Dual Metal Site Fe Single Atom Catalyst with Improved Stability in Acidic Conditions

Abstract

Dual atom catalysts (DACs) not only retain uniform active sites and high atomic utilization efficiency as the single atom catalysts, but the two adjacent metal sites also cooperate and play a synergistic role to achieve additional benefits. However, the relationships connecting their dual-site synergistic effects on catalytic performance are not well rationalized due to limited pairs available from experiments. Herein, Fe/M dual sites supported by nitrogen doped carbon (Fe/M-N-C whereby M from 3 d–5 d electron containing transition metals) have been screened as an oxygen reduction reaction (ORR) catalyst. The results show that the absorption strength of ORR intermediates on four nitrogen coordinated metals is weaker than the three coordinated metals, which promotes favourable ORR activities. As a result, we recommended FeIr, FeRh, FeRu and FeOs as promising ORR catalysts. Ab initio molecular dynamic (AIMD) simulations suggest Fe/M-N-C (M = Ir, Rh, Ru and Os) catalysts with encouraging structural stability at room temperature. Furthermore, it is found that the nitrogen atoms in-between metals are vulnerable sites for proton attacking, yet the protonation process demands high energy, even under O₂ atmosphere, which underlines good tolerance under acidic conditions. This work provides a broad understanding of Fe based catalyst and a new direction for catalytic design.

1. Introduction

Advanced energy conversion and storage systems have received considerable attention due to their significant roles in the field of clean energy [1,2,3]. Among various clean energy technologies, fuel-cells are considered as promising candidates due to their high energy density and eco-friendliness [4,5]. However, slow kinetics of oxygen reduction reaction (ORR) limits the discharge power density of the device, and it is extremely challenging to overcome this obstacle [6,7]. Till now, noble metal-based materials are selected as the commercial electrocatalysts, dominating a large percentage of the fuel-cell market price, which demands the development of cost-effective and high performance non-precious metal catalysts.

Single atom catalyst (SAC) has been proposed to achieve 100% atomic utilization for precious metal [8], and recent studies have mentioned the upsurge of the atomically dispersed transition metal nitrogen doped carbon materials (known as M-N-C, wherein M stands for transition metals), derivatives from the biological systems which have shown great promise in the field of ORR catalysis [9,10,11,12,13,14,15,16,17,18,19,20]. In particular, well-dispersed Fe-N-C is enabled with competitive catalytic activity as commercial Pt-based catalysts [21,22,23,24,25]. However, Fe-N-C catalyst lacks long-term stability in acid medium, which tends to be involved in various side-reactions, hindering its industrial application in fuel cells.

Dual atom catalysts (DACs), an extension of SACs, are receiving increasing attention due to additional benefits rooted from electronic structure modulation [26,27,28,29,30,31,32,33,34]. For example, Fe/Co-N-C [26] was first synthesized by an Li group, which exhibited improved catalytic performance compared with its SAC counterparts. Following closely, it was used as a prototype to achieve bifunctional catalysis for both ORR and oxygen evolution energy (OER) [27]. Recently, Fe/Mn-N-C [28] was further demonstrated with excellent ORR performance in both alkaline and acidic media, and the key factors are ascribed to the spin related electronic structure adjustment. A unique half metallic quantum state was reasonably designed as in Fe/Zn-N-C [29] catalyst, which outperformed in both catalytic activity and long-life stability, providing new insights for the development of high-performance electrocatalysts. However, the reaction mechanism is not well rationalized in Fe based DACs due to limited pairs available from experiments, not to mention the structure evolution during working conditions.

In this study, based on systematic density functional theory (DFT) calculations, 13 Fe/M pairs of two distinct configurations are investigated as Fe/M-N-C OERR catalysts, with M covering from 4 d to 6 d transition metals (Cr, Mn, Co, Ni, Cu, Zn, Mo, Ru, Rh, W, Re, Os, and Ir). In addition to the Fe/Co, Fe/Mn, Fe/Ni, Fe/Zn and Fe/Cu which are available from experiments, four new candidates, including Fe/Ir (η = 0.26 V), Fe/Rh (η = 0.30 V), Fe/Ru (η = 0.42 V) and Fe/Os (η = 0.57 V) have been found with satisfied ORR activity, which are thermodynamically stable at 300 K from dynamic simulation. Attractively, the simulation of protonation process gives rise to high free energy demanded for surface hydrogenation, demonstrating improved structure stability in acid conditions.

2. Results and Discussion

Schematic models for FeMN₆(I) and FeMN₆(II) are shown in Figure 1. In FeMN₆(I), two N atoms are shared by metal sites, forming four N coordinated metal centers, yet, in FeMN₆(II), each metal is three coordinated with N atoms. In total, 13 metal elements are paired with Fe, yet the configurations of FeReN₆(II) and FeWN₆(II) are not capable in stabilizing *OOH (Table S1), which is thus beyond the following study for the 4e reaction pathway.

The optimized structures, along with the key reaction intermediates (*O₂, *OOH, *O and *OH), are depicted in Table S1. For the first H⁺+e step of ORR, the capture of O₂ molecules is analyzed by the O₂ adsorption energy, as observed in Figure 2 (Table S2). According to Sabatier’s law [35], the ORR catalyst requires moderate adsorption energy for O₂, as strong adsorption will cause the catalyst surface inhabitant by O₂, but weak binding will make it difficult to initiate the catalytic reaction. A favorable O₂ adsorption energy is estimated to be between −0.6~−0.7 eV. [27] Active centers which meet this precondition include FeCoN₆(I), FeNiN₆(I), FeRhN₆(I), FeOsN₆(I), FeRuN₆(I), and FeIrN₆(I), with ∆E_ads of −0.63 eV, −0.79 eV, −0.67 eV, −0.78 eV, −0.82 eV, and −0.78 eV, respectively. For a specific metal pair, FeMN₆(II) configuration exhibits stronger O₂ absorption ability as evidenced by larger adsorption energy. The calculated ΔE_ads values for FeMN₆(II) are between −1.5 eV and −4.0 eV, and thus FeMN₆(II) is not conducive to a smooth ORR. Here, it is found that the O-O bond is elongated from 1.23 Å in gas-phase to XX in the absorbant as shown in Figure 2, demonstrating that the O₂ is activated. In addition, the elongation of O-O bond is more pronounced in FeMN₆(II) than that in FeMN₆(I) because of the stronger O₂ adsorption in the former. From the results of Bader Charge (Table S3), it is seen that after the *O₂ binding on the metal sites, the oxidation states of Fe are enhanced. This is understandable because the O₂ extracts electrons from metal sites.

It is aware that the adsorption sites of O₂ and the adsorption configurations are different (Figure 3). For most FeMN₆(I), the *O₂ is absorbed on the metal with side-on configuration, where one O interacts with one metal site while the other O points away from the surface (although *O₂ adopts end-on configuration in Cr case, other intermediates interact with Cr site, instead of both metals). The following reaction steps proceed with the terminal O atom. The key intermediates (*OH, *OOH, *O) tend to bind to the M site, when M locates on the left of Fe in the period table (M = Cr, Mn, Mo, Ru, W. Re, Os); otherwise, Fe is preferred. By contrast, the two oxygen atoms in the O₂ molecular can interact with both Fe and M atoms in FeMN₆(II) (except FeMo couple), so do the *O and *OH species, which well explains the stronger interaction between adsorbates and catalysts.

ORR catalytic activity was assessed by the free energy changes of the key oxygen-containing intermediates, i.e., *OOH, *O and *OH, (Table S4 and Figure S1). The standard equilibrium potential for reaction steps R1–R4 are calculated from the free energy value, as shown in Figure 4 for FeMN₆(I) and Figure S2 for FeMN₆(II). The ideal reaction free energy value for the elementary step is 1.23. Since each intermediate plays dual roles, namely, the product of the former elementary reaction and the reactant of the following elementary reaction, the variation from the ideal value should be small. In other words, the closer to the standard equilibrium potential value, the better the catalytic activity. It is witnessed that, for FeMN₆(I) with FeCo, FeMn, FeCu, FeZn, FeRu, FeRh, FeOs and FeIr pairs, the ΔG values are close to each other, suggesting good activity. Unfortunately, a huge difference in standard equilibrium potential for the elementary steps can be found for all considered FeMN₆(II) configurations (Figure S2), excluding potential candidates with satisfied performance.

A close inspection of Figure 4 finds out that the lowest ΔG values are in general for the *OH→H₂O step, which corresponds to the desorption of the *OH intermediates. This suggests that the formation of the second H₂O is the potential determining step (PDS). Therefore, a volcano relationship between the ΔG_*OH and the negative catalytic overpotential (−η) is plotted in Figure 5. It is seen that all FeMN₆(II) configurations are located on the left side of the volcano, underlining strong adsorption of *OH, which hampers the further protonation of *OH. On the contrary, most FeMN₆(I) configuration sites on the right side are relative to those of FeMN₆(II), illustrating alleviate *OH binding. As a result, FeIrN₆(I) (η = 0.26 V), FeRhN₆(I) (η = 0.30 V), FeRuN₆(I) (η = 0.42 V), FeCoN₆(I) (η = 0.34 V) and FeZnN₆(I) (η = 0.41 V) give smaller overpotential than that of commercial Pt/C (η = 0.45 V) and other Fe based dual site counterparts (Table S5). In addition, the overpotential of FeOsN₆(I) (η = 0.57 V), FeCuN₆(I) (η = 0.48 V) and FeMnN₆(I) (η = 0.54 V) is also small, holding great promise for high activity reaction centers.

Among these reaction centers, the combinations of FeCo, FeMn, FeCu and FeZn are available from experiments, which exhibit excellent ORR performance, in accordance with our work. However, for FeRuN₆(I), FeRhN₆(I), FeIrN₆(I) and FeOsN₆(I), which are excluded from current experiments, AIMD simulation is performed to address their stabilities (Figure S3). It is seen that the total energy oscillates in a small range within 10 ps at 300 K, indicating solid stability kinetically. Moreover, the active center and carbon skeleton of the catalyst are remained as in the initial configuration (inset of Figure S3), another piece of evidence for the high stability.

Besides reaction activity, the structure stability under working condition is also an important criterion for a good catalyst. When operating in the cathode of a fuel cell, the catalyst is exposed under acid conditions, the attacking of H⁺ in the reaction center will cause the structure degradation and the change of the electronic structure. Therefore, the protonation is simulated for the screened FeRuN₆(I), FeRhN₆(I), FeIrN₆(I) and FeOsN₆(I) reaction centers with one and two protons adsorbed on the surface. It is revealed that the shared N sites are favorable for binding *H (Figure 6 and Table S6), yet the positive free energy change demonstrates that the protonation process is endothermic. Moreover, compared with the first *H binding, the energy demanded for another *H is higher, indicating good tolerance in acid environments.

Furthermore, since the structure of the reaction center is maintained after protonation and the interaction with one *H is flexible in the presence of O₂, a free energy diagram of ORR is simulated and compared with those of bare catalysts (Figure 7 and Table S7). It is shown that a small negative influence appears for the FeRh and FeIr pairs, and the overpotential becomes 0.54 V and 0.52 V, respectively. Attractively, the overpotentials for FeRuN₆(I) and FeOsN₆(I) become even lower than the bare metal centers, namely 0.35 V and 0.38 V, respectively, suggesting appealing acid performance.

3. Methodology

All the spin-polarized calculations were carried out by using the Vienna ab-initio simulation package (VASP) [36,37]. The Perdue–Burke–Enzerhof (PBE) [38] function and the projection-enhanced wave (PAW) [39] were used to describe the exchange-correlation potential and ion-electron interactions, respectively. The kinetic energy cut-off was set as 500 eV. A 5 × 5 supercell of graphene was used as the substrate for the diatomic centers. To avoid inter-layer interactions, a vacuum as large as 20 Å is set above the graphene slab. The dispersion correction for Van der Waals (vdW) [40] interactions was performed by the DFT-D2 [41] method. For structural optimization, a Γ centered 3 × 3 × 1 Monkhorst-Pack k-point sampling was used in the first Brillouin zone, which was increased to 7 × 7 × 1 for the electronic properties’ calculations. The energy and force tolerance are 1 × 10⁻⁵ eV and 0.02 eV Å⁻¹, respectively [42,43]. For the assessment of kinetic stability, ab initio molecular dynamics (AIMD) [44] simulations were used, and the simulation time lasts 10 ps at 300 K.

The adsorption energy (ΔE_ads) was calculated as follows:

ΔE_ads = E_{adsorbate+support} − (E_adsorbate + E_support)

where E_{adsorbate+support} is the total energy of the catalyst with the adsorbed intermediates, and E_adsorbate and E_support are the energy of the catalyst and the free adsorbate, respectively. The change of the Gibbs free energy (ΔG) was calculated via the computational hydrogen electrode (CHE) model [35]:

ΔG = ΔE_ads + ΔE_ZPE − TΔS

where ΔE_ZPE and ΔS are the correction of zero-point energy and entropy, respectively. The free energy of O₂ (4.92 eV) was extracted from the reaction O₂ + 2H₂ = 2H₂O. The key elementary step for a 4e ORR pathway is as follows:

(R1) *O₂ + H⁺ + e → *OOH ΔG₁ = ΔG_*OOH − 4.92

(R2) *OOH + H⁺ + e → *O + H₂O ΔG₂ = ΔG_*O − ΔG_*OOH

(R3) *O + H⁺ + e → *OH ΔG₃ = ΔG_*OH − ΔG_*O

(R4) *OH + H⁺ + e → H₂O ΔG₄ = −ΔG_*OH

Among them, the * denotes that the ORR species is adsorbed on the catalyst surface. The thermodynamic overpotential η [27,45] is thus determined by

η = (1.23 − G/e) V wherein G = min {|ΔG₁|, |ΔG₂|, |ΔG₃|, |ΔG₄|}.

In conclusion, a set of Fe-based dual metal site reaction centers supported by carbon have been studied using DFT calculations. Besides FeCo, FeMn, FeCu and FeZn pairs, which are available from experiments, FeIr (η = 0.26 V), FeRh (η = 0.30 V), FeRu (η = 0.42 V) and FeOs (η = 0.57 V) pairs are predicted with desirable ORR activity, which outperform even in an acid environment. AIMD simulations demonstrated their good structural stability at room temperature, providing new research directions for further catalytic design.