalysts, we performed theoretical investigations on possible catalytic reaction pathways of CO oxidation on Fe1/FeOx. We examined the chemical interactions using DFT, with our previous calculated results of Pt1/FeOx as a reference.
On the basis of our theoretical calculations, we show that among the fundamental differences in the catalytic activity toward CO oxidation between Pt1/FeOx and Fe1/FeOx, the regeneration of single-atom active centers is the key factor, which determines the catalytic activity for CO oxidation. These findings help to develop an understanding of the high activity of Pt1/FeOx for CO oxidation, which will guide the design of new FeOx supported single-atom catalysts for potential practical applications.
As used in our previous work [1], the (0001) surfaces of α-Fe2O3 were represented by a periodic slab model, constructed from bulk cell dimensions: a = b = 5.04 Å and c = 13.72 Å. Because α-Fe2O3 is antiferromagnetic and has an atomic magnetic moment on iron atoms, we used the primitive rhombohedral unit cell of Fe2O3 with a magnetic configuration (+ --+) to build the surface slab. This arrangement has previously been shown to be the most energetically favored magnetic configuration for α-Fe2O3 [39]. A vacuum distance 12 Å-width in the direction perpendicular to the surface was set to eliminate the interaction between layered nanostructures in adjacent cells. Considering the large relaxation typical of Fe2O3 surfaces [1, 40], we selected slabs containing 12 layers of Fe atoms, and 7 atomic layers of O3 (see below) to model the O3-terminated surfaces [1]. The 10 top-layer slabs of the surface were allowed to relax while the other layers beneath the surface were frozen during the geometry optimizations. On the basis of previous studies [1, 8], here the relatively favorable O3-terminated surface of Fe2O3(0001) was selected, where each Fe1 atom is coordinated by three surface oxygen atoms (i.e. the O3 atoms), with a third-layer of Fe atoms lying below the Fe1 atoms. The Fe1 can be viewed as a single-atom adsorbed to the α-Fe2O3 substrate.
The theoretical calculations were performed by DFT using the Vienna Ab-initio Simulation Package (VASP) [41, 42]. The core and valence electrons were represented by the projector augmented wave (PAW) [43] method and plane-wave basis functions with a kinetic energy cut-off of 400 eV [1]. Considering that both Pt and Fe have relativistic effects, we included the mass-velocity and Darwin relativistic effects through the PAW potentials. The generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) [44] exchange-correlation functional was used in the calculations. A Monkhorst-Pack grid of size of 3×3×1 was used to sample the surface Brillouin zone [1]. Ground-state atomic geometries were obtained by minimizing the forces on the atoms below 0.02 eV/Å. Because of the strong d-electron correlation effects for Fe, the calculations were performed with the DFT+U method, based on the formalism suggested by Liechtenstein and Dudarev et al. [45]. The parameters were set at U = 4 eV and J = 1 eV according to the previous study [1]. The transition states were obtained by relaxing the force below 0.05 eV/Å via the dimer method [46].
We begin our study using the previous calculated model of M1/FeOx (M = Pt, Ir, Au, Ni) [1, 3-5], which derives from the primitive structure of the α-Fe2O3 (0001) surface with an oxygen vacancy (Ovac) on its O-termination surface. The FeOx substrate in this model is similar to other M1 in M1/FeOx SACs with a single atom of Fe1 anchored to the surface, denoted as Fe1/FeOx. After intensive searching for adsorption and transition states, the predicted catalytic cycle and calculated reaction energies of CO oxidation on the α-Fe2O3 substrate are shown in Fig. 1 (top view). The corresponding bond lengths of the species are presented in Fig. 2 (side view) and Table 1. In the initial step of the reaction (step ⅰ), an oxygen vacancy (Ovac) near the Fe atoms is constructed to represent the reduction of the hematite surface, and the optimized bond lengths Fe1-OA1 and Fe1-OA2 are found to be 1.78 and 1.77 Å, respectively. Owing to the presence of the oxygen vacancy, O2 molecule prefers to dissociatively adsorb to the surface of FeOx.That is, the calculated adsorption energy of O2 dissociation (-3.75 eV) is much higher than molecule activation of O2 on the oxygen vacancy (-2.14 eV). With the dissociative adsorption of O2 on the surface (step ⅱ), the distance between the OB atom (light green) and OC atom (light green) is 2.70 Å (as seen Fig. 2), which indicates the formation of O adatoms on the surface. After this adsorption of O2, the first CO molecule is co-adsorbed on the single-atom Fe1(step ⅲ) through the C atom (pink) of CO, with a Fe-C distance of 2.18 Å. The optimized distance of the C and OB atoms on the single-atom Fe1 is 2.36 Å. As the C atom of CO approaches OB to generate the first CO2 molecule (CO + OB → CO2), only a negligibly barrier energy (0.08 eV) is needed to be overcome through the first transition state (TS-1) with an imaginary vibration frequency at 345i cm-1. The activation barrier to the first formation of CO2 on the single-atom Fe1 is low enough for the reaction to occur at a relative low temperature. After climbing TS-1, the first CO2 molecule desorbs from the surface of FeOx, leaving behind the dissociated OC atom to heal over the Ovac on the FeOx surface (step ⅳ).
In addition to reactions of COad and Oad on Fe1 single active site through the Langmuir-Hinshelwood (L-H) mechanism, the Eley-Rideal (E-R) mechanism might also be involved in the catalytic process as suggested by our previous study on SAC Ir1/FeOx [3]. The E-R mechanism was also considered for formation of the first CO2on the Fe1/FeOx surface, owing to vertical adsorption of the dissociated OB atom on the single-atom Fe1. The calculated E-R reaction pathway and the selected distances of atoms in these species are presented in Fig. 3. The distance between the C atom of CO and the OB atom vertically adsorbed on the single-atom Fe1 is 3.33 Å, indicating only weak adsorption of the CO molecule in the initial reactant pattern (R), as shown in Fig. 3. By overcoming a barrier energy of 0.23 eV, the conversion of CO and OB to CO2 can be accomplished through a transition state (TS) on the surface of FeOx to give the final product (P). Thus, both the L-H and E-R mechanisms might be involved in the formation of the first CO2 molecule because the reaction activation energies of these two processes are low enough for the reaction of CO + OB → CO2 to occur at a low temperature.
After the formation of a perfect Fe1-terminated surface of α-Fe2O3 (0001) (step ⅳ), the second CO molecule can favorably adsorb on the single-atom Fe1 as illustrated in step ⅴ. The optimized adsorption energy of CO is -0.73 eV, and the distance between the C atom from CO and the OC atom dissociated from O2 is 3.11 Å, as shown in Figs. 1 and 2. The adsorbed CO can approach OC at Ovac, leading to a small barrier energy of 0.66 eV for the reaction of CO + OC → CO2 via the L-H mechanism on the surface of FeOx(TS-2). This transition state features an imaginary vibration frequency of 391i cm-1. The distance between the C atom and OC is shortened from 3.11 to 1.69 Å in TS-2, indicating the formation of a new C-O bond and the generation of a second CO2 analogue on the surface of FeOx; however, the CO2 molecule remains adsorbed to the FeOx surface (step ⅵ). Before the CO2 is released from the FeOx surface into the gas phase, an energy barrier of 1.09 eV must be overcome in TS-3. Finally, the whole catalytic cycle for CO oxidation is complete and the original catalyst with Ovac (step ⅰ) of Fe1/FeOx is regenerated.
The reaction rate-determining step over the whole catalytic cycle is clearly the desorption of the second CO2 molecule from the surface of FeOx (TS-3). Thus, the FeOx substrate without a supported single-atom Pt1 has a much lower catalytic activity than that of Pt1/FeOx SAC in CO oxidation reaction, which agrees well with previous experimental results [1]. When the standard enthalpies of formation of the oxides are compared, the lower catalytic activity of Fe1/FeOx for CO oxidation is also likely related to the relatively high oxygen affinity of Fe (250-300 kJ/mol) [47].
To determine why SAC Pt1/FeOx has a much higher catalytic activity than that of the substrate α-Fe2O3 (Fe1/FeOx), and elucidate the nature of the binding of the single-atom Pt1 to the FeOx support, we compared the mechanism of catalysis and electronic structures in the proposed reaction pathways of CO oxidation. The relative binding energies for CO oxidation over these two catalysts are presented in Table 1. As shown in Fig. 1 and Table 1, there are several differences between these two different catalysts. First, O2 is adsorbed in a molecular form onto the single Pt1mechanism, but is dissociatively adsorbed on the single-atom Fe1. Second, the difference in the mode of O2 activation leads to a discrepancy in the energy barriers for formation of the first CO2 molecules on Pt1 and Fe1 single-atom active sites. As shown in Figs. 1 and 2, the activation barrier for formation of the first CO2 molecule on single-atom Fe1 (0.07 eV, TS-1) is much lower than that on Pt1 (0.59 eV, TS-1) because the dissociated OB atom reacts with the adsorbed CO much more easily. Third, the activation barriers of the rate-determining step of these SACs for the second CO2 formation are clearly different for the two SACs M1/FeOx (M = Fe, Pt). The activation energy of COad + OC → CO2 (TS-3) on Pt1/FeOx (0.79 eV) is much lower than that of COad + OC → CO2 (TS-3) on Fe1/FeOx (1.09 eV), indicating that the α-Fe2O3 substrate has a low catalytic activity for CO oxidation.
As seen from the calculated Bader charges for M1 single-atoms in M1/FeOx (M = Fe, Pt) in Table 2, the valence electron loss of the Fe1 single-atom (1.488 |e|) is larger than that of a single-atom Pt1 species (1.045 |e|) after CO adsorption on these single atoms. This result indicates the different interaction strengths between the adsorbed CO molecules and the single atoms of Pt1 and Fe1. To further understand the chemical interactions between the CO molecule and Pt1 or Fe1 single-atoms, we analyzed the electronic properties of these two different surfaces before (step ⅳ) and after (step ⅴ) CO adsorption. The formation of the second CO2 molecule in these two steps governs the major differences in the process of CO oxidation. According to the density of states (DOS) distributions in Figs. 4 and 5, both the 3d and 5d states of the single Fe1 and Pt1atoms mix well with the 2p states of the neighboring oxygen atoms, suggesting strong chemical interactions between the single metal atoms with the FeOx support, and that the stability of Pt1/FeOx is equivalent to that of the FeOx substrate. The C 2p states of the adsorbed CO also mix with the d-states of single-atom metal species, indicating chemical bonding of CO on both SACs. As shown in Table 1, the calculated CO adsorption energy on Pt1/FeOx is larger than that on the FeOx substrate. From Fig. 2, the stronger CO adsorption on the Pt1 center makes it easier for the adsorbed CO to bind to the surface O atom in the OC…O(surface) of TS-3. Indeed, for the TS-3 structure the OC…O(surface) distance is 0.05 Å shorter for Pt1/FeOx than that for Fe1/FeOx, which accounts for the lower activation barrier of CO oxidation on Pt1/FeOx.
Finally, we use a (MeO)3M←CO (M = Fe, Pt) molecular system to model the bonding interactions between CO and the single metal atom at the surface. In this model, the M atom is clearly in the +3-oxidation state, while the real oxidation states of Pt and Fe in Pt1/FeOx and FeOx are likely to be somewhat different. However, this molecular model has a similar coordination environment to that of the surface Pt and Fe atoms, and can be used to provide insight into the orbital interactions involved at these types of active sites with similar local structure. The structures optimized by the PBE functional and TZ2P basis sets as implemented in ADF 2016 are shown in Fig. 6 and the bond parameters are given in Table 3. The orbital energy-level correlation diagram is shown in Fig. 7, with the corresponding molecular orbitals listed in Table 4. As can be seen from Fig. 7 and Table 4, the 5s orbital of CO is stabilized by a donor interaction with (MeO)3M, and the main interactions in the frontier MOs are back-donation to 2π orbitals of CO, which contribute to the HOMO-1 and LUMO+1 of (MeO)3Fe←CO as well as the HOMO-2 and LUMO of (MeO)3Pt←CO. The calculated energy-level splittings between the bonding Fe 3d or Pt5d and the anti-bonding orbitals CO (2π*) of (MeO)3M←CO (M = Fe, Pt) are 4.73 and 5.24 eV, respectively. Thus, the orbital interactions between 2π(CO) and (MeO)3Pt are stronger than those between 2p(CO) and (MeO)3Fe owing to the larger extent of the Pt 5d orbital than that of the Fe 3d. This effect cancels the lower energy of the 5d atomic orbitals (AOs) for Pt than that of the 3d orbitals for Fe. Therefore, the stronger adsorption of CO on the Pt single-atom active center and the higher activity of Pt1/FeOx for CO oxidation are related to differences in the orbital interactions between the various species and Pt 5d and Fe 3d orbitals.
We have performed systematic theoretical investigations using relativistic density functional theory (DFT) to develop an understanding of why single-atom Pt1 on FeOx possesses excellent catalytic activity for CO oxidation, but the FeOx substrate alone is not an effective catalyst. Our calculations indicate that the formation mechanisms of the first CO2 molecules on the surfaces of these SACs are quite different owing to the different activation of O2 at these two single-atom sites. Molecular adsorption occurs on the single-atom Pt1 while dissociative adsorption is unavoidable at the single-atom Fe1. More importantly, a relatively high activation energy (1.09 eV) is required to remove the O atom from the perfect (0001) surface of α-Fe2O3 through the formation of the second CO2 molecule, owing to the weak Fe(OC…Osurface) interaction. A relatively high oxygen affinity for Fe (250-300 kJ/mol) and the weak strength of CO adsorption on the single-atom Fe1 are the key factors contributing to the lower catalytic activity of Fe1/FeOx for CO oxidation. The present theoretical study offers an explanation for why the FeOx support is effectively inert in CO oxidation compared with the SAC Pt1/FeOx. These findings provide insights that will be useful for the design of new FeOx supported single-atom catalysts.
The calculations were done using supercomputers at Tsinghua National Laboratory for Information Science and Technology, the State Key Laboratory of Physical Chemistry of Solid Surfaces (Xiamen University), and Guizhou Provincial High-Performance Computing Center of Condensed Materials and Molecular Simulation.