Theoretic calculations. The structure models of the pristine NiPc and NiPc with substituents of hydroxyl (NiTHPc) and amino (NiTAPc) are illustrated in Fig. 1a. In order to explore the substituent effect on the NiPc system, the electronic density of Ni was analyzed (Supplementary Fig. 1). Hrishfeld and Mulliken charge of Ni (Fig. 1b) indicates that the electron can be localized on Ni atom by the electron-donating substituents of hydroxyl or amino, and NiTAPc exhibits more apparent electronic localization than that of NiTHPc due to the stronger electron-donating ability of amino. The CO2 adsorption energy of molecules was also examined (Fig. 1c and Supplementary Fig. 2). NiTAPc shows the best CO2 adsorption ability (0.228 eV) in contrast to those of NiTHPc (0.22 eV) and NiPc (0.218 eV). To understand the CO2RR performances of those molecule catalysts, the free energies of the CO2RR pathway were calculated (Fig. 1d). The rate-determining step on all of the molecule catalysts is the step of CO2 activation to *COOH. The substituted NiPc with an electron-donating substituent of hydroxyl or amino can lower the free energy of *COOH. NiTAPc with the lowest energy barrier for CO2 activation manifests the rapid process for CO2RR in comparison to the NiTHPc and NiPc. According to these theoretic results, the electron-donating substituent of hydroxyl or amino-modified NiPc can induce the electronic localization of Ni site and thus enhances CO2 adsorption and activation to boost CO2RR.
Molecule catalysts synthesis and characterization. To synthesize the proposed molecule catalysts, a facile solvothermal reaction was conducted by employing the substituted phthalonitrile and nickel acetate as the reactants (see Methods for details). Matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF MS) of as-prepared molecule catalysts (Fig. 2a) shows the mass of molecule ion at 630 and 634 from NiTHPc and NiTAPc, respectively. This result confirms the successful synthesis of the substituted NiPc (Fig. 1a). Fourier transform infrared (FTIR) spectra (Fig. 2b and Supplementary Fig. 3) prove the presence of hydroxyl and amino in NiTHPc and NiTAPc, respectively, without a characteristic peak of phthalonitrile at around 2230 cm-1. Ultraviolet-visible (UV-Vis) spectroscopy was used to characterize the NiPc, NiTHPc and NiTAPc owing to the special optical property of corresponding substituted NiPc. As shown in Fig. 2c, the substituted NiPc exhibits the characteristic Q band absorption at around 600–800 nm and B band absorption,36 without the absorption of nickel acetate (Supplementary Fig. 4). Furthermore, the Q band absorption of π→π* is attributed to the electron transition from the highest occupied molecule orbital (HOMO) to the lowest occupied molecule orbital (LOMO) of the phthalocyanine ring.37 The Q band absorption without splitting indicates the symmetrical substituent position of the synthesized molecule.38 There is a redshift in the Q band absorption among NiTHPc and NiTAPc due to the increased electron-donating ability of amino,39-41 which is highly matches the results of theoretic calculations (Supplementary Fig. 5). X-ray powder diffraction (XRD) patterns (Fig. 2d) show similar signal peaks of all samples below 10o, suggesting an analogous molecule structure of the fabricated samples. Raman spectra of NiPc, NiTHPc and NiTAPc (Supplementary Fig. 6) exhibit the same sensitive band at 1545 cm-1, which can be caused by the Ni interaction with the phthalocyanine ring.42,43 Based on the above-mentioned results, we have demonstrated the successful preparation of the target molecule catalysts of NiTHPc and NiTAPc.
To further investigate the structure-dependent electronic effects of molecules, XPS and XAS were measured. From the XPS survey scan and N 1s spectra of NiPc, NiTHPc and NiTAPc in Supplementary Fig. 7 and Supplementary Note 1, the amino and hydroxyl are successfully substituted in NiTHPc and NiTAPc, respectively.23 The C and N K-edge of molecules were measured (Supplementary Fig. 8 and Supplementary Note 2). The substituents have no measurable effects on the C K-edge spectra of the molecules. The N K-edge spectra of NiTHPc and NiTAPc can be greatly influenced by the electron-donating substituents. The Ni 2p spectra of NiPc, NiTHPc and NiTAPc (Fig. 3a) exhibit similar signal peaks, which can be indexed to the Ni confined to the ring of phthalocyanines. Noticeably, the binding energy of Ni 2p has a negative shift in NiTHPc and NiTAPc compared with that of the pristine NiPc. Accordingly, XAS spectra of Ni L-edge (Fig. 3b) also show a negative shift of absorption peaks at NiTHPc and NiTAPc in contrast to the counterpart of the pristine NiPc. NiTAPc displays a remarkable shift of absorption energy compared with that of NiTHPc owing to the preferable electron-donating ability of amino. To further confirm the electronic localization of the central Ni atom, X-ray absorption near-edge spectra (XANES) were conducted. XANES spectra of Ni K-edge (Fig. 3c) show a negative shift in NiTHPc and NiTAPc compared to that of the traditional NiPc, which demonstrates that the electron of phthalocyanine ligands could be localized on the Ni site by electron-donating substituents. NiTAPc shows the most apparent electronic localization among the NiPc-based molecules, in good agreement with the results of XPS and XAS of Ni L-edge. In addition, extended X-ray absorption fine structure (EXAFS) spectra of Ni K-edge (Fig. 3d) and the wavelet transform plot (Supplementary Fig. 9) demonstrate the same coordination number of Ni in those catalysts. Thus, the electronic localization on the Ni site is derived from the peripheral substituents rather than the changed coordination number of Ni. These results indicate that the substituents of hydroxyl or amino with electron-donating ability can arouse the electronic localization on the central Ni atom of molecules, and the degree of electronic localization is positively correlated with the electron-donating ability of substituents.
To improve the conductivity of molecules for electrocatalytic CO2RR, NiPc, NiTHPc and NiTAPc were dispersed onto carbon nanotube to form catalysts of NiPc/CNT, NiTHPc/CNT and NiTAPc/CNT. Scanning electron microscope (SEM) in Supplementary Fig. 10 and the elemental mapping (Supplementary Fig. 11) from high-angle annular dark field-scanning transmission electron microscope (HAADF-STEM) demonstrate that the molecules can be well dispersed by carbon nanotubes. The results of XRD and Raman from the composite catalysts (Supplementary Fig. 12) suggest the presence of carbon.44 XANES spectra of Ni K-edge from molecule catalysts and its mixture with the conductive carbon (Supplementary Figs. 13a-c) indicate the stable electronic structure of Ni. Accordingly, EXAFS spectra of Ni K-edge (Supplementary Fig. 13d) manifest the NiPc-like structure of catalysts. It can be concluded that molecule catalysts are well dispersed onto carbon nanotubes, and they maintain the electronic structure of Ni sites in composite catalysts.
CO2 adsorption measurements. To investigate the CO2 adsorption ability of different molecule catalysts, the CO2 adsorption measurement was performed (Supplementary Fig. 14) and normalized by the surface area from N2 adsorption/desorption analysis (Supplementary Table 1). The pristine NiPc with the electron-deficient Ni site displays negligible CO2 adsorption (Fig. 4a). On the other hand, NiTHPc and NiTAPc possess the electronic localization of Ni sites that are induced by the electron-donating substituents, and thus both show a significantly enhanced ability of CO2 adsorption. NiTAPc shows the optimal CO2 adsorption due to its conspicuous electronic localization among the three of NiPc-based catalysts. Furthermore, the CO2 adsorption responses of NiPc/CNT, NiTHPc/CNT and NiTAPc/CNT were also explored (Supplementary Fig. 15). N2 was used as the control gas to study the CO2 adsorption responses. NiTAPc/CNT exhibits a remarkable response in the CO2 atmosphere (Fig. 4b), outperforming the moderate response of NiTHPc/CNT and the inferior response of NiPc/CNT. These consequences indicate that the enhanced CO2 adsorption is enabled by the electronic localization of Ni sites, in accordance well with the results of computational simulations.
CO2RR performance. To evaluate the catalytic performance of developed catalysts, the electrochemical measurements were carried out in the typical H-cell. Linear sweep voltammetry (LSV) curves in the CO2 and Ar-saturated 0.5 M KHCO3 electrolyte (Supplementary Fig. 16) demonstrate the composite catalysts with apparent catalytic activities for CO2RR. However, the blank carbon nanotube shows no activity for CO2RR (Supplementary Fig. 17). Electrochemical active surface area (ECSA) measurements (Supplementary Fig. 18) indicate a similar active surface of composite catalysts. Therefore, the catalytic performance results from the intrinsic activity of molecule catalysts. From the LSV in the CO2-saturated electrolyte (Fig. 3a), the amino-substituted NiPc with superior electronic localization of Ni site exhibits the best performance of CO2RR than those of other catalysts. The electron-donating substituents of hydroxyl and amino can greatly improve the catalytic activity of NiPc. To obtain the product distribution at different potentials of catalysts, chronoamperometry measurements (Supplementary Fig. 19) were carried out to detect the products of CO2RR. Carbon monoxide (CO) can be detected as the only product from the results of gas chromatograph (Supplementary Fig. 20) and 1HNMR spectroscopy (Supplementary Fig. 21). Fig. 3b shows the faradaic efficiency of CO (FE(CO)). As a byproduct of CO2RR, the FE of hydrogen (H2) is also displayed in Supplementary Fig. 22. It can be seen that the FE(CO) of catalysts maintain a level of above 90% in the potential range of 0.58–0.90 V. NiTHPc/CNT and NiTAPc/CNT demonstrates the enhanced selectivity of CO for CO2RR in comparison to the NiPc/CNT, which can be attributed to electron-donating substituents induced electronic localization of Ni site. Although a high selectivity of CO in NiPc/CNT can be achieved, the stability (Fig. 3c) is much lower than those of NiTHPc/CNT and NiTAPc/CNT, undergoing a fast degradation during the continuous CO2RR catalysis.
NiTAPc/CNT shows a high activity of −12.2 mA cm-2 at −0.58 V and remarkable CO selectivity of almost 100%, exceeding the other molecule catalysts and even the most performances of representative in literatures (Supplementary Table 2). The electrochemical impedance spectrum (EIS) was employed to explore the reaction dynamics of catalysts during CO2RR. As exhibited in Fig. 3d, the charge transfer resistance of catalysts is ranked as the order: NiPc/CNT > NiTHPc/CNT > NiTAPc/CNT, which demonstrates the fastest reaction dynamics of NiTAPc/CNT. The substituted NiPc with electron-donating groups encourages the electronic localization of the Ni atom and therefore promotes CO2RR performance in terms of activity, selectivity and stability.
To study CO2RR performances at high current density, the composite molecule catalysts were supported on the gas diffusion layer (GDL) to form gas diffusion electrodes. In this way, a fast CO2 diffusion can be obtained via the gas phase to the catalytic sites, which overcomes the limited mass transfer in H-cell. The schematic illustration of the flow cell is displayed in Supplementary Fig. 23. The pristine NiPc/CNT loaded on the GDL shows the stable CO2RR performance from LSV measurements in CO2, but this catalyst suffers from serious degeneration during LSV tests in the Ar without CO2 atmosphere (Supplementary Fig. 24a). Furthermore, NiPc/CNT with adequate CO2 in the flow cell displays a stable potential at a low current density of −60 mA cm-2 (Supplementary Fig. 24b). When the current density increased at −200 mA cm-2, the performance of NiPc/CNT is dramatically declined due to limited CO2 adsorption (Fig. 4a). Thus, the inferior stability of the pristine NiPc is rooted in the poor CO2 adsorption of the active site. The GDL coated with NiTAPc/CNT or NiTHPc/CNT can be operated at the current densities from −20 to −500 mA cm-2 at applied potentials less than −0.80 V vs. RHE (Fig. 4a). An ultrahigh current density of –500 mA cm-2 can be obtained from NiTAPc/CNT with the FE(CO) of 94.6%. NiTAPc/CNT displays a remarkable performance with high FE(CO)>99.8% between the current densities from –20 to –400 mA cm-2 (Fig. 4b), which is much better than the counterparts of other NiPc catalysts and even the reported catalysts (Supplementary Table 3). Accordingly, the turnover frequency of CO (TOFCO) from various catalysts (Fig. 6c) is calculated according to the results of inductively coupled plasma-optical emission spectrometry (ICP-OES) in Supplementary Table 4. The optimal catalyst of NiTAPc/CNT can achieve a remarkable TOFCO of 41.9 s−1, outperforming that of NiTHPc/CNT (28.0 s−1) and NiPc/CNT (10.0 s−1). To obtain the stability measurement of NiTAPc/CNT, the CO2RR performance was conducted at −150 mA cm-2. The catalysts can maintain 12, 000 s with the selectivity of CO over 99.8% (Fig. 6d). On the basis of these electrochemical data, the NiTAPc/CNT delivers most superb performance for CO2RR among the molecule catalysts.
To further verify the significance of the electronic localization of Ni site for CO2RR, the tert-butyl, a typical electron-withdraw substituent in the phthalocyanine,29,45 to substituted NiPc (NiTBPc) was synthesized (Supplementary Fig. 25). MALDI-TOF MS spectrum, FTIR spectra, UV-Vis spectrum and Raman spectrum of NiTBPc (Supplementary Fig. 26) prove the successful preparation of NiTBPc. SEM and HAADF-STEM characterizations (Supplementary Fig. 27) suggest that NiTBPc can be well dispersed onto the carbon nanotubes. NiTBPc possesses an interior electron-withdraw ability of tert-butyl without an obvious shift in the binding energy of Ni 2p and the absorption energy of Ni L-edge (Supplementary Figs. 28a and b). However, XANES spectra of Ni K-edge in NiTBPc (Supplementary Fig. 28c) show a positive shift of near-edge absorption compared with that of NiPc, and this shift can be more obvious (Supplementary Fig. 28d) after dispersing the molecule catalysts onto carbon nanotubes, indicating that there is an electronic delocalization on the Ni site in NiTBPc/CNT when compare to NiPc/CNT. EXAFS spectrum and wavelet transform plot of NiTBPc (Supplementary Figs. 29) demonstrate the similar structures of NiPc. As expected, from the H-cell performance of CO2RR, NiTBPc/CNT with poor stability exhibits low activity and selectivity in contrast to NiPc/CNT (Supplementary Fig. 30). In flow cell, the activity and selectivity of NiTBPc/CNT are lower than the counterparts of NiPc/CNT (Supplementary Fig. 31) due to its electron-withdraw substituent of tert-butyl. Thus, the electron-withdraw substituent that induces the electronic delocalization of Ni site can inhibit the catalytic activity of NiPc for CO2RR. These conclusions further demonstrate that the electronic localization of the Ni site plays a significant role in boosting CO2RR.