Realizing a high OER activity in new single-atom catalysts formed by introducing TMN_x (x = 3 and 4) units into carbon nanotubes using high-throughput calculations

Abstract

Exploring highly efficient electrocatalysts for the oxygen evolution reaction (OER) is of great significance for hydrogen production through water splitting. By means of high-throughput density functional theory (DFT) calculations, we investigated the OER catalytic activity of a series of one-dimensional carbon nanotube (CNT)-based systems containing TMN₄ or TMN₃ functional units. Through the screening of 3d/4d/5d transition metals (TMs) from Group IVB to Group VIII, eight newly obtained TMN_x@CNT (x = 3 and 4) systems were found to exhibit excellent OER activity, with very low overpotentials in the range 0.29–0.51 V, where the Co, Rh, Ir, Ti, Fe, and Ru atoms could be used as active sites. It was found that under the framework of TMN₃@CNTs, the pre-adsorption of some species from water dissociation on the relevant TM sites (TM = Ti, Fe, and Ru) could lead to a high OER catalytic activity, which was different from the general situation where OER reactions directly occur on the clean surfaces of the remaining systems with Co/Rh/Ir metal centers. Moreover, the catalytic mechanisms were analyzed in detail. This work can be conducive to obtaining low-cost and high-performance OER single-atom electrocatalysts based on excellent CNT nanomaterials.

---

Introduction

In order to solve the global energy crisis and the environmental problems caused by fossil fuel consumption, researchers have made many efforts to find sustainable alternative energy sources.^1–3 Hydrogen can be considered as a promising energy carrier for sustainable development owing to its high energy density and zero carbon emissions.^4–6 Currently, electrochemical water splitting is recognized as one of the most effective methods for hydrogen production.^7–9 As one half-reaction for water splitting, the oxygen evolution reaction (OER) at the anode has a very slow kinetics owing to its multi-electron transfer process. Therefore, to ensure the smooth progress of the reaction, it is necessary to employ highly efficient catalysts to reduce its overpotential.^10,11 It is well known that RuO₂ and IrO₂ can be viewed as the best-performing OER catalysts due to their low overpotential and high efficiency. However, their widespread application is largely limited because of the small reserves and high price of the metals.^12,13 Therefore, developing cost-effective OER catalysts is critical for eliminating the bottleneck in the practical application of electrochemical water splitting.

Generally, there are two ways for developing ideal OER electrocatalysts. One approach is to use fewer precious metals without reducing the catalytic efficiency. Another approach is to design nonprecious metal alternatives.^14–16 In the development process of OER electrocatalysts, the emergence of single-atom catalysts (SACs) can provide new opportunities for achieving nonprecious, highly efficient OER catalysts. The concept of SACs was first proposed by Qiao and coworkers in their report on the incredible performance of Pt₁/FeO_x in CO oxidation.¹⁷ It is well known that SAC systems can have the unique advantages of a high utilization rate of active atoms, uniform active center, low coordination number of active center atoms, and high selectivity.^18,19 At present, some progress has been made in the research on SACs for the OER.^20–23

Among them, the development of low-dimensional carbon-based SACs for the OER has attracted much attention owing to their enormous potential applications.^24–26 So far, some typical carbon structures have been used to construct the SAC electrocatalysts for the OER, such as graphene,^27,28 graphyne,^29–32 macroporous T-carbon,³³ and fullerene.³⁴ As a classic type of nanocarbon, carbon nanotubes (CNTs) can be considered as one of the most important members of the low-dimensional carbon-based material family. They can be regarded as a one-dimensional (1D) cylindrical structure with a nanoscale diameter, which has a large area π-conjugated skeleton composed of sp² hybridized C atoms. Due to the unique structure, CNTs can exhibit many unprecedented properties, such as a high Young's modulus, large tensile strength, and excellent electrical and thermal conductivity.^35,36 In particular, CNTs have become attractive support materials because of their high conductivity, structural stability, large surface area, low cost, and corrosion resistance in acid and alkali solutions.^37–39 In addition, it can be found that the electronic structure and reactivity of CNT materials is closely related to their inherent structural curvature.^40,41 It is worth mentioning that the CNT-based SAC systems have been used as effective electrocatalysts for both the HER^42–44 and ORR.^45–48 However, relevant research is quite scarce, especially on the OER electrocatalytic activity of SACs based on CNTs.

Unlike the previous investigations, in this study, we propose an effective strategy to construct novel one-dimensional (1D) SAC catalysts for the OER by introducing the structural units TMN_x (x = 3 or 4) into CNTs. High-throughput DFT calculations were carried out to screen the series of 3d/4d/5d transition metals from Group IVB to Group VIII. It is worth mentioning that some different structures with TMN_x units exhibited good OER catalytic performance, such as M-C₃N₄ (M = Cr, Mn, Fe, Co, Ni, Cu, and Zn),⁴⁹ Mn-CoN,⁵⁰ M-COF (M = Mn, Fe, Co, Ni, and Cu),^51,52 and MOF (M = Fe, Co, Ni, and Ru).^53–55 It is highly expected that these typical CNT-based SAC systems would possess excellent OER catalytic activity. Indeed, our calculation results indicated that a series of CNT-based SACs involving the TMN₃ or TMN₄ units (TM = Co, Rh, Ir, Ti, Fe, or Ru) would exhibit excellent OER catalytic performance with very low overpotentials (η) ranging from 0.29 to 0.51 V, which are comparable to or even lower than IrO₂ (0.56 V).^56,57 Obviously, this work can provide new ideas for achieving low-cost and high-performance OER electrocatalysts.

Computational method

All calculations were carried out using density functional theory (DFT), as implemented in the Vienna Ab initio Simulation Package (VASP).^58,59 The exchange correlation energy was described by the Perdew–Burke–Ernzerhof (PBE) functional within the Generalized Gradient Approximation (GGA),⁶⁰ which includes the semiempirical van der Waals (vdW) method proposed by Grimme (DFT-D2) to account for the dispersion interactions.^61–63 The nuclei–electron interactions were described by the projector augmented wave (PAW) pseudo-potentials. The 1 × 1 × 5 Monkhorst–Pack k-points were used to sample the Brillouin zone for the structural optimization, and the 21 Line-mode k-points along the (0, 0, 0) → (0, 0, 0.5) high symmetry line were employed to calculate the density of states (DOSs). The convergence criteria of the total energy and force were set to 10⁻⁴ eV and 0.05 eV Å⁻¹, respectively. The cutoff energy for the plane wave basis set was 450 eV. A vacuum space of 20 Å along the x- and y-directions was utilized to prevent spurious interactions between the periodically repeated images.

To evaluate the structural stability of the studied TMN_x@CNT (x = 3 and 4) systems, the corresponding binding energy (E_b) was calculated by the following formula:

E_b = E_TM + E_Nx@CNT − E_TMNx@CNT (1)

where E_Nx@CNT and E_TMNx@CNT represent the energies of the N-substituted CNT systems without or with TM-doping, respectively, E_TM is the energy of an isolated TM atom, and x is number of N atoms.

In acid media, the overall OER can be expressed as:

2H₂O → O₂ + 4H⁺ + 4e⁻ (2)

The OER processes have the following four electron reaction paths:^64–66

H₂O + * → HO* + H⁺ + e⁻ (a)

HO* → O* + H⁺ + e⁻ (b)

O* + H₂O → HOO* + H⁺ + e⁻ (c)

HOO* → * + O₂ + H⁺ + e⁻ (d)

where * denotes the adsorption site for oxygen-containing intermediates, and OH*, O*, and OOH* represent three different catalytic intermediates.

The adsorption energies of the oxygen-containing intermediates on the TMN_x@CNT (x = 3 and 4) structures were defined as:⁶⁷

ΔE_OH* = E_OH* − E_* − (E_H2O − 1/2E_H2) (3)

ΔE_O* = E_O* − E_* − (E_H2O − E_H2) (4)

ΔE_OOH* = E_OOH* − E_* − (2E_H2O − 3/2E_H2) (5)

where E_*, E_O*, E_OH*, and E_OOH* represent the total energies of the catalyst substrate without and with the adsorption of O, OH, and OOH, respectively, and E_H2O and E_H2 are the total energies of H₂O and H₂ molecules in the gas phases, respectively.

Then, the adsorption free energy can be obtained by including the zero-point energy (ZPE) and the entropy (S) corrections in the following equation:⁶⁵

ΔG = ΔE + ΔZPE − TΔS + ΔG_U + ΔG_pH (6)

where ΔZPE and ΔS are the changes of zero-point energy and the entropy contribution, respectively; ΔE represents the reaction energy difference of the reactant and product, which can be directly calculated from DFT calculations; and ΔG_U is the free energy change caused by the applied potential U (ΔG_U = –eU, where e is the elementary charge). ΔG_pH = k_BT ln 10 × pH, where k_B is the Boltzmann constant and pH = 0 was used in this study, according to convention.^24,68–70

The reaction free energies of the four steps (a)–(d) for the OER were defined as ΔG_a, ΔG_b, ΔG_c, and ΔG_d, respectively. In view of ΔG_a + ΔG_b + ΔG_c + ΔG_d = 4.92 eV, the best scenario corresponded to the case where they are all the same and equal to 1.23 eV, which is an ideal OER catalyst. Then the overpotential (η) can be calculated to evaluate the performance for the OER by:⁶⁷

η = max{ΔG_a, ΔG_b, ΔG_c, ΔG_d}/e − 1.23 V (7)

In addition, we also conducted a computational test to explore the solvation effect on the OER catalytic activity by sampling the RhN₃@CNT system (Fig. S1†). The calculation results showed that the overpotential of the OER in solvation could be as small as 0.41 V, which was close to the corresponding result in vacuum (0.46 V), indicating a negligible solvation influence. Therefore, to make the computational cost less demanding, in this work we calculated the η value under vacuum conditions for estimating the OER activity of the studied systems.

Results and discussion

Geometric structures and stability

Initially, we optimized the structure of a pristine (6, 0) CNT, and then removed two adjacent C atoms to form a divacancy defect. Further, one 3d/4d/5d TM atom ranging from Group IVB to Group VIII was inserted into the divacancy defect, while the surrounding C atoms were replaced with three or four N atoms (Fig. 1a and b). All the TM atoms considered are listed in Fig. 1c. For convenience, these new systems are labeled as TMN_x@CNT (x = 3 and 4), according to the number of N atoms. As shown in Fig. 1a and b, the TM atom was usually located at the center of the divacancy after optimization, and slightly protruded from the plane formed by the four surrounding atoms, in view of the curvature of the nanotube. The involved TM-N/TM-C bond lengths ranged from 1.848 to 2.238 Å. The calculated lattice parameter a of TMN₃@CNT and TMN₄@CNT systems could be in the range of 12.789 to 12.868 Å, very close to that of a pure CNT (12.793 Å), which indicated that doping a TM atom into the CNT did not cause significant structural deformation. Moreover, the binding energy (E_b) of the TM atom was calculated to estimate the stability of TMN_x@CNT (x = 3 and 4). As shown in Fig. 1d, all these doped structures exhibited large binding energies in the range of 6.027–11.198 eV and 5.132–10.218 eV for TMN₃@CNT and TMN₄@CNT, respectively. This indicates that all these doped systems had high structural stability, with the TMN₃@CNT system generally slightly more stable than its corresponding counterpart in the TMN₄@CNT series.

Fig. 1 The structures of (a) TMN₄@CNT and (b) TMN₃@CNT. The brown, blue, and gold balls represent the C, N, and TM metal atoms, respectively. (c) The schematic illustrations of the TM dopants considered in this work. Note that different situations related to the thermodynamically optimal pre-adsorption process are marked with the corresponding colors according to the adsorption species. (d) The corresponding binding energies for TMN₄@CNT and TMN₃@CNT.

Furthermore, the calculated electron location function (ELF) results revealed that all the C–C and C–N bonds exhibited strong covalent characteristic, in view of the fact that the ELF isovalues at the center of these chemical bonds were approximately 0.8 (Fig. S4 and S5†). In contrast, strong ionic bonds could be formed between the TM and N/C atom, where the ELF isovalues around TM atom could be close to zero, while the ELF isovalues around its adjacent N/C atoms could be larger than 0.9. The formation of all these strong chemical bonds could jointly promote the high structural stability of these doped systems. This is advantageous for them to become potential SACs for the OER in the process of water splitting, where the TMN_x (x = 3 and 4) part is very promising to become high active sites.

OER catalytic activity

In this section, we conducted a systematic study on the OER catalytic activity of these stable TMN₃@CNT and TMN₄@CNT structures by screening almost all 3d/4d/5d transition metals. According to the approach proposed by Rossmeisl et al.,⁵⁶ the OER process typically consists of four elementary reaction steps involving some intermediates (i.e., OH*, O*, and OOH*), each of which involves a proton/electron coupled transfer process, as shown in eqn (a)–(d) above. The overpotential (η) was used to evaluate the catalytic activity of the material systems, which could be obtained by assessing the difference between the minimum voltages required for the four reaction steps.

Here, the intermediates with an OH group adsorbed at the TM sites were optimized for the TMN₃@CNT and TMN₄@CNT series, where 21 different TM metal atoms were considered (Fig. 1c).

The corresponding Gibbs free energies are shown in Fig. 2. It could be found that the calculated ΔG_OH* value generally presented a periodic change trend with the number of valence electrons of the TM atom. In these doped CNT systems, the TM atoms with fewer valence electrons could have negative ΔG_OH* values, while those with more valence electrons could have positive ΔG_OH* values. Among them, we first focused on the TMN₄@CNT systems (TM = Mn, Fe, Co, Ni, Ru, Rh, Pd, Ir, and Pt) with positive ΔG_OH* values (0.04–2.03 eV), considering that for a good catalyst, all the steps in the OER reaction should usually move uphill when no bias is applied. As shown in Fig. S6,† the OER free energy diagrams of these systems were constructed, revealing that the CoN₄@CNT, RhN₄@CNT, and IrN₄@CNT systems exhibited very small overpotential (η) values, which were 0.51, 0.29, and 0.33 V, respectively. It is worth mentioning that all of them were even smaller than that of the state-of-the-art IrO₂ system (0.56 V),^56,57 indicating that these SAC systems doped with TM atoms with 9 valence electrons in the Group VIII could exhibit a considerably high OER catalytic activity. Besides, the neighboring NiN₄@CNT system, doped with Ni atom in the Group VIII, could also present good OER activity, in view of the comparable η value (0.64 V) to IrO₂. In contrast, the other TMN₄@CNT systems (TM = Mn, Fe, Ru, Pd, and Pt) could be considered to have poor OER electrocatalytic performance, as reflected by their large η values (0.92–1.51 V).

Fig. 2 The adsorption Gibbs free energy of the OH group at the TM site for the TMN₄@CNT and TMN₃@CNT series. The dark cyan, light yellow, and red columns represent the doped systems with transition metals of 3d, 4d, and 5d, respectively.

In addition, the remaining TMN₄@CNT systems (TM = Ti, V, Cr, Zr, Nb, Mo, Tc, Hf, Ta, W, Re, and Os) were also investigated. All of these could have negative ΔG_OH* values (−2.42 to −0.08 eV), indicating that these TM sites may have relatively strong interactions with some related species in the solution, which may cause additional reaction processes before the regular OER occurs. As is well known, the dissociation of water can usually form OH, O, and H species. Therefore, before the OER process, their possible occupation of TM sites should be investigated by calculating the ΔG_OH*, ΔG_O*, and ΔG_H* values based on the formulas (1)–(3) in the ESI.† All the calculation results are shown in Fig. 1c and Table S1.†

Specifically, we found that the TMN₄@CNT systems doped with Ti, V, Cr, Zr, or Hf atoms exhibited the most negative ΔG_OH* value, suggesting that the corresponding TM site had the strongest interaction with OH*. Comparatively, the CNT systems doped by the 4d/5d TM atoms (i.e., Nb, Mo, Tc, Ta, W, and Re) from Group VB to VIIB had the strongest interaction with O*, as reflected by their more negative ΔG_O* values (Table S1†). This indicates that during the electrocatalytic OER process, the TM site of the former tends to be occupied by OH species, while that of the latter favors being occupied by O species. Furthermore, given the more negative ΔG_H* values, the Os atom in Group VIII elements is more likely to be occupied by H species. Clearly, different from the TM sites with more valence electrons (such as Group VIII elements), these TM sites with less valence electrons will be first occupied by the related different species in the OER process, which could be expected to have an important impact on the OER catalytic performance.

Then, the overpotential (η) values of these TMN₄@CNT systems (TM = Ti, V, Cr, Zr, Nb, Mo, Tc, Hf, Ta, W, Re, and Os) were estimated based on the relevant structural models, where the corresponding TM site was first occupied by the OH, O, or H species before the OER reaction occurs. By performing a computational screening (Fig. S6†), it was found that these TMN₄@CNT systems could exhibit large η values in the range of 0.70–2.09 V, indicating their relatively poor OER catalytic activity.

Subsequently, we investigated the OER catalytic activity of the TMN₃@CNT series, and explored whether more TM-doped CNT structures with high OER performance could be achieved by changing the functional unit from TMN₄ to TMN₃. Initially, we focused on the TMN₃@CNT systems with positive ΔG_OH* values (TM = Mn, Co, Ni, Rh, Pd, Ir, and Pt). Our computed results reveal that when the number of N atoms decreased, the obtained CoN₃@CNT and RhN₃@CNT systems could still exhibit small η values as low as 0.39 and 0.46 V, indicating that they could maintain a high OER catalytic activity. Additionally, other TMN₃@CNT systems doped with TM = Ir, Mn, Ni, Pd, and Pt could possess relatively inert OER catalytic activity, in view of their large overpotential values (0.72–1.47 V).

Further, we found that the TM sites with less valence electrons would also be occupied by the OH, O, or H species, when converting the functional unit from TMN₄ to TMN₃. However, some differences could also be observed. For instance, the doped CNT systems by the Nb, Ta, or Os atom will be occupied by the OH* group instead of the original O* or H*. Especially, different from the clean surface of TMN₄@CNT, the Fe and Ru sites will also be occupied by OH* species within the TMN₃@CNT framework (Fig. 1c), which could cause the extremely low η value of about 0.35 V, indicating the excellent OER activity. Obviously, the occupation of OH at the TM site can be considered as a self-optimization process for the OER catalytic activity of FeN₃@CNT and RuN₃@CNT in the process of electrochemical water splitting, which can effectively modulate the adsorption strength of the relevant species involved in the OER process, ultimately leading to a high OER catalytic performance. This is significantly different from the general situation that the OER reaction can occur directly on the surface of a clean material through the four elementary reaction steps. Moreover, the TiN₃@CNT system could also exhibit high OER catalytic activity, as reflected by the small η value of about 0.48 V. Obviously, by converting the functional unit from TMN₄ to TMN₃, more new carbon nanotube-based SAC structures with high OER activity could be obtained.

In addition, by using the overpotential (η) as a function of ΔG_OOH* − ΔG_O*, activity volcano plots were constructed to display the overall OER catalytic activity in the TMN₄@CNT and TMN₃@CNT frameworks (Fig. 3). As is well known, constructing a volcano plot has been widely used as a powerful method for screening high-performance electrocatalysts. Usually, the top of the volcano plot indicates the optimal catalytic activity of the catalyst, because a lower overpotential implies a better electrochemical activity. From Fig. 3a, it can be found that the calculated η values of the TMN₄@CNT series were in the range of 0.29–2.10 V. The CoN₄@CNT, RhN₄@CNT, and IrN₄@CNT systems were located at the peak of the volcano, with very small η values of 0.51, 0.29, and 0.33 V, respectively. Additionally, the η values of the TMN₃@CNT series were in the range of 0.35–1.57 V (Fig. 3b). We can observe that more TMN₃@CNT (TM = Ti, Fe, Ru, Co, and Rh) systems could be located at the peak of the volcano plot, exhibiting very small overpotentials of approximately 0.48, 0.35, 0.35, 0.39, and 0.46 V, respectively. All these TMN₄@CNT and TMN₃@CNT systems could even have lower η values than that of the state-of-the-art IrO₂ (0.56 V), indicating their excellent OER catalytic activity.

Fig. 3 The volcano plots of the calculated overpotential η versus ΔG_OOH* − ΔG_O* for (a) TMN₄@CNT and (b) TMN₃@CNT systems.

Furthermore, the stability of these eight one-dimensional TMN₄@CNT (TM = Co, Rh, and Ir) and TMN₃@CNT (TM = Ti, Fe, Ru, Co, and Rh) systems with high OER activity were further evaluated by AIMD simulations at 500 K, as shown in Fig. S8.† This temperature has been widely accepted for evaluating thermal stability.^71–74 It was found that after 5000 fs, they could still maintain structural integrity, and their average potential energy remained almost unchanged, indicating their high thermal stabilities. Moreover, we also investigated the electronic behavior of these eight structures by calculating their density of states (DOSs). Our computed results show that they uniformly exhibited a metallic behavior in view of the relevant states crossing the Fermi level (Fig. S9†), suggesting their high conductivity. All of these can be beneficial for the high OER catalytic activity.

Obviously, introducing the TMN₄ or TMN₃ units into one-dimensional CNTs with curvature can be considered as a new and effective strategy to achieving SACs with high OER performance. As the number of N atoms decreased from four to three, more new SAC structures with high OER activity could be obtained. In particular, compared to the general case of the OER on a clean material surface, the pre-adsorption of some species from water dissociation on the TM sites could play an important role in realizing high OER catalytic activity in some SAC systems.

OER catalytic mechanisms

From the above discussions, we know that the TMN₄@CNT and TMN₃@CNT systems doped with Co/Rh/Ir/Fe/Ru atoms in Group VIII and Ti atom in Group IVB would possess excellent OER catalytic performance. In this section, we conducted a detailed analysis of the catalytic mechanism to reveal the underlying reasons behind their high OER catalytic activity.

Initially, the influence of different TM metals on the overpotential η value was explored by plotting the relationship between η and the number of valence electrons (N_val) of TM in the TMN₄@CNT and TMN₃@CNT frameworks. It can be seen from Fig. 4a that for the TMN₄@CNT series, the η values showed a fluctuating trend with the increase in the number of valence electrons of TM. To be specific, when employing the TM atoms with four valence electrons, the doped systems exhibited moderate η values. Then, as the number of valence electrons increased from four to six in succession, the overpotential η value increased monotonically, and then decreased when N_val further increased from six to nine, and increased again for the system with N_val = 10. Obviously, the number of valence electrons of the TM atom can play an important role in determining the overpotential η value of the TMN₄@CNT system, where the CNT systems doped by Co/Rh/Ir atoms with nine valence electrons exhibited the optimal OER catalytic performance.

Fig. 4 Relationship between the calculated overpotential (η) and the valence electrons (N_val) of TM on (a) TMN₄@CNT and (b) TMN₃@CNT.

A similar fluctuation could also be observed in the TMN₃@CNT series. As shown in Fig. 4b, the relationship curve between η and N_val still showed a trend of first rising, then falling, and then rising with the increase in the number of valence electrons of the TM atom. The difference is that when N_val increased from four to five, the η value monotonously increased, and then monotonically decreased until N_val increased to eight. Eventually, the η value monotonically increased again when increasing N_val from eight to ten. Clearly, the effective influence of the N_val of the TM atoms on the η value was reflected again in the TMN₃@CNT series.

Further, the relationships of ΔG_OH*vs. ΔG_O* and ΔG_OOH*vs. ΔG_O* were fitted for the two series including TMN₄@CNT and TMN₃@CNT. It can be found that they exhibited good linear correlations, as shown in Fig. S10.† Specifically, for ΔG_OH* and ΔG_O*, the linear relationships could be expressed as ΔG_OH* = 0.55ΔG_O* − 0.49 and ΔG_OH* = 0.66ΔG_O* − 0.58 for TMN₄@CNT and TMN₃@CNT, respectively, where the corresponding correlation coefficients (R) were about 0.86 and 0.87. As for ΔG_OOH* and ΔG_O*, the relational expressions of ΔG_OOH* = 0.47ΔG_O* + 2.83 (R = 0.90) and ΔG_OOH* = 0.50ΔG_O* + 2.97 (R = 0.88) could be obtained for TMN₄@CNT and TMN₃@CNT, respectively. Clearly, regardless of the number of N atoms, both ΔG_OH* and ΔG_OOH* were linearly proportional to ΔG_O*, thus ΔG_O* can be regarded as a valid descriptor for determining the overpotential η values for the two series. This situation can also be reflected by the volcano diagram of ΔG_O* and η, as shown in Fig. 5a and b.

Fig. 5 The volcano curves of η versus ΔG_O* for (a) TMN₄@CNT and (b) TMN₃@CNT systems after considering a possible pre-adsorption process of the relevant species (e.g., OH, O, and H). The relationships between ΔG_O* and valence electrons (N_val) of TM for (c) TMN₄@CNTs and (d) TMN₃@CNTs. The partial density of states (PDOS) about the p orbitals of O atoms and the d orbitals of TM atoms after the O* adsorption of TiN₄@CNTs (e), CoN₄@CNTs (f), RhN₄@CNTs (g), IrN₄@CNTs (h), TiN₃@CNTs (i), FeN₃@CNTs (j), RuN₃@CNTs (k), CoN₃@CNTs (l), and RhN₃@CNTs (m). The Fermi level was set to zero (black dashed line). Inset: molecular orbitals of the O atom adsorbed at the TM site in different energy ranges marked by green dashed lines. Green arrow represents the p–d overlap center.

In Fig. 5a, we can observe that the TMN₄@CNT systems were located on the left side of the volcano plot when the TM atoms come from Group IVB (Ti, Zr, and Hf), VB (V, Nb, and Ta), VIB (Cr, Mo, and W), and VIIB (Mn, Tc, and Re) elements, as well as some atoms in Group VIII, such as Fe, Ru, and Os. This means that the interaction between O* and these TM atoms is relatively strong. In contrast, the TMN₄@CNT systems doped with Ni, Pd, and Pt in Group VIII were located on the right side of this volcano map, indicating weak interactions between O* and these TM sites. Obviously, the adsorption strength of the intermediates O* can play a crucial role in determining the OER activity of such CNT-based SACs. As is well known, too strong an adsorption of reactants can usually block the catalytic sites and hinder the catalytic reaction, while too weak an adsorption cannot provide sufficient driving force to bind with the adsorbent, resulting in a low catalytic efficiency. Therefore, the appropriate adsorption strength of O* is necessary for efficient catalysis. Indeed, the TMN₄@CNT systems doped by Co, Rh, and Ir were located at the peak of the volcano plot, and exhibited a suitable adsorption state for O*, leading to their high OER catalytic activity. Similarly, the CoN₃@CNT and RhN₃@CNT systems were also located at the peak of the volcanic curve (Fig. 5b). Besides, the TiN₃@CNT, FeN₃@CNT, and RuN₃@CNT systems were located at the peak of the volcano plot, different from the corresponding TMN₄@CNT systems. Clearly, the appropriate adsorption of O* can also endow these five TMN₃@CNT systems with high OER catalytic activity.

These findings raise the question, why can doping these TM atoms induce the appropriate adsorption state of O* within the framework of TMN₃@CNT and TMN₄@CNT? To solve this issue, we plotted the relationships between ΔG_O* and the valence electron number N_val of TM atom, as shown in Fig. 5c and d. It was found that for the TMN₄@CNT series, the ΔG_O* values showed a trend of first decreasing and then increasing with the increase in N_val, and the systems doped by Cr, Mo, and W with N_val = 6 displayed the smallest ΔG_O* values. A similar trend could also be observed in the TMN₃@CNT series, where the systems doped by V, Nb, and Ta with N_val = 5 had the lowest ΔG_O* values. Obviously, the filling of valence electrons can play a key role in determining the ΔG_O* value. To be specific, when there are fewer valence electrons, the bonding ability between O* and TM metals is relatively strong, thus showing relatively small ΔG_O* values. In contrast, a continuous increase in valence electrons in the d orbitals will lead to the existence of antibonding characteristic, which can weaken the bonding ability between O* and the TM metal, and lead to an increase in the ΔG_O* value. Clearly, during the gradient process, moderate ΔG_O* values were obtained in the TMN₄@CNT (TM = Co, Rh, and Ir) and TMN₃@CNT (TM = Ti, Fe, Ru, Co, and Rh) systems, resulting in high OER catalytic activity.

In order to better understand the catalytic mechanism, we further conducted bonding analysis on all the TMN₄@CNT and TMN₃@CNT systems with excellent OER catalytic performance. For comparison, the TiN₄@CNT system with poor OER catalytic activity was also considered. From Fig. 5e, we can find that for TiN₄@CNT, the overlapping center of the O-p and Ti-d orbitals was located in the π bonding area, meaning a relatively strong interaction between the O* and Ti-site, which is the reason for the inert OER catalytic performance. Comparatively, for the three TMN₄@CNT (TM = Co, Rh, and Ir) systems, the overlapping centers of the O-p and TM-d orbitals were located in the π* antibonding area, which could effectively weaken the interaction between the O* and TM, and bring about the appropriate the adsorption state of O*, leading to their excellent OER catalytic activity (Fig. 5f–h).

When converting the functional unit from TiN₄ to TiN₃, the overlapping center of the O-p and Ti-d orbitals also moved into the π* antibonding region (Fig. 5i). The appearance of π* antibonding characteristic endowed the TiN₃@CNT system with the moderate ΔG_O* values, ultimately leading to a high OER catalytic performance. In addition, a similar situation was further reflected in the TMN₃@CNT systems (TM = Fe, Ru, Co, and Rh), where all the overlapping centers of the O-p and TM-d orbitals were also located in the π* antibonding region, as shown in Fig. 5j–m. Obviously, for all these 1D TMN₄@CNT and TMN₃@CNT structures, the valence electrons of the TM can play a crucial role in determining the overpotential η by effectively influencing the adsorption state of O*, regardless of the number of N atoms.

Obviously, for a series of TMN_x@CNT systems, the adsorption state of O* (ΔG_O*) as an effective descriptor can be not only affected by the number of valence electrons in the TM atom, but also by the chemical environment surrounding the TM atom (such as the coordinated N atoms and the possible adsorbed species, including OH*, O*, and H*). The effective synergy of these factors will induce the appropriate adsorption state of O* (usually exhibiting a certain antibonding characteristic), which can lead to a high OER catalytic performance with a low overpotential η.

Finally, based on the linear relationship in Fig. S11, Tables S2, 3,† and eqn (3)–(7), a two-dimensional volcano diagram was constructed to evaluate the limiting overpotential η of the TMN₄@CNT and TMN₃@CNT series. A similar approach has also been successfully employed in previous works.^75–77Fig. 6 shows that the 2D volcano map could be divided into four regions according to the different potential determination steps (PDSs). Most the TMN₄@CNT and TMN₃@CNT samples were distributed in the PDS3 region (O* → OOH*). We can find that the adsorption strength of O* is an important factor to determine the PDS step of the OER reaction. The TMN₄@CNT and TMN₃@CNT systems with weaker O* absorption were located in the PDS1 and PDS2 regions, while the related systems with stronger O* absorption were located in the PDS3 and PDS4 regions for TMN₄@CNT and PDS2 and PDS3 regions for TMN₃@CNT, respectively. Especially, the structures located at the boundary between PDS2 and PDS3 regions could display the best OER catalytic activity, including the systems doped by Ti, Fe, Ru, Co, Rh, and Ir. Here, the red area at the center of these two volcano plots predicts the limiting overpotential η of the TMN₄@CNT and TMN₃@CNT systems, both of which could uniformly be as small as 0.27 V. All of these provide good in-depth insights for designing SACs with high OER activity within the frameworks of TMN₄@CNT and TMN₃@CNT.

Fig. 6 The 2D colored contour plots of OER activity volcanos for (a) TMN₄@CNT and (b) TMN₃@CNT by presenting the η value as a function of the Gibbs free energy.

Conclusions

In summary, high-throughput DFT calculations were carried out to systematically investigate the structures and OER catalytic activities of a total of 42 1D CNT-based systems with the TMN₃ or TMN₄ functional units. The following interesting findings can be noted:

(1) All these TMN_x@CNT (x = 3 and 4) systems possessed high structural stability, in which TM atoms were stably anchored to form the TMN₃ or TMN₄ units. In addition, the DOS calculations indicated that they would exhibit metallic conductivity. All of these advantages are beneficial for the catalytic performance of the OER.

(2) By screening a series of 3d/4d/5d TM atoms, we found that when employing Co, Rh, and Ir atoms in Group VIII, the overpotential η values of the TMN₄@CNT systems could be as low as 0.29–0.51 V, which is comparable to or even lower than the state-of-the-art IrO₂ (0.56 V), suggesting their considerably high OER catalytic activity.

(3) In contrast, a high OER activity could also be observed in the TMN₃@CNT framework systems, which were doped not only with Co and Rh atoms, but also with Ti, Fe, and Ru atoms, as revealed by the corresponding small η values (0.35–0.48 V). In particular, different from the general situation of the OER, the pre-adsorption of some species from water dissociation on the TM centers could play a key role in realizing the high OER activity within the framework of TMN₃@CNT (TM = Ti, Fe, and Ru).

(4) ΔG_O* can be considered to be a good descriptor for the OER catalytic activity of TMN_x@CNT (x = 3 and 4) systems. The filling of the valence electrons of the TM atom can have an important impact on the ΔG_O* value, and the appropriate ΔG_O* values led to the low overpotentials for all these eight TMN₃@CNT or TMN₄@CNT systems, showing excellent OER catalytic activity.

Clearly, introducing the functional units TMN₃ or TMN₄ into 1D CNTs can be regarded as an effective strategy for obtaining nonprecious and highly efficient SACs for the OER. This study can offer in-depth theoretical insights into the OER activity and mechanism of CNT-based SACs, which is advantageous for promoting the practical application of important CNT-based nanomaterials in OER catalytic processes.

Author contributions

Xia Yang performed investigation, data curation and writing – original draft preparation. Guangtao Yu and Wei Chen performed conceptualization, investigation, writing – review and editing, supervision, project administration and funding acquisition.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was financially supported by the Natural Science Foundation of Fujian Province (2020J01147 and 2022J01167), Research Foundation of Academy of Carbon Neutrality of Fujian Normal University (TZH2022-05), Minjiang Scholar and startup fund for high-level talent at Fujian Normal University, and Fujian-Taiwan Science and Technology Cooperation Base of Biomedical Materials and Tissue Engineering (2021D039). We acknowledge the Computing Center of Jilin Province for supercomputer time.

References

1. Z. G. Yang, J. L. Zhang, M. C. W. Kintner-Meyer, X. C. Lu, D. W. Choi, J. P. Lemmon and J. Liu, Chem. Rev., 2011, 111, 3577–3613.
2. G. S. Liu, S. J. You, Y. Tan and N. Q. Ren, Environ. Sci. Technol., 2017, 51, 2339–2346.
3. N. L. Panwar, S. C. Kaushik and S. Kothari, Renewable Sustainable Energy Rev., 2011, 15, 1513–1524.
4. S. Chu and A. Majumdar, Nature, 2012, 488, 294–303.
5. G. W. K. Moore, S. E. L. Howell, M. Brady, X. Xu and K. McNeil, Nat. Commun., 2021, 12, 1.
6. L. Schlapbach, Nature, 2009, 460, 809–811.
7. J. Li, C. A. Triana, W. Wan, D. P. Adiyeri Saseendran, Y. Zhao, S. E. Balaghi, S. Heidari and G. R. Patzke, Chem. Soc. Rev., 2021, 50, 2444–2485.
8. G. Qian, G. Yu, J. Lu, L. Luo, T. Wang, C. Zhang, R. Ku, S. Yin, W. Chen and S. Mu, J. Mater. Chem. A, 2020, 8, 14545–14554.
9. H. Chang, Z. Liang, L. Wang and C. Wang, Nanoscale, 2022, 14, 5639–5656.
10. H. Dau, C. Limberg, T. Reier, M. Risch, S. Roggan and P. Strasser, ChemCatChem, 2010, 2, 724–761.
11. M. K. Debe, Nature, 2012, 486, 43–51.
12. N. T. Suen, S. F. Hung, Q. Quan, N. Zhang, Y. J. Xu and H. M. Chen, Chem. Soc. Rev., 2017, 46, 337–365.
13. X. Yin, L. Yang and Q. Gao, Nanoscale, 2020, 12, 15944–15969.
14. L.-L. Feng, G. Yu, Y. Wu, G.-D. Li, H. Li, Y. Sun, T. Asefa, W. Chen and X. Zou, J. Am. Chem. Soc., 2015, 137, 14023–14026.
15. L. Yang, G. T. Yu, X. Ai, W. S. Yan, H. L. Duan, W. Chen, X. T. Li, T. Wang, C. H. Zhang, X. R. Huang, J. S. Chen and X. X. Zou, Nat. Commun., 2018, 9, 5236.
16. K. Sathiyan, T. Mondal, P. Mukherjee, S. G. Patra, I. Pitussi, H. Kornweitz, R. Bar-Ziv and T. Zidki, Nanoscale, 2022, 14, 16148–16155.
17. B. T. Qiao, A. Q. Wang, X. F. Yang, L. F. Allard, Z. Jiang, Y. T. Cui, J. Y. J. Y. Liu, J. Li and T. Zhang, Nat. Chem., 2011, 3, 634–641.
18. X. J. Cui, W. Li, P. Ryabchuk, K. Junge and M. Beller, Nat. Catal., 2018, 1, 385–397.
19. X. N. Li, Y. Q. Huang and B. Liu, Chem, 2019, 5, 2733–2735.
20. X. Gong, Z. Jiang, W. Zeng, C. Hu, X. Luo, W. Lei and C. Yuan, Nano Lett., 2022, 22, 9411–9417.
21. X. Liang, F.-K. Chiang, L. Zheng, H. Xiao, T. Zhang, F. Zhang and Q. Gao, Chem. Commun., 2022, 58, 7678–7681.
22. R. Zhang, W. Liu, F.-M. Zhang, Z.-D. Yang, G. Zhang and X. C. Zeng, Appl. Catal., B, 2023, 325, 122366.
23. T. Zhang, B. Zhang, Q. Peng, J. Zhou and Z. Sun, J. Mater. Chem. A, 2021, 9, 433–441.
24. L. Cao, Q. Luo, J. Chen, L. Wang, Y. Lin, H. Wang, X. Liu, X. Shen, W. Zhang, W. Liu, Z. Qi, Z. Jiang, J. Yang and T. Yao, Nat. Commun., 2019, 10, 4849.
25. H. Fei, J. Dong, Y. Feng, C. S. Allen, C. Wan, B. Volosskiy, M. Li, Z. Zhao, Y. Wang, H. Sun, P. An, W. Chen, Z. Guo, C. Lee, D. Chen, I. Shakir, M. Liu, T. Hu, Y. Li, A. I. Kirkland, X. Duan and Y. Huang, Nat. Catal., 2018, 1, 63–72.
26. X. L. Zhang, Z. X. Yang, Z. S. Lu and W. C. Wang, Carbon, 2018, 130, 112–119.
27. K. Wang, Z. Lu, J. Lei, Z. Liu, Y. Li and Y. Cao, ACS Nano, 2022, 16, 11944–11956.
28. J. Liu, J. Xiao, B. Luo, E. Tian and G. I. N. Waterhouse, Chem. Eng. J., 2022, 427, 132038.
29. M. Guo, M. Ji and W. Cui, Appl. Surf. Sci., 2022, 592, 153237.
30. S.-L. Li, Q. Li, Y. Chen, Y. Zhao and L.-Y. Gan, Appl. Surf. Sci., 2022, 605, 154828.
31. H. Zhang, W. Wei, S. Wang, H. Wang, B. Huang and Y. Dai, J. Mater. Chem. A, 2021, 9, 4082–4090.
32. Z. Zhang, S. Qi, X. Song, J. Wang, W. Zhang and M. Zhao, Appl. Surf. Sci., 2021, 553, 149464.
33. Z. Zhao, J. Hao, B. Jia, X. Zhang, G. Wu, C. Zhang, L. Li, S. Gao, Y. Ma, Y. Li and P. Lu, J. Energy Chem., 2023, 83, 79–89.
34. H. Xiao, H. Li, X. Li and J. Jiang, J. Phys. Chem. Lett., 2022, 13, 7392–7397.
35. C. J. Shearer, A. Cherevan and D. Eder, Adv. Mater., 2014, 26, 2295–2318.
36. D. M. Sun, C. Liu, W. C. Ren and H. M. Cheng, Small, 2013, 9, 1188–1205.
37. D. He, S. Mu and M. Pan, Carbon, 2011, 49, 82–88.
38. D. P. He, C. Zeng, C. Xu, N. C. Cheng, H. G. Li, S. C. Mu and M. Pan, Langmuir, 2011, 27, 5582–5588.
39. T. Li, Y. Chen, W. Hu, W. Yuan, Q. Zhao, Y. Yao, B. Zhang, C. Qiu and C. M. Li, Nanoscale, 2021, 13, 4444–4450.
40. G.-L. Chai, Z. Hou, D.-J. Shu, T. Ikeda and K. Terakura, J. Am. Chem. Soc., 2014, 136, 13629–13640.
41. O. Gülseren, T. Yildirim and S. Ciraci, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 65, 153405.
42. H. J. Liu, L. M. Zhao, Y. H. Liu, J. Xu, H. Y. Zhu and W. Y. Guo, Catal. Sci. Technol., 2019, 9, 5301–5314.
43. C. J. Oluigbo, Y. G. Xu, H. Louis, A. B. Yusuf, W. Yaseen, N. Ullah, K. J. Alagarasan, M. Xie, E. E. Ekpenyong and J. M. Xie, Appl. Surf. Sci., 2021, 562, 150161.
44. L. M. Zhao, S. Guo, H. J. Liu, H. Y. Zhu, S. F. Yuan and W. Y. Guo, ACS Appl. Nano Mater., 2018, 1, 6258–6268.
45. S. Aoyama, J. Kaiwa, P. Chantngarm, S. Tanibayashi, H. Saito, M. Hasegawa and K. Nishidate, AIP Adv., 2018, 8, 115113.
46. H. Hu, P. Zhang, B. B. Xiao and J. L. Mi, ACS Appl. Mater. Interfaces, 2023, 15, 23170–23184.
47. A. V. Kuzmin and B. A. Shainyan, ACS Omega, 2021, 6, 374–387.
48. J. T. Niu, W. J. Qi, C. Li, M. Mao, Z. G. Zhang, Y. Chen, W. L. Li and S. S. Ge, Int. J. Energy Res., 2021, 45, 8524–8535.
49. Y. Zheng, Y. Jiao, Y. Zhu, Q. Cai, A. Vasileff, L. H. Li, Y. Han, Y. Chen and S.-Z. Qiao, J. Am. Chem. Soc., 2017, 139, 3336–3339.
50. Y. Zhang, B. Ouyang, G. Long, H. Tan, Z. Wang, Z. Zhang, W. Gao, R. S. Rawat and H. J. Fan, Sci. China: Chem., 2020, 63, 890–896.
51. D. Huang, Q. Wang, W. Dong, L. Liao, H. Chen and J. Ye, Energy Fuels, 2022, 36, 11601–11608.
52. Y. Li, J. Tan, X. Zhao, L. Zhao and X. Chen, Mater. Today Commun., 2023, 37, 107157.
53. T. Wang, Y. Wu, Y. Han, P. Xu, Y. Pang, X. Feng, H. Yang, W. Ji and T. Cheng, ACS Appl. Nano Mater., 2021, 4, 14161–14168.
54. K. Adpakpang, L. Pukdeejorhor, L. Ngamwongwan, S. Suthirakun, S. Impeng, S. Wannapaiboon, P. Chakthranont, K. Faungnawakij and S. Bureekaew, Chem. Commun., 2022, 58, 7124–7127.
55. B. Boro, M. K. Adak, S. Biswas, C. Sarkar, Y. Nailwal, A. Shrotri, B. Chakraborty, B. M. Wong and J. Mondal, Nanoscale, 2022, 14, 7621–7633.
56. J. Rossmeisl, Z. W. Qu, H. Zhu, G. J. Kroes and J. K. Nørskov, J. Electroanal. Chem., 2007, 607, 83–89.
57. X. Li, Z. Su, Z. Zhao, Q. Cai, Y. Li and J. Zhao, J. Colloid Interface Sci., 2022, 607, 1005–1013.
58. G. Kresse and J. Furthmüller, Comput. Mater. Sci., 1996, 6, 15–50.
59. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
60. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
61. S. Grimme, J. Comput. Chem., 2006, 27, 1787–1799.
62. G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 558–561.
63. G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 49, 14251–14269.
64. G. Gao, E. R. Waclawik and A. Du, J. Catal., 2017, 352, 579–585.
65. M. Li, L. Zhang, Q. Xu, J. Niu and Z. Xia, J. Catal., 2014, 314, 66–72.
66. J. K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J. R. Kitchin, T. Bligaard and H. Jónsson, J. Phys. Chem. B, 2004, 108, 17886–17892.
67. I. C. Man, H. Y. Su, F. Calle-Vallejo, H. A. Hansen, J. I. Martinez, N. G. Inoglu, J. Kitchin, T. F. Jaramillo, J. K. Norskov and J. Rossmeisl, ChemCatChem, 2011, 3, 1159–1165.
68. L. Wu, T. Guo and T. Li, Adv. Funct. Mater., 2022, 32, 2203439.
69. Y. Zhu, Z. He, Y. Choi, H. Chen, X. Li, B. Zhao, Y. Yu, H. Zhang, K. A. Stoerzinger, Z. Feng, Y. Chen and M. Liu, Nat. Commun., 2020, 11, 4299.
70. C. Xuan, Q. Xu, L. Han and B. Hou, Chem. Eng. J., 2023, 464, 142620.
71. X. Zhai, L. Li, X. Liu, Y. Li, J. Yang, D. Yang, J. Zhang, H. Yan and G. Ge, Nanoscale, 2020, 12, 10035–10043.
72. Y. Ying, K. Fan, X. Luo, J. Qiao and H. Huang, J. Mater. Chem. A, 2021, 9, 16860–16867.
73. Q.-J. Fang, J.-k. Pan, W. Zhang, F.-l. Sun, W.-x. Chen, Y.-f. Yu, A.-F. Hu and G.-l. Zhuang, J. Colloid Interface Sci., 2022, 617, 752–763.
74. F.-l. Sun, Q.-j. Fang, Y.-f. Yu, W. Zhang, J.-k. Pan, W.-X. Chen and G.-l. Zhuang, J. Mater. Chem. A, 2022, 10, 9565–9574.
75. A. Kulkarni, S. Siahrostami, A. Patel and J. K. Nørskov, Chem. Rev., 2018, 118, 2302–2312.
76. Z. Xue, X. Zhang, J. Qin and R. Liu, J. Mater. Chem. A, 2019, 7, 23091–23097.
77. Z. Xue, X. Zhang, J. Qin and R. Liu, J. Energy Chem., 2021, 55, 437–443.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3nr04396g

---