Electrocatalytic reduction of CO2 under ambient conditions using renewable energy has emerged as an attractive way to maintain carbon balance, which is regarded as one of the cleanest and efficient approaches1–3. Recently, numerous electrocatalysts have been developed for CO2 reduction reaction (CO2RR)4–7. However, the majority of these catalyst systems have low efficient for multicarbon (C2+) products due to the multiple pathways involved in the reaction process8. Cu-based catalysts are mostly reported to convert CO2 into C2+ products such as ethylene, alcohols and acetic acid9–11. Up to now, non-copper catalysts are rarely reported to generate C2+ products12,13.
Iron-based catalysts have been studied extensively for CO2RR14–17. The reported Fe-based catalysts, such as atomically dispersed Fe − N−C, single-crystalline Fe3N, Fe-porphyrin carbon materials (FeN4-C), mainly produce CO with high FE up to 90%18–20. The element dopants together with the local coordination environment of Fe center play remarkable roles in affecting the electronic structures and catalytic performances21. Currently, there have been few reports about selective formation of multicarbon products over Fe based catalyst. A FeP nanoarray was reported to be capable of producing C2H5OH with FE of 14.1% and a low current density (2 mA cm− 2)13. In addition, Fe2P2S6 nanosheet was also reported with a maximum FE for C2H5OH of 23.1% and very low current density (< 0.5 mA cm− 2)22. It is highly attractive but still remains a big challenge for electrocatalytic CO2RR to multicarbon (C2+) products efficiently over non-copper catalysts.
In this work, we designed and prepared Fe-based moieties behaving core-shell architecture inlaid with nitrogen (N) doped γ-Fe2O3 and wrapped with carbonitride (CN) as shell. The as-prepared catalysts containing Fe-O-N sites with N coordination and oxygen defects exhibited outstanding performance toward selectivity for electrochemical conversion of CO2 to C2H6.
Preparation and characterization of the iron-based catalysts.
We designed a pre-assemble strategy to fabricate iron-based catalysts through the in situ polymerization of dopamine (DA) in a solution with Fe source (Fe3+) and Tris (Tris(hydroxymethyl)aminomethane) buffer, using Pluronic F127 as template. After calcination of the as-prepared nanoparticles under N2 atmosphere, the N doped Fe2O3 with CN shell were obtained, as illustrated in Fig. 1a. The iron-based moieties were denoted as xFe2O3-N@CN, where the subscript x corresponds to the mole fraction of Fe in all the elements after carbonization, according to the results obtained by the inductively coupled plasma optical emission spectroscopy (ICP-OES). A series of catalysts with different Fe content, 0.05Fe2O3-N@CN, 0.11Fe2O3-N@CN, 0.17Fe2O3-N@CN, 0.2Fe2O3-N@CN and 0.3Fe2O3-N@CN were prepared by this method via controlling the amounts of Fe source (Fe2(SO4)3) used.
The scanning electron microscopy (SEM) images showed that the xFe2O3-N@CN exhibited core-shell structure with size distribution of 300 ~ 400 nm (Supplementary Fig. 1). As showed in Supplementary Fig. 2, the X-ray diffraction (XRD) patterns of the as-prepared xFe2O3-N@CN were consistent with the characteristic peaks of γ-Fe2O3 (PDF # 39-1346). The relative diffraction intensity of Fe2O3 peaks in the XRD patterns was enhanced by the increase of iron content. A fine study on 0.3Fe2O3-N@CN with the highest content of Fe atom was carried out by using the transmission electron microscopy (TEM) and SEM, and the images indicated the core-shell architecture of the catalyst with a shell thickness of about 30 nm (Fig. 1b and Supplementary Fig. 3). The catalyst of 0.3Fe2O3-N@CN showed an intermediate mode between type I and type IV, indicative of the mesoporous and microporous features23. Brunnauer–Emmett–Teller (BET) surface area (SABET: 248 m2 g ) analysis and pore distribution were presented in Supplementary Fig. 4. In the high-resolution TEM (HRTEM) images (Fig. 1c), the emerged fringe spacings of 0.208 nm and 0.295 nm matched well with the (400) and (220) lattice planes of Fe2O3. The scanning transmission electron microscopy (STEM) image indicated that the Fe2O3 nanoparticles were coated by CN shell with rough surface. The associated energy dispersive X-ray spectroscopy (STEM-EDX) elemental mapping analysis (Fig. 1d) demonstrated that the existence of Fe, C, O and N elements in the core and shell, with Fe element as the primary component, which results were approximately equal to that of ICP-OES (Fe 65.7 wt%). In addition, the EDX analysis also displayed the distribution of each components (Supplementary Fig. 5), which showed an obvious separation interface between the N-doped Fe2O3 in the core and C, N elements in the shell section.
To explore the electronic states and local coordination environments of Fe atoms in the xFe2O3-N@CN, the X-ray absorption near edge structure (XANES) and X-ray photoelectron spectroscopy (XPS) measurements were conducted. The Fe K-edge XANES curves of the as-prepared xFe2O3-N@CN demonstrated near-edge absorption energy between those of Fe foil and Fe2O3, implying that the oxidation state of Fe was Fe3+ (Fig. 1e)24,25. The Fourier transform (FT) curves at the Fe K-edge of extended X-ray adsorption fine structure (EXAFS) spectra were further fitted to investigate the coordination information of Fe (Supplementary Fig. 6). The FT-EXAFS curves displayed a broad peak at about 1.4 ~ 1.7 Å which could be attributed to Fe–O/N coordination, and the peak around 2–3 Å in the FT-EXAFS was assigned to the Fe–Fe distance in Fe oxides26,27, as illustrated in Fig. 1f. It is ambiguity to distinguish the Fe-O and Fe-N coordination in the EXAFS data28. Moreover, the peak around 2.22 Å was observed in both samples of 0.05Fe2O3-N@CN and 0.11Fe2O3-N@CN, which verified the coexistence of zero-valent iron crystalline phase (Fe0)29, but no Fe–Fe bond derived from Fe0 was detected in xFe2O3-N@CN with higher content of Fe (x ˃ 0.11). Quantitative EXAFS fitting analysis (Supplementary Table 1) demonstrated that all the coordination numbers of Fe-O/N in xFe2O3-N@CN were lower than that of Fe2O3 reference, suggesting the unsaturated coordination of Fe-O/N. The XPS spectra for xFe2O3-N@CN also showed the existence of Fe, N, O and C elements (Supplementary Fig. 7a). The Fe 2p peaks (Supplementary Fig. 7b) were attributed to the status of Fe3+ (712.7 and 724.8 eV), which was identical with the results of XANES. The high-resolution N 1s spectra of xFe2O3-N@CN (Fig. 1g) were mainly assigned to four components corresponding to pyridinic N (398.1 eV), pyrrolic N (400.8 eV), Fe-N coordination with pyridinic (399.6 eV) and pyrrolic (398.6 eV)30. The graphitic N (402.1 eV) existed only in xFe2O3-N@CN with much lower mole fraction of Fe, and no peak of graphitic N was observed in samples of 0.2Fe2O3-N@CN and 0.3Fe2O3-N@CN. The O 1s spectra of all catalysts (Supplementary Fig. 7c) exhibited typical peaks at ~ 529.8 eV, 531.5 eV and 533.5 eV, which were assigned to lattice oxygen, oxygen vacancy, and surface-adsorbed oxygen species, respectively31. Notably, the oxygen vacancy (531.5 eV) occupied noticeably larger proportion in the O 1s components, which showed an increasing trend with the increase of Fe content. In brief, no obvious difference was observed in the XRD patterns, XANES and XPS spectra of xFe2O3-N@CN with high content of Fe, demonstrating virtually identical iron oxynitrides in 0.17Fe2O3-N@CN, 0.2Fe2O3-N@CN and 0.3Fe2O3-N@CN. Based on the above analysis, the as-prepared catalysts possessed N-doped γ-Fe2O3 centers with different coordination of Fe-O-N structures and rich oxygen defects.
Atomic structure analysis of xFe2O3-N@CN by using 57Fe Mössbauer spectroscopy.
57Fe Mössbauer spectroscopy is sensitive to iron species of similar coordination environments but with different electronic states32,33. To better identify the local coordination of Fe species in the catalysts, 57Fe Mössbauer spectroscopy measurements were conducted to distinguish the detail structures of Fe-based catalysts. The Mössbauer spectra of 0.05Fe2O3-N@CN, 0.17Fe2O3-N@CN, 0.2Fe2O3-N@CN and 0.3Fe2O3-N@CN were collected and analyzed. The spectrum of 0.05Fe2O3-N@CN was fitted with one doublet and three types of sextets (Fig. 2a). According to the isomer shift (IS) and quadrupole splitting (QS) values (Supplementary Table 3), the doublet (doublet 1, D1) was assigned as characteristic for the presence of square-planar FeN434,35. It is well known that N atoms are very mobile to diffuse into deeper layers forming a magnetic oxynitride, in which the N content is close to that in the initial iron nitrides36. Sextet 1 could be identified as iron oxynitride phase, FeOmN4−m (0.2 ≤ m ≤ 0.58) (Supplementary Text) centers enriched with N, which formation might be attributed to the diffusion of N atoms from the surface region of FeN4 into deeper layers37. In addition, the hyperfine field (H) parameters of FeOmN4−m agreed well with the values of iron nitrides given in the literatures36–39. Signals of sextet 2 (IS -0.11 mm/s, QS 0.03 mm/s) and sextet 3 were assigned to γ-Fe2O3 and ɑ-Fe, respectively40, which was in agreement well with the results of XRD and EXAFS.
The increase of Fe content resulted in significant variations in the Mössbauer spectra. As illustrated in Fig. 2b and Supplementary Figs. 8a and 8b, a second doublet (D2) and another different sextet (sextet 3) signals emerged in the Mössbauer spectra of xFe2O3-N@CN with Fe mole fraction higher than 0.17. The D1 and D2 in the Mössbauer spectrum of 0.3Fe2O3-N@CN were assigned to square-planar FeN4 coordination with Fe in low spin states with different iron position41. The extra sextet 3 was assigned to a O-rich composition FeO1.5−nNn (0.34 ≤ n ≤ 0.55) (Supplementary Text) with small amount of N incorporated into Fe2O3. The discrimination derived from the combination of isomer shifts and hyperfine field (H) was significantly different from that of FeOmN4−m with N-rich (Supplementary Table 3)34,42. Nitrogen atoms are more mobile at some elevated temperature, and part of them gradually diffuse into deeper layers, which resulted in the formation of O-rich FeO1.5−nNn phase, leading to O vacancies simultaneously36. A comparison between these catalysts with low and high content of Fe displayed totally different signals in the Mössbauer spectra (Fig. 2c). The spectra of 0.17Fe2O3-N@CN, 0.2Fe2O3-N@CN and 0.3Fe2O3-N@CN exhibited similar signals signifying the presence of identical iron oxy-nitride structures. The signals attributed to γ-Fe2O3 (sextet 2), FeN4 (D1, D2), FeOmN4−m (sextet 1) and FeO1.5−nNn (sextet 3) were all visible by the identical features in Mössbauer spectra of xFe2O3-N@CN with high loading of Fe element. We focus on the variation of resonance area to estimate the different iron oxy-nitride centers in depth (Supplementary Table 3). The γ-Fe2O3 composition exhibited slight increase ranging from 20.6–30.5% with increasing iron content. Interestingly, it could be observed that both of the contents of FeN4 and FeOmN4−m decreased obviously with the increase of iron content in xFe2O3-N@CN. In contrast, the proportion of FeO1.5−nNn sites displayed a remarkable tendency of increase (Fig. 2d), along with the generation of more oxygen defects (Supplementary Fig. 7c). In brief, the coordination environment varied with iron content, and there were more O-rich FeO1.5−nNn sites and oxygen defects in the catalysts with higher Fe content.
Electrocatalytic CO2 RR over the various catalysts.
The catalytic CO2RR performance of 0.3Fe2O3-N@CN with the highest Fe content was evaluated in a three-electrode system in an ionic liquid (IL) based electrolyte (1-Butyl-3-methylimidazolium hexafluorophosphate ([Bmim]PF6/acetonitrile (MeCN)/H2O (W/W/W = 30/65/5)). The linear sweep voltammetry (LSV) curves of 0.3Fe2O3-N@CN in N2- and CO2- saturated [Bmim]PF6/MeCN/H2O electrolyte were presented in Fig. 3a. In the CO2- saturated electrolyte, a much higher current density was generated over 0.3Fe2O3-N@CN, indicating the efficient activity of CO2RR. Controlled potential electrolysis of CO2 was conducted to study the reduction products at potentials ranging from − 1.9 to -2.6 V vs Ag/Ag+. Gas-phase and liquid products were analyzed by gas chromatography (GC) and 1H NMR spectroscopy, respectively. Only gaseous products were detected in this study, and no liquid product was generated. The products were quantified after electrolysis for 2 hours. C2H6 was the sole product from CO2 at − 2.0 V vs Ag/Ag+, and hydrogen was the only by-product (Fig. 3b). The FE of C2H6 was 42% at − 2.0 V vs Ag/Ag+ with a current density of 32 mA cm− 2. As the potential became more negative, the FE of C2H6 decreased gradually, accompanied with the production of CH4 and CO. Control experiments were performed at the same conditions to confirm that the products were indeed generated by electrochemical reduction of CO2 over 0.3Fe2O3-N@CN, and there was no carbonous product at any potential in N2-saturated electrolyte. Subsequently, the catalytic performances of the catalysts with various Fe content were further evaluated systematically. All the as-prepared xFe2O3-N@CN displayed much higher current density in CO2-saturated electrolyte (Supplementary Fig. 9). Catalysts of 0.05Fe2O3-N@CN and 0.11Fe2O3-N@CN with relatively low content of Fe element produced CO and CH4 at all applied potentials, as illustrated in Supplementary Figs. 10a and 10b. Notably, FEC2H6 and the current density increased as a function of Fe content in the xFe2O3-N@CN (Figs. 3c and 3d). At -2.0 V vs Ag/Ag+, the FEC2H6 over 0.17Fe2O3-N@CN, 0.2Fe2O3-N@CN, and 0.3Fe2O3-N@CN were 22%, 27%, and 42%, respectively. The catalysts with low loading of Fe element produced C1 products such as CO and CH4. However, the high content of Fe3+ in xFe2O3-N@CN led to high FE of C2H6 and current density. In other words, Fe content and the resulting Fe-O-N coordination were the primary differences in the as-prepared xFe2O3-N@CN, which determined the selectivity for generating C1 or C2 products in CO2RR. The porous CN shells of catalysts exhibited unconspicuous role in the selectivity of products, which could prevent the aggregation of the magnetic iron-based materials. A combination of the atomic structures obtained by Mössbauer analysis and catalytic performances of the various xFe2O3-N@CN demonstrated that FeO1.5−nNn sites with small amount of incorporated N favored the generation of C2H6 in CO2RR. Consequently, we can deduce that the active sites are related to the structure of FeO1.5−nNn.
The Nyquist plots under the open circuit potential in CO2-saturated electrolyte were measured to explore the properties of the electrode/electrolyte interface. The experimental impedance data were fitted by using an equivalent circuit R(C(R(Q(RW)))) (Supplementary Fig. 11 and Supplementary Table 2)43. Catalyst of 0.3Fe2O3-N@CN showed the lowest interfacial charge transfer resistance (Rct) (Fig. 3e), indicating the facile charge transfer on the catalyst. The FEC2H6 and current density showed no obvious change after 10 hours of electrolysis, illustrating the excellent stability of 0.3Fe2O3-N@CN (Fig. 3f). Meanwhile, the quasi-operando XPS spectra acquired after reaction over 0.3Fe2O3-N@CN displayed that Fe3+ in the as-prepared materials maintained the oxidation state at all potentials (Supplementary Fig. 12), verifying the stability of the catalyst.
To further test the above hypothesis, we also prepared a sample by pyrolysis of 0.3Fe2O3-N@CN under air at 600°C to change the N coordination (denoted as Fe2O3-CNF). The XRD pattern for the material showed characteristic peaks of ɑ-Fe2O3, indicating a transformation of configuration for ferric oxide centers (Supplementary Fig. 13a). No core-shell structure was observed in the morphology of Fe2O3-CNF (Supplementary Figs. 13b and 13c). Trace amounts of N (< 0.03 at %) was observed in the XPS spectra (Supplementary Fig. 14). Notably, the lattice oxygen escalated dramatically and oxygen vacancy decreased after the calcination in air (Supplementary Fig. 14b). Whereas only two types of sextet assigned to Fe2O3 (Supplementary Fig. 15 and Supplementary Table 4) were observed in the Mössbauer spectrum of Fe-CNF, and no signal about the coordination of Fe-O-N was observed. The electrocatalytic performance of Fe-CNF without FeO1.5−nNn sites exhibited only C1 productions (CO, CH4) with a maximum FE of 68%, but C2 products were not detected (Supplementary Fig. 16). The results further indicated that the FeO1.5−nNn sites with oxygen vacancy in the catalysts of xFe2O3-N@CN were crucial for the generation of C2H6. Meanwhile, the polydopamine during preparation in this method provided CN shell and N coordination with iron species to obtain the particular Fe-O-N sites.
DFT studies on the iron oxy-nitrides (FeO1.5−n Nn) for CO2 reduction.
Density functional theory (DFT) calculations were carried out to unveiling the electronic properties of Fe species and the intrinsic property of active species in iron oxy-nitrides (FeO1.5−nNn) for CO2 reduction to ethane44,45. According to the XAFS results and the different N, O coordination with iron species analyzed by 57Fe Mössbauer spectroscopy, two catalyst models of FeO1.5−nNn and pristine Fe2O3 were constructed for the calculations. Fe2O3 (400) surface was considered based on the HRTEM and XRD results. The energy profiles of the most possible route for the activity of CO2 reduction over N-doped Fe2O3 were presented, as shown in Fig. 4. All the corresponding structures of the two models are also given in Supplementary Fig. 17. The existence of oxygen vacancy in FeO1.5−nNn sites leaded to the emergence of two adjacent surface of Fe atom, of which configuration existed in the lowest free energy (Supplementary Fig. 17c). The first hydrogenation of *CO2 preferred to take place on the emerged Fe atom (*COOH) with ΔG = 0.20 eV. Further, the *COOH intermediate was adsorbed and stabilized at the same surface. Subsequently, the energy barrier (1.37 eV) in the rate-determining step of the C − C coupling process (*CO*CO + H++e−→*COCOH) over FeO1.5−nNn sites was obviously lower than that on catalyst of pristine Fe2O3 (1.99 eV). Then the hydrogenation process of *COHCOH to *CCOH and *CHCOH to *CCH were downhill pathway with ΔG = 0.07 and 0.05 eV respectively. Thereafter, the rest hydrogenation steps involving exothermic or endothermic process were facile with decreased energy barrier (Fig. 4). However, the hydrogenation of *COHCOH to *CCOH was difficult with ΔG = 1.33 eV over pristine Fe2O3, also implying that the C − C coupling process was not favorable. The electronic characteristics of Fe species were regulated by the different coordination environments, which inevitably affected the adsorption properties of intermediates46. As illustrated in Supplementary Fig. 18, The density of states (DOS) of FeO1.5−nNn were distinctly different from that of pristine Fe2O3. The N-doped Fe2O3 centers exhibited a shifting of the 3d orbitals of central Fe indicative of the manipulation of electronic characteristics of Fe atom due to the N doping. The varied electronic properties in the Fe-O-N sites benefited the donation and back-donation of electrons between the catalyst and the reactant, which satisfied the adsorption and stabilization of *CO intermediate (1.09 eV) produced in the hydrogenation process of CO2 (Supplementary Fig. 19). Additionally, with regards to Fe2O3, the desorption of *CO is facile with lower energy (-0.07 ev), demonstrating high selectivity for C1 products. However, the difficult desorption of *CO is conducive to the proceeding of C-C coupling reaction.