Synthesis of COPs with well-defined FeN 4 configuration and ascending DDE. Instead of a high-temperature pyrolysis process randomly riveting Fe in the carbon matrix with inevitable connection with diverse N moieties, we adopted symmetrical acid anhydride or cyan groups as reactive groups to prepare-oriented iron, nitrogen co-coordinated single-atom catalysts with well-designed configuration. Briefly, we deliberately assembled Tetracyanoethylene, Pyromellitic dianhydride, 1,4,5,8-Naphthalenetetracarboxylic dianhydride, and 3,4,9,10-Perylenetetracarboxylic dianhydride with Fe centers into extended sp2 carbon networks by solid phase synthesis, denoted by COP-Ene, COP-Ppcfe, COP-Nap, and COP-Pyr, respectively (Fig. 1).45, 46 The 13C solid-state nuclear magnetic resonance (NMR) spectra (Supplementary Fig. 1) confirmed that four structures have been successfully synthesized, where C = N links connecting building block were smoothly arranged in corresponding networks, which was also favored by stretch vibration peaks of C = N covalent bonds at ~ 1600 cm− 1 in Fourier transform infrared spectroscopy (FT-IR spectra, Supplementary Fig. 2).46, 47 More detailed characterization can be found in the supplementary materials.
Unlike conventional Fe-N-C catalysts, our pyrolysis-free strategy escapes unpredictable active-site configuration as well as accompanying diverse catalyst activities due to the pyrolysis process. The as-prepared COPs possess exclusive single Fe atom constrained in the conjugated 2D networks. Meanwhile, building blocks were orderly connected by rigid covalent bonds. As expected, Fe and N atoms kept highly homogeneous distribution in the carbon matrix for four as-synthesis samples, which was demonstrated by the high dispersion image of Fe and N atoms in high-resolution transmission electron microscopy (HRTEM, Fig. 2a, 2b and Supplementary Figs. 4–8). Furthermore, we took COP-Ppcfe sample as a typical case and performed the aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) test, which identified single Fe atoms from monodispersed bright spots images without metal clusters being observed (Fig. 2c and Supplementary Fig. 9). To verify the coordination environment and chemical states of Fe atom in as-synthesis sample, we carried out Fe K-edge analysis of X-ray absorption fine structure (XAFS). The COP-Ppcfe sample was demonstrated with high similarity to FePc benchmark, exhibiting dominated peaks corresponding to the Fe-N (~ 1.53 Å) and Fe-N-C (~ 2.7 Å) scattering paths (Fourier transform-extended XAFS spectra, Fig. 2d and Supplementary Fig. 11). Besides, a predominant intensity maximum of wavelet transform (WT) contour plots of Fe K-edge appearing at approximately 4.0 Å−1 in k space, which assigned to Fe-N(O) coordination (Fig. 2e), synergistically confirmed the existence of mononuclear Fe atoms. EXAFS quantitative least-squares fitting analysis further supported that the constructed active centers (Fig. 2f) were anastomotic with experimental EXAFS spectra of COP-Ppcfe samples, i.e., single Fe center coordinated with four N atoms and one adsorbed O atom (Fig. 2f and Supplementary Table 1).8, 10, 48
Although these as-synthesized COPs present semblable FeN4 coordination structure, they have obviously discrepant linked-carbon environment adjacent to FeN4 sites. As investigated by FT-IR spectra (Supplementary Fig. 2), sequential ascending C = C stretching vibrations at ~ 1400 cm− 1 manifest successively incremental planar π-electron abundance in four polymers. In addition, continuously ascending deconvoluted π-π* peaks in C1s XPS spectra (Fig. 2g) consistently demonstrated that the DDE gradually increases with carbon skeleton covalent-connected FeN4 sites in as-synthesized COPs. The typical sp2 carbon conjugated peak at 284.3 eV in C1s XPS spectra is closely related to the transition of π-band (1s → π*),49, 50 therefore, full-width at half maxima value (FWHM) of conjugated C = C bond peaks can work as a criterion to quantificationally evaluate corresponding delocalization of π electrons in carbon matrix.51 The narrower FWHM embodies fewer topological defects, higher DDE as well as stronger electron-donating capability in the carbon plane. As expected, observed significant shrink of FWHM along with the extension of sp2 carbon-conjugated frameworks doubtlessly demonstrates that the DDE indwelled in polymers obviously increases (Fig. 2h). Based on the elemental analysis results of the C element, the gradually increasing C atom contents (C wt.%, 33.18%, 47.53%, 56.65% and 66.65% for COP-Ene, COP-Ppcfe, COP-Nap and COP-Pyr, respectively) also consistently reflects structural expansion in carbon skeletons. Therefore, a series of FeN4 sites with defined configurations yet increasingly extending surrounding carbon networks were design-oriented and systematically synthesized.
Oxygen reduction performance and adsorption energy evaluation. The ORR measurements for four COPs were carried out in a harsh acidic medium (0.1 M HClO4) to evaluate their performance discrepancy (Fig. 3a). The Linear sweep voltammetry (LSV) curves revealed that the catalytic activity of the COP-Ppcfe sample was significantly ahead among the four COP samples, followed by COP-Ene; frustrating half-wave potentials mean that COP-Nap and COP-Pyr samples are difficult to undergo acidic oxygen reduction reaction. The turnover frequency (TOF) and mass activity (Ma) per mass of metal at 0.7 V versus RHE were further investigated to compare the discrepancy in the electrochemical intrinsic activity of the active sites in the four as-synthesized COPs. As observed in Fig. 3b, both TOF and Ma exhibit a volcano plot that first increases and then decreases with extension of sp2 carbon skeletons connected to FeN4 sites. In addition, the comparison in kinetic current density (Jk) of as-synthesized four COPs also embodies the volcano relationship between the electrochemical kinetics and DDE in carbon matrix connected to FeN4 sites (Supplementary Fig. 18). Aiming at this consistent volcano relationship plot, we next try to explain it from the adsorption strength of the reactant species obtained experimentally.
Theoretically, due to continuous multi-proton/-electron coupling reaction in the ORR process, large overpotential is ultimately stemmed from the mismatch of the adsorption energy.52 Due to the restriction of scaling relations in oxygen intermediates, a volcano relationship53 regularly emerges between ORR activity and adsorption energy, quantitatively illustrating Sabatier principle19, 20, 54. Namely, the binding between catalysts and oxygen intermediates is neither too strong nor too weak. However, the direct measurement for adsorption strength in experiment is a great challenge. Promisingly, this plight was eliminated by migrations of Fe2+/3+ redox potential (Eredox).55 As an effective indicator, the high Eredox manifests weak Fe-O binding, and vice versa. As observed, a negative growth tendency in Eredox appears with an order of COP-Ene, COP-Ppcfe, COP-Nap, and COP-Pyr in square wave voltammetry experiment (Supplementary Fig. 19), which indicates their growing adsorption strengths along with the extension of sp2 carbon skeletons connected to FeN4 sites.
Subsequently, we theoretically preliminarily identified the reliability about above catalytic activity and adsorption strength. As observed, the theoretical Uon−set potential by DFT calculations is consistent with experimental results (Fig. 3d). Specially, the negative Uon−set potentials of COP-Nap and COP-Pyr samples imply that higher electrode potentials are required to overcome the grand barrier and afterward occur four-electron reaction. Meanwhile, the Tafel slopes were observed to surge from COP-Ppcfe to COP-Nap sample (50.42 v.s. 112.65 mV•decade− 1), which mainly was ascribed to the sudden oxytropism in metal sites (Fig. 3c).56 As a result, the sluggish the desorption of OH* (OH* + H+ + e− → H2O) into RDS (vide infra, Supplementary Fig. 20), severely restricting the overall oxygen reduction kinetics. The experimental exchanged current densities (j0ept) directly scaled with theoretic ∆Gmax− 1 (Fig. 3e) also indicates transformation about the RDS.28 Certainly, in addition to the four-electron reaction, O2 also can be reduced to H2O2 through incomplete reduction. Therefore, the tropism of OOH* intermediates towards O*, described by ∆GO* - ∆GOOH*, is a criterion of the selectivity of O2.57, 58 A more negative value indicates that oxygen is more facilely cleaved and completely reduced to H2O. Therefore, as expected, a linear relationship between ∆GO* - ∆GOOH* and H2O2 yields (obtained by rotation ring-disk electrode) can be found (Fig. 3f and Supplementary Fig. 21). Hereto, these results manifest that the oxygen reduction activities of the four COPs exhibit a volcano plot with the expansion of the carbon skeletons regardless in experiment or theory.
The relationship between DDE and activity demonstrated by experiments. Theoretically, the electron-donating capability of electrocatalysts determines the interfacial electronic transfer with oxygen intermediates, with an extensive effect on the oxygen reduction.17, 59 Generally, a catalyst with a small work function is more apt to undergo four proton-electron coupled steps, because electrons are effortless to escape from the catalyst surface.32 According to the Ultraviolet Photoelectron Spectroscopy (UPS) spectra (Supplementary Figs. 22–23), the obtained work function values (Ø) markedly decreased along with the increment of DDE in the carbon skeleton, thus FeN4 sites with larger DDE are easier to react with oxygen60. Moreover, since the valence orbital is involved in the binding with associate oxygen intermediates, we further evaluated the valence band energy level, which are determined to -1.37, -1.36, -1.27, and − 1.02 eV relative to the Fermi level (Ef, Fig. 4a), respectively, for four COP-based catalysts. The obvious upshift of valence orbital reflects an inherent discrepancy of d-orbital level (Ed) along with DDE in the carbon matrix.61 Therefore, the valence band level is employed as a criterion for evaluating the transition of the d-orbital level of the monometallic Fe atom in the synthesized COPs (vide infra).
In fact, the changes in adsorption are a direct cause towards the ORR activity discrepancy. Therefore, we first associated oxygen adsorptions with DDE. As aforementioned, the DDE was obtained based on FWHM of C = C bond (284.7 eV) while oxygen adsorption was quantified by Fe2+/3+ redox potential (Eredox). In order to be clearer and more intuitive, we then employed FWHM− 1 and Eredox−1 to express intensity of DDE and adsorption strength: the larger their value, the stronger their strength. As observed from Fig. 4b, an increasing linear tendency between the Eredox−1 and FWHM− 1 emerges, manifesting that the DDE in the carbon matrix connected to the FeN4 sites is decisive to their adsorption strength. On the other hand, since d-orbital electronic states inherently determined the catalytic activity of FeN4 sites, thereby the relationship between the valence band level and adsorption was next regressed and fitted (Fig. 4c). As a result, lifting valence band level ultimately induces stronger interactions between active sites and oxygen intermediates, which also corresponds to the drastically reduced work function values (Fig. 4a). Therefore, we have speculated that the DDE adjusts oxygen reduction activity largely by changing the valence-band or d-orbital electrons in active centers. As a result, although configurations of FeN4 sites are similar, their electronic states are different and thus their intrinsic activities may also be diverse. Subsequently, we constructed the relationship between TOF and FWHM− 1. We noted that intrinsic catalytic activities in FeN4 catalysts presented an approximate volcano plot with the increment of DDE near FeN4 sites (Fig. 4d). The phenomenon is likely to be ascribed to that the relationship of DDE, d-orbital level, and adsorption are unidirectionally determined in sequence (Figs. 4b-c), while electrocatalytic activity and adsorption are restricted by Sabatier principle (Fig. 4e). Hereto, we speculate that the DDE in the carbon matrix connected to the FeN4 sites affects the d-state level in the single iron atom, and thus tailors the adsorption strength between the moieties and oxygen intermediates; eventually, catalytic activity appears a volcano plot with DDE.
Electronic structure analysis and mechanism study. To comprehensively study the in-depth effect of the adjacent carbon environment on the ORR performance, we accomplished detailed DFT calculations. In addition to the four COPs discussed above, we further in silico designed two more types of asymmetric carbon matrix structures connected to FeN4 sites that are difficult to synthesize experimentally, named as COP-Nap1 and COP-Nap2, to study mechanism of DDE towards FeN4 electronic structures (Fig. 5a). The climbing total electron numbers observed from the total density of states (TDOS, Supplementary Fig. 25) and electronic localization functions (ELF) diagrams (Supplementary Fig. 26) theoretically demonstrated the DDE in carbon matrix is gradually increasing with the expansion of the carbon skeleton connected to FeN4 sites, which also inosculates with the FWHM of conjugated C = C bond peaks experimentally (Fig. 2g-h). The Bader charge population analysis manifests that these carbon matrix functions as a ‘motor’, gradually donating more electrons from themselves into FeN4 sites thereby altering the net charge of the single Fe atom (Fig. 5b), strengthening iron oxophilicity and binding with adsorbates.62
Generally, for most Fe-N-C catalysts, protonation of O2* (O2* + H+ + e- → OOH*) or desorption of OH* (OH* + H+ + e− → H2O) is RDS in oxygen reduction.63 The protonation of O2* is the initial step and directly affects the next electron transfer. Therefore, we initially focused on the changes in the electron states of oxygen adsorption to unveil the effect of the DDE on oxygen reduction. Among six systems, COP-Ene, COP-Ppcfe, COP-Nap1 and COP-Nap2 display an end-on adsorption models, whereas COP-Nap and COP-Pyr exhibit side-on adsorption models (Fig. 5c). For FeN4 sites with abundant DDE, such as COP-Nap and COP-Pyr samples, the Fe atom occupied orbitals transfer more electrons into anti-bonding orbitals of O2 via two O atoms, thereby adsorbed dioxygen adopts side-on configuration on Fe atom (transferred electrons ~ 0.65 |e| in side-on configuration v.s. 0.3–0.4 |e| in end-on configuration). Partial density of state analysis (Fig. 5d) indicates that electrons mainly transfer from 3dxz and 3dyz orbitals in Fe atom to 1π* orbitals in dioxygen through π-back bonding. In contrast, for other COPs, superoxide species form when O2 molecules are absorbed on Fe atom by end-on interaction, where electrons mainly transfer from the Fe 3dz2 orbitals to the oxygen 1π* orbitals and form σ bonds. These completely antithetical manners imply that even if active sites possess semblable geometric structure, their electronic transfer paths and bonding patterns may be completely different (Fig. 5d and Supplementary Figs. 27–28).64
Distinguish from the interaction between Fe atoms and O2*, the 1π valence orbital in lone pair O 2px, 2py electrons, and 3σ orbital in H 1s - O 2pz of OH* intermediate occurred renormalizing when OH* reacted on 3d orbitals of Fe atom (Fig. 5e and Supplementary Fig. 29).65 This coupling makes their energy level split into bonding states under the Fermi level and antibonding states above the Fermi level (Fig. 5f).66 The higher the energy of Ed relative to Ef means the less electronic occupancy in the antibonding states and corresponding stronger adsorption.19 This precise tailor pointing at Ed energy level was also validated by a series of increasing absolute value of integrated crystal orbital Hamilton population value (ICOHP, Fig. 5e).67 Thus, it can be concluded that although the d-orbitals electron in FeN4 sites determines the progress of ORR by affecting oxygen intermediates, π-conjugated ligand configurations connected to FeN4 sites can relocate electronic filling of antibonding states in Fe atom and modulate electronic configurations of FeN4 sites.
Finally, we studied ORR pathways and catalytic activities of the six well-defined catalytic models in oxygen reduction process. Although four-electron/ proton coupling reaction theoretically includes the associative mechanism and the dissociative mechanism, the grand barriers of directly breaking the O = O bond means that the dissociative mechanism for uniform and continuous single-atom catalysts is insurmountable (Supplementary Fig. 20). Nevertheless, the delocalized π electrons in the carbon matrix can change the RDS of the four-electron oxygen reduction reaction (Fig. 6a-b, Supplementary Fig. 30). As we all known, the intrinsic activity of the electrocatalysts can be determined by the limiting reaction barrier in the RDS. For COP-Ene, COP-Ppcfe, COP-Nap1, the RDS is oxygen protonation to OOH* with limiting barriers of 0.64, 0.58 and 0.56 eV, respectively (Fig. 6c). However, when the degree of delocalization in carbon matrix continues increasing, the RDE of oxygen reduction reaction were transformed from protonation of O2* into desorption of OH*, with limiting barrier of 0.78 eV for COP- Nap2, even as large as 1.24 and 1.37 eV for COP-Nap and COP-Pyr. In addition, the free energy of the six COP models is more negative, successively suggesting that the DDE in carbon matrix induce a stronger chemical adsorption with oxygen intermediates, which is also the direct reason for the transition of RDS. By correlating the onset potential and Gibbs free energy of OH* intermediates, a typical volcano plot was constructed. Notedly, the ‘JUST’ DDE in carbon matrix connected to FeN4 sites makes COP-Ppcfe, and COP-Nap1 have the optimum d-state level, thereby exhibit ‘RIGHT’ adsorption, and lie at the apex of volcano plot (Fig. 6d). Therefore, our calculation results manifest that the carbon matrix connected to FeN4 sites remarkably contributes much to the catalytic activity of electrocatalysts and the RDS in oxygen reduction processes.