Introduction

Currently, with the human demand for energy bigger and bigger, the traditional fossil energy cannot meet the demand of human society in the future energy [1, 2]. Therefore, water electrolysis [3, 4], metal air batteries [5, 6], fuel cells [7,8,9] and other sustainable energy production technologies are increasingly being people's attention. The electrocatalysts oxygen evolution reaction (OER) and oxygen reduction reaction (ORR) are the two key reaction of the current advanced energy conversion technology [10, 11]. It is well known that both reactions kinetics rate is slow, which limits the development of related energy conversion and storage technology [12]. So far, usual choice is given priority with noble metal or noble metal oxide catalysts to accelerate the reaction rate of OER and ORR. Among them, the most widely used catalysts are Pt for ORR and Ru or Ir oxides for OER [13,14,15,16]. However, the high prices and scarcity of these noble metals restrict their large-scale and wider industrial applications. Hence, development of low-cost and high catalytic activity noble metal-free catalysts has very important practical significance and urgency [5]. Previous research has shown that carbon-based materials, transition metal hydroxides, transition metal sulfides and composites composed of these materials all have good catalytic activity [17,18,19,20].

The deposition of transition metal (TM) atoms on the surface of materials to form single-atom catalysts (SACs) is a new frontier in catalytic research, which opens a new door for us to reveal the catalytic process on atomic scale [21,22,23,24,25]. In the past few years, a growing number of reports indicated that SACs can be widely used for CO oxidation [26, 27], fuel cells [5] and water splitting. SACs can ensure maximum atomic utilization and form more adjustable active sites. However, as the metal particle size reduces to a single atom, the increase in the surface free energy is likely to lead to metal aggregation [24]. Consequently, to keep SACs excellent catalytic performance, choosing reasonable metal atoms and selecting the appropriate substrate materials are very crucial to stabilize the isolated metal atom. In recent years, catalysts based on two-dimension (2D) materials have attracted widespread attention due to their high activity and a large number of reaction sites [28,29,30,31,32,33,34,35]. 2D materials as SACs substrate have been extensively studied, which has opened up a new frontier field for the search of efficient catalysts [36,37,38,39,40,41]. More importantly, 2D catalysts are potential bifunctional catalysts with both OER and ORR activities. For energy conversion technologies, bifunctional catalysts have higher catalytic performance than two single functional catalysts [42, 43]. In addition, compared with two separate single functional catalysts, bifunctional catalysts can reduce product costs due to less equipment usage and simpler production process.

The catalytic activity of single atom and substrate combinations requires a simple and efficient descriptor for design estimation [42, 44,45,46]. Nevertheless, a complete description of how SACs and their coordination environment affect OER and ORR activity remains unclear. In this regard, there is an urgent need to fully understand the catalytic activity trends of different single atom and substrate combinations, which can help guide the direction of adjusting the performance of SACs. Recently, B4N and B2O monolayers materials have been successfully predicted using first-principles calculation methods [47, 48], and a large number of theoretical studies have been carried out to study their applications as cathode materials for batteries and catalysts for hydrogen evolution reactions (HER) [49,50,51,52,53]. All previous studies indicated that B4N and B2O monolayers were two promising SACs substrates [47, 48]. Based on this, using 2D B2O and B4N as substrate materials, this paper utilizes density functional theory (DFT) to study the OER and ORR catalytic activities of TM atoms supporting on its surface, and to explore the mechanism of OER and ORR. We have studied the 3d and 4d group TM atoms, including Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Zn, Cu and Pd. The scale relationship between the Gibbs free energy (ΔG) of OER and ORR intermediate adsorption is used to describe the trend of catalytic activity. It is found that the d band center (εd) of the doped TM atoms is strongly correlated with the catalytic activity of OER and ORR. Finally, our calculated results suggest that Ni@B4N and Pd@B2O systems could be used as bifunctional electrocatalysts with excellent performance, and the overpotentials (η) of ORR and OER are even as low as 0.43/045 V and 0.47/0.56 V, respectively. The effect of single TM atom doping on ORR and OER catalytic activity of 2D materials is further investigated. Our study not only provides electrochemists with a promising bifunctional SACs for ORR and OER, which may stimulate the electrochemists to synthesize it experimentally, but also provides theoretical basis and more opportunities for the design of efficient bifunctional catalysts.

Computational methods

All the calculations in the present study were carried out on spin-polarized DFT using the Vienna ab initio simulation package (VASP) code [54, 55]. The projector augmented wave (PAW) pseudopotentials were used to treat the core electrons [56]. The Perdew–Burke–Ernzerhof (PBE) [57] functional of the generalized gradient approximation (GGA) [58] was used to describe the electronic exchange and correlation interactions. A cutoff energy was set as 500 eV for the plane-wave basis set, and the convergence criteria for energy and force were less than 10−5 eV and 10−2 eV/Å, respectively. Moreover, van der Waals (vdW) corrections were considered by using the dispersion correction of DFT-D3 method of Grimme [59, 60]. A vacuum distance was not less than 15 Å to avoid the interlayer interaction in the z-direction. The Brillouin zones were sampled with 3 × 3 × 1 Monkhorst–Pack meshes for B2O monolayer and 5 × 5 × 1 for B4N monolayer, respectively. The dynamic stability of B4N and B2O monolayers was verified by using the PHONOPY code [61] implemented in VASP using the density functional perturbation theory (DFPT) [62]. A more precise k-mesh of 7 × 7 × 1 and 3 × 3 × 1 was used for electronic structure calculations and ab initio molecular dynamics (AIMD), respectively. In AIMD simulations, integration of the equations of motion proceeds with time steps of 2.0 fs for different temperature ranges. Typical simulations ran for 3000 steps with the time scale about 6 ps. The isokinetic ensemble (NVT) was employed for the ions, where the ion temperature Ti was fixed using velocity scaling. For more computational details, please see the Supporting Information.

Results and discussion

As shown in Fig. 1a, for the atomic structure of 2D B2O, there are two B and one O atoms in a primitive unit, and this structure has the space group Cmmm (No. 65). The optimized B-B (1.709 Å), B-O (1.338 Å) bond lengths, the B-B-B angle (107.5°) and lattice parameters (a = b = 3.937 Å) obtained from our calculations are in accordance with the previous work [48, 51, 53]. For the 2D B4N atomic structure, the structure reveals Cmmm symmetry with orthogonal lattices. The optimized structural parameters are a = 2.97 Å, b = 10.70 Å, and B-N bond length is 1.334 Å (see Fig. 1b). These calculation results are consistent with the previous report [47, 50]. To evaluate the dynamic stability of B2O and B4N monolayers, we calculated their phonon spectra, as shown in Fig. 1c. The absence of negative frequencies confirmed the dynamic stability of B2O and B4N monolayers. The highest frequency of B2O and B4N monolayers is about 50 THz, indicating that B − O, B − B and B − N in B2O and B4N monolayers have strong interactions. The thermal stability of B2O and B4N monolayers was simulated by AIMD. Figure 1d shows that the mean value of the total potential energy during the AIMD simulation oscillates within a narrow range, and the overall configuration of B2O and B4N monolayers remains good after 6 ps AIMD simulation, which confirms their thermal stability. Then, the structural optimization of B2O and B4N monolayers is performed based on a 4 × 4 × 1 and 2 × 2 × 1 supercell.

Figure 1
figure 1

The top and side views of the optimized primitive cell a B2O b B4N. c Phonon band structure of the 2D B2O and B4N. d The fluctuation of total energy during AIMD simulations at 300 K. The insets show the structures of the B2O and B4N monolayers at 0 and 6 ps, respectively

Full size image

To explore the catalytic performance of the B2O and B4N monolayers, we study the active sites on the surface. According to the symmetry of the structure, there are three possible sites of TM atoms adsorbed on the monolayers of B2O and B4N. These possible sites are marked as top (T), bridge (B) and center sites, respectively, as shown in Fig. S2(a-b). The stability of TM atoms doping on B2O and B4N monolayers by analyzing the binding energy (Eb), thus finding the optimal active site. Finally, it can be found from Table S1 that after geometry optimization, only the center site of the hexagon has the lowest Eb, while the other sites have higher Eb. In addition, existing studies have shown that atomic defects are regarded as the active site of catalysis [53, 63], and both the B2O and B4N monolayers have the advantage of natural defect structure, so we finally select the defect center site to stabilize solid TM atoms, as shown in Fig. 2a and c. Irreversible detachment of the central transition metal atom from the catalyst significantly limits catalyst durability. To investigate the stability of TM@B4N and TM@B2O systems, we calculated their Eb of TM atoms supporting on the center site. The calculated Eb are shown in Fig. 2b and d, where all negative values of Eb indicate that the B4N and B2O monolayers stabilize these doped TM atoms against clustering. This may be because the boron atom forms a very strong coordination bond when it provides 2p electron pairs to the vacant d orbital of the TM atoms. In order to better understand the interaction between TM atoms and B4N and B2O monolayers, we also plot the charge density difference distribution. In both cases, it has been observed that the center TM atoms are exposed and depleted to the B2O and B4N monolayers, with the yellow and cyan surfaces showing electron accumulation and loss, respectively. The bonding between TM atoms and O or N atoms is more stable than that between TM atoms and B atom so that TM atoms are bound around O and N atoms, thus residing at the center site and reducing the aggregation between TM atoms. These results indicate that there is a strong interaction between TM atoms and B4N and B2O monolayers, and the structure is stable after doping TM atoms.

Figure 2
figure 2

a The top view of the optimal TM adsorption site on B4N monolayer and corresponding charge density difference profiles. b The Eb of TM@B4N systems. c The top view of the optimal TM adsorption site on B2O monolayer and corresponding charge density difference profiles. d The Eb of TM@B2O systems

Full size image

Different TM atoms anchoring on B4N and B2O monolayers have different electronic structures, which provides a new method for understanding their catalytic properties. We calculate the values of the densities of states of partial wave (PDOS) for d orbitals of different TM atoms anchored on the B4N and B2O monolayers, as shown in Figs. 3 and 4. All the TM@B4N and TM@B2O systems considered in this work exhibit metallic properties, indicating that all structures have good electrical conductivity. It is well known that the εd is used to analyze the descriptor of the strength of the interaction between adsorbates and substrates [64,65,66,67]. The large number of d electrons in TM atoms and low εd energy levels usually result in weaker interactions with OH *, O* and OOH* adsorbates [68]. This is because the interaction between TM atoms and the adsorbates occurs through their electron level hybridization and charge transfer from TM atoms to the adsorbates. The PDOS in the figure displays that as the number of d electrons in TM atoms increases, we can see the εd shift towards the lower energy relative to the Fermi level. The results indicate that the interaction strength of adsorbates on B4N and B2O monolayers presents the following trend: Sc > Ti > V > Cr > Mn > Fe > Co > Ni > Pd. Nørskov et al. believed that the interaction intensity of adsorption intermediates on the catalyst was too strong, and the active center of the catalyst would be poisoned [69]. However, if the interaction is too weak, it is difficult to activate the reactants or intermediates and initiate the catalytic process. Therefore, it is a challenge to identify promising materials with suitable adsorption energies of the reaction intermediates. In order to verify the above conclusions, the ΔG of OER and ORR reaction intermediates corresponding to the εd of TM@B4N and TM@B2O systems is calculated based on Eqs. (S23)-(S25), as shown in Fig. S3. One can find from Fig. S3 that there are good relationships between the εd and ∆GO*, ∆GOH* as well as ∆GOOH*. We can conclude that for TM atoms in the same row of the periodic table, an increase in the number of d electrons tends to weaken the adsorption free energy of adsorbates. Therefore, these results exhibit the negative correlation between the εd and ∆GO*, ∆GOH* and ∆GOOH*. Considering the adsorption properties of intermediates can determine the reaction activity, such linear relationship can help to design catalysts more efficiently. This phenomenon has also been observed in experiments [70] and other theoretical studies [42, 71]. Therefore, we can adjust the interaction strength to the optimal values of OER and ORR performance by changing the doped TM atoms. As shown above, the ΔG of the adsorbed states (OH*, O* and OOH*) on the catalyst determine the rate-determining steps and catalytic activity of OER and ORR. For the ideal catalyst, the total formation energy of ORR is 4.92 eV, and the ideal adsorption values of ΔGOH*, ΔGO* and ΔGOOH* are 1.23, 2.46 and 3.69 eV, respectively, and the energy distance between the two adjacent intermediate states is 1.23 eV (when U = 0 V). Therefore, OER and ORR reaction can occur at the thermodynamic limit, where the corresponding η value is zero. However, in reality, the energy distance between adjacent intermediate states is difficult to satisfy the condition of equality, resulting in limited reaction. The ηOER is determined by the maximum energy distance between two adjacent intermediate states, while the ηORR is determined by the minimum energy distance between two adjacent intermediate states. We calculate the ηOER and ηORR of all TM@B4N and TM@B2O systems as shown in Fig. 5. Our results demonstrate that the stable position of the intermediate on the catalyst is at the top of the TM atoms, and only Ni@B4N and Pd@B2O system has good catalytic activity of ORR and OER, which are two potential bifunctional catalysts. The ηOER of Ni@B4N system is 0.45 V, comparable to the calculated RuO2 catalyst (0.42 V) [72]. The ηORR of Pd@B2O system (0.56 V) is slightly higher than that of conventional RuO2 catalyst. The best catalyst for ORR is Ni@B4N system with the lowest ηORR value of 0.43 V, followed by Pd@B2O system (0.47 V). It is worth noting that their ηORR values are both lower than the reported value of the best catalyst Pt (0.48 V) [73].

Figure 3
figure 3

Calculated the partial density of states (PDOS) of the TM@B4N systems. The Fermi level is set at zero (pink dashed line) and the d band center (εd) and are marked by cyan dashed line

Full size image
Figure 4
figure 4

Calculated the partial density of states (PDOS) of the TM@B2O systems. The Fermi level is set at zero (pink dashed line) and the d band center (εd) and are marked by cyan dashed line

Full size image
Figure 5
figure 5

Summary of overpotentials of TM@B4N and TM@B2O systems for a OER and b ORR

Full size image

The relationship among the Gibbs free energy ΔGOH*, ΔGO* and ΔGOOH* of the intermediate can be used to analyze the OER and ORR trends of the catalyst. The scaling relationships of ΔGOH* and ΔGOOH*, ΔGOH* and ΔGO* in Ni@B4N and Pd@B2O systems are shown in Fig. 6. Obviously, for the Ni@B4N system, ΔGOOH* can be expressed as a function of ΔGOH* by the equation: ΔGOOH* = 0.67ΔGOH* + 2.75. The relationship between ΔGOOH* and ΔGOH* has a very high coefficient of certainty (R2 = 0.94), indicating a strong linear relationship. Similarly, for Pd@B2O system, there is a strong linear relationship between ΔGOOH* and ΔGOH*, which is expressed by the equation: ΔGOOH* = 0.61ΔGOH* + 1.84, and has a high determination coefficient (R2 = 0.77). The slope and constant separation in the relation between ΔGOOH* and ΔGOH* indicate that the intermediate OOH* and OH* form a single bond between oxygen and TM atoms. Then, through linear fitting, we can obtain the scaling relationship of ΔGOH* and ΔGO* for Ni@B4N system: ΔGO* = 1.19ΔGOH* + 1.29, and the R2 is 0.75. For the Pd@B2O system, the scaling relationship is ΔGO* = 0.94ΔGOH* + 0.97 with the corresponding R2 of 0.74. Note that the linear relationship between the ∆GOH* and ∆GO* is not as good as that of ∆GOH* and ∆GOOH* due to the different single metal–oxygen bond nature of the OH* and O*. Different oxidation states have different contributions to the binding strength of oxygen atom and TM atoms, which more or less leads to the deviation from the linear scaling between them. As stated in Eq. (S17) and (S22), the ηOER and ηORR are two function of four variables (ΔG1, ΔG2, ΔG3 and ΔG4). According to the scaling relationship of ΔG1 + ΔG2 + ΔG3 + ΔG4 = 4.92 eV, the ηOER and ηORR can be reduced to two descriptors. On the grounds of the definitions of Eqs. (S13-S16) and (S18-S21), the two descriptors ΔGOH* and ΔGO*—ΔGOH* are sufficient to describe the catalytic performance of OER and ORR. For the TM@B4N and TM@B2O systems, we drew 2D volcano maps. As seen from Fig. 7, it can be found that under the optimum condition (ΔG1 = ΔG2 = ΔG3 = 1.34 eV), limiting ηOER of TM@B4N and TM@B2O systems are as low as 0.39 and 0.41 V, respectively. The blue region of the 2D volcano map shows the highest active area. Among the systems considered, Ni@B4N and Pd@B2O systems are potential catalysts, with ηOER of 0.45 and 0.56 V, respectively. As shown in Fig. 8, for the Ni@B4N system, the fourth step (desorption of O2) are the rate-determining steps, while the second step (O* formation) is the rate-determining step for the Pd@B2O system. Figure 9 shows the trend of activity of ORR for TM@B4N and TM@B2O systems by using ηORR as a function of the independent descriptors of ΔGOH* and ΔGO*—ΔGOH*. Under the optimal conditions (ΔG1 = ΔG4 = -0.91 eV), the theoretical ηORR is as low as 0.39 V. The first step in the ORR pathway is that O2 molecule is hydrogenated to OOH* through coupled proton electron transfer. Then OOH* is attacked by an additional proton electron pair to form O* atom and H2O molecule. O* is then further hydrogenated to OH* before the second H2O molecule formed. The most suitable catalysts for ORR are Ni@B4N and Pd@B2O systems. The ηORR of Ni@B4N system is found to be below 0.43 V, and the rate-determining step is to release of adsorbed OH*, while the release of adsorbed OH* is the rate-determining step for the Pd@B2O system (ηORR = 0.47 V) (see Fig. 8). Obviously, Ni@B4N and Pd@B2O systems, as potential efficient bifunctional catalysts for OER and ORR, can catalyze both OER and ORR at the same active site. Calculated intermediates of ΔGOH*, ΔGOH* and ΔGOOH* for the other systems are summarized in Fig. S4 and Table S2-S3. For the other systems, the formation of O* and OOH* in the second and third steps is the most difficult because it shows the maximum energy range and therefore becomes the most likely rate-determining step. The corresponding ηOER of these systems is too high to effectively catalyze OER. As OER electrocatalysts in practical application, they not only need more energy input, but also cause the lag of kinetic rate. In addition, due to the large required energy and slow reaction kinetics, it is difficult to remove the reaction intermediate products, thus leading to the active center deactivation and making the catalyst poison. Our results reveal that Ni@B4N and Pd@B2O systems are highly efficient and stable OER/ORR bifunctional catalysts and provide useful guidance for screening metal active centers of SACs on 2D materials.

Figure 6
figure 6

Scaling relationships between the adsorption free energies of adsorbates of a TM@B2O and b TM@B4N systems. The dashed horizontal line shows the O2 initial state at 4.92 eV

Full size image
Figure 7
figure 7

Colored contour plots of OER activity volcanoes for a TM@B4N and b TM@B2O systems showing the ηOER as a function of adsorption Gibbs free energies of the intermediates. The color bar represents the value of η

Full size image
Figure 8
figure 8

Gibbs free energy diagrams during the OER and ORR reaction pathways of a Ni@B4N and b Pd@B2O systems at 0 V versus standard hydrogen electrode

Full size image
Figure 9
figure 9

Colored contour plots of ORR activity volcanoes for a TM@B4N and b TM@B2O systems showing the ηORR as a function of adsorption Gibbs free energies of the intermediates. The color bar represents the value of η

Full size image

Conclusions

In summary, we have systematically studied the stability and catalytic behavior of TM atoms doped at the center defect site on B2O and B4N monolayers based on the DFT method, aiming to search for potential bifunctional SACs for OER/ORR with high activity and stability. The results show that the adsorption strength of the intermediates on the active center of TM atoms is closely related to the εd. As the number of d electrons of doped TM atoms on the defect B2O and B4N monolayers increases, the εd moves to the lower energy level, which weakens the interaction between the intermediate and the active center of TM atoms. Therefore, the ideal OER and ORR catalysts can be selected by adjusting the doped TM elements. According to the volcanic maps of OER and ORR, it is found that Ni@B4N system (ηOER = 0.45 V, ηORR = 0.43 V) and Pd@B2O system (ηOER = 0.56 V, ηOER = 0.47 V) as ORR/OER bifunctional SACs have a high application prospect, and their catalytic performance can be comparable to that of traditional noble metal-based catalysts. Considering the tons of structural candidates for such 2D substrates as well as the simple and tunable strain-engineering strategy, we wish the present theoretical basis will spur more interests and efforts in designing efficient bifunctional SACs in overall water splitting from both the theory and experiments.