We focused on the single-atom metal and N doped carbon nanomaterials (M-N-C) due to their tunable porosity, N dopant, coordination environment of isolated metal-N sites, as well as robust structure.[9] These features endowed M-N-C with great potentials in catalysis, separation and purification, which involved the key step of adsorption. Especially, many studies had demonstrated that the bonding energy of graphitic N in M-N-C generally lied in 400.9~401.3 eV, [9c, 10] comparable with that for the pyridinium and ammonium ion,[11] where N was positively charged in both cases. This implied a positive charge feature of graphitic-N.[12] Apart from graphitic-N, the pyridinic and/or pyrrolic N often existed in M-N-C, which could be protonated into the pyridium and pyrrolium cations, respectively, due to their basic feature. In theory, incorporating positive-charged graphitic, pyridium or pyrrolium N into the π-electron-rich carbon support tended to induce polarization effect, thus modulating the local ligand and electric field, which probably improved the adsorption barrier and kinetics.
By employing a facile annealing of Cu-based ionic liquid strategy at the temperature of 400-1000 oC,[13] we prepared the carbon nanomaterial with isolated Cu-N, graphitic and protonated N sites. After pyrolysis, the obtained material was further treated by nitric acid to remove the possibly aggregated Cu-related particles. The XRD pattern displayed that when the pyrolysis temperature attained 1000 oC, the obtained sample obviously contained Cu-related particles (Figure S1). To focus on the single-atom Cu-N-C with optimum adsorption performance, we thus mainly investigated the protonated Cu-N-C prepared at 800 oC (denoted as Cu-N-C-800), which was taken as a representative sample to analyze its physic-chemical structure.
The scan electron microscope (SEM) images disclosed the bulk texture of Cu-N-C-800 with the rough surface (Figure S2). The transmission electron microscope (TEM) also demonstrated this bulky morphology and observed the obvious nanopores (Figure S3). To further identify the existence form of Cu, we conducted spherical aberration corrected high-angle annular dark-field scanning TEM (HAADF-STEM) observation. The images at several different areas directly observed numerous isolated Cu single atoms with an average size of 0.15 nm (Figure 1a and Figure S4), in according well with the X-ray diffraction (XRD) result, which revealed that there were no obvious Cu-related crystalline peaks in Cu-N-C-800 except two graphitic carbon peaks (Figure S5). The Cu along with N and C was well distributed in Cu-N-C-800, as confirmed by the energy-dispersive X-ray spectroscopy (EDS) mapping (Figure 1b). Moreover, according to the ICP analysis, we determined the Cu content of 0.82 wt.% in Cu-N-C-800. In addition, the N2 adsorption–desorption experiment confirmed that Cu-N-C-800 was a micro-mesopore-dominated carbon with a high surface area (1008.2 m2/g) (Figure 1c), which were expected to promote the exposure of adsorption sites and accelerate mass transfer during adsorption.
The X-ray photoelectron spectrometry (XPS) was further applied to investigate the chemical state of Cu and N in Cu-N-C-800. The Cu2p XPS spectrum observed the obvious Cu2+ species except Cu0/1+ (Figure S6). In the N1s XPS spectrum (Figure S7), there existed 16.6% pyridinic N (398.2 eV) and 18.3% Cu-Nx (399.7 eV).[13–14] In addition, another peak at 401.1 eV was also observed (Figure S7), assigned to the graphitic N and/or the protonated pyridinic-NH+ species due to their similar bonding energies.[15] Interestingly, the obvious oxidative N (405.5 eV) with the 16.6% content was also found.[14b] To better identify distinguish these oxidative N species in Cu-N-C-800, the N K-edge X-ray absorption near edge structure (XANES) analysis was further carried out. Figure 1d showed that in addition to the pyridinic and graphitic N (or protonated pyridinic-NH+), another two peaks assigned to NO3− anion (405.1 eV) and -NO2 (403.5 eV) were obviously presented in Cu-N-C-800.[16] Importantly, the presence of NO3− strongly illustrated that during removing the possible Cu-based nanoparticles via nitric acid treatment, the partial pyridinic N of Cu-N-C-800 was successfully protonated due to that the pyridinic-N possesses lone-pair electrons. The corresponding contents of graphitic N and protonated pyridinic-NH+ species were determined through the alkali treatment, which were described in the following discussion.
To deeply elucidate the coordination environment of Cu, we performed extended X-ray absorption fine structure (EXAFS) analysis. The Cu K-edge XANES spectrum (Figure 1e) illustrated that the rising absorption edge intensity of Cu-N-C-800 was situated between those of Cu2O and CuO, indicating the dominant valence state of single Cuδ+ (1 < δ < 2) atom supported on the protonated N-doped carbon. This data along with the XPS result further revealed the dominant co-existence of Cu2+ and Cu+ (Figure S6). Through k3-weighted Fourier-transform, a dominant Cu-N first coordination shell at 1.53 Å appeared in the R space EXAFS spectrum of Cu-N-C-800 without obvious Cu-Cu characteristic peaks (Figure 1f). In the wavelet transform (WT) contour plots, the maximum intensity of Cu-N-C-800 at 3.9 Å−1 arising from Cu-N bonds was also presented (Figure 1h). This agreed with the XPS, HAADF-STEM and XRD results (Figure S4, S5 and S7), again implying that the copper was mainly dispersed on the protonated N-doped carbon matrix in the Cu-N single-atom form. The further fitting result disclosed that the coordination number of Cu atom approached four (Table S1), indicating that the Cu atoms were mainly dispersed on the protonated N-doped carbon matrix in the Cu-N4 configuration, similar to that of copper phthalocyanine (CuPc) (Figure 1g) and many reported studies.[17] All of the above structural characterizations jointly demonstrated that the single-atom Cu-N center, positive-charged graphitic N and protonated pyridinic-NH+ species had been successfully incorporated into the N-doped carbon matrix.
Taking DBT as a model molecule, we evaluated the adsorption capability of Cu-N-C-800. For comparison, the adsorption performance of commercial carbon black (CB) with a high surface area, copper (II) meso-tetraphenylporphine (CuTPP) with single-atom Cu-N site and the hybrid of CB and CuTPP (CuTPP@CB) were also investigated. Figure 2a revealed that the desulfurization rate of Cu-N-C-800 attained 91.3%, notably superior to those of CB, CuTPP and CuTPP@CB. During adsorbing DBT, Cu-N-C-800 followed the Langmuir adsorption model, indicating its monolayer adsorption mechanism (Figure S8).[18] According to this model, we further evaluated the maximum adsorption capacity (Qm). It can be seen that the Qm of Cu-N-C-800 (35.32 mg S/g) was much larger than that of CB, CuTPP and CuTPP@CB, and even surpassed or approached that of many reported carbon nanomaterials (Figure 2b and Table S2).[3a, 19] The corresponding breakthrough result obtained by a fixed bed experiment also disclosed the desulfurization capacity of Cu-N-C-800, which could reach 32.02 mg S/g (Figure S9), approaching to the theoretical maximum adsorption capacity of Langmuir isotherm model. Furthermore, Cu-N-C-800 still displayed the excellent stability after the continuous six recycling (Figure 2c), due to its robust structural feature.
The adsorption kinetics represented another crucial parameter to evaluate the adsorption capability of material. Figure S10 and Table S3 revealed that Cu-N-C-800 obeyed the pseudo-second-order model with a rate constant (K2) of 0.1054 g/mg/min during removal of DBT, which surpassed that of the counterparts of CB and CuTPP (Figure S10). Such a high K2 presented by Cu-N-C-800 almost exceeded most of the reported adsorbents (Figure 2d) by a factor of 2.4-310 times.[3a, 18a, 18b, 18g−i, 18n, 19a−d, 20] These facts directly suggested the ultrafast adsorption kinetics of Cu-N-C-800 during desulfurization, which was also verified by the fixed bed experiment. It showed that Cu-N-C-800 rapidly reduced the DBT content to a lower than 10 ppm level of the U.S. Environmental Protection Agency within 70s, far outstripping the commercial CB and the recently reported Fe-N-C-750 adsorbents (Figure 2e).[3a]
The real fuels probably involved different organosulfur compounds. Thus, we also examined the adsorption capabilities of Cu-N-C-800 for other organosulfur such as thiophene (Th), benzothiophene (BT) and 4, 6-dimethyldibenzothiophene (4, 6-DMDBT). According to Figure 2f, it can be seen that the desulfurization rate followed the order of Th < BT < 4, 6-DMDBT~DBT. Due to the possibly competitive adsorption, we further paid our attention to explore how the interfering organosulfur influenced the capture capability of Cu-N-C-800. By choosing Th and 4, 6-DMDBT as interfering adsorbate, we investigated the adsorption selectivity of Cu-N-C-800 for DBT. Figure S11 displayed that Th showed a negligible influence on the selective adsorption of DBT. In presence of 20% 4, 6-DMDBT, Cu-N-C-800 could maintain the high desulfurization selectivity for DBT. In addition, we also tested the desulfurization performance of Cu-N-C-800 in the model oil which contained toluene. Figure S12 unveiled that the desulfurization rate declined as the content of toluene increased, indicating desulfurization rate was subject to the competition adsorption and strong solvent effect of toluene that was similar to our previous work.[3a]
To further examine the adsorption mechanism of Cu-N-C-800, Fourier-transform infrared (FT-IR) and XPS spectrum measurement were carried out. As shown in Figure S13, the strong S2p peak existed in the sample of Cu-N-C-800 (adsorbed). Figure 2g showed that the aromatic ring peak of DBT presented an obvious red shift after adsorbed by Cu-N-C-800, suggesting the strong interaction between the adsorption sites and DBT. When the Cu-N center of Cu-N-C-800 was poisoned by SCN− through a coordination effect,[3a, 9c] its desulfurization rate obviously decreased (Figure S14). From the Cu2p XPS spectra, it can be further observed that the Cu2+ peak of Cu-N-C-800 (adsorbed) shifted to the higher bonding energy in comparison with that of Cu-N-C-800 (fresh) while there was no shift for the Cu+ peak (Figure 2h). Combined with the density of state (DOS) calculation, these implied that the Cu2+ acted as adsorption sites to remove DBT via a Cu-S coordination bond between Cu3d and S3p orbitals (Figure S15). The obvious peak shift was also found for the positive-charged graphitic N and pyridinic-NH+ once the adsorptive desulfurization occurred in Cu-N-C-800, signifying that the pyridinic-NH+ and graphitic N species were also the desulfurization sites for removing DBT via the cation-π interaction (Figure 2i).
Due to the existence of graphitic structure, we further evaluated the possible contribution of the π-π interaction between Cu-N-C-800 and DBT for adsorptive desulfurization. According to the elemental analysis, N1s and ICP-MS results, it could be calculated that 1g Cu-N-C-800 theoretically contained 2.12 mmol adsorption sites, which would remove 11.49 mmol DBT molecules, corresponding to 64.12 mg sulfur. This obviously outstripped the Qm (35.32 mg/g) of Cu-N-C-800, implying that the adsorption driven by the π-π interaction between Cu-N-C-800 and DBT could be negligible. Such a result was consistent well with the theoretical calculation that the π-π interaction presented the poorest capability compared with other sites, because of its lowest adsorption energy (Figure S16). On the basis of the above results, it can be concluded that the Cu2+-N and positive-charged graphitic N/pyridinic-NH+ were the adsorption sites of Cu-N-C-800 during desulfurization.
To further determine the critical role of protonation in the adsorptive desulfurization, we used 1M NaOH to eliminate the proton of pyridinic-NH+. As displayed in the N1s XPS spectra, the relative content of oxidative N species drastically declined in the alkali-treated Cu-N-C-800 (denoted as Cu-N-C-800*), while the pyridinic-N percentage significantly increased (Figure 3a). The N K-edge XANES analysis further disclosed that Cu-N-C-800* did not contained any NO3− (Figure 3b), hinting the complete removal of proton. On the basis of N K-edge XANES and N1s XPS results, we again verified the coexistence of graphitic N and protonated pyridinic-NH+ species in Cu-N-C-800. Moreover, the subsequent adsorption experiment indicated that Cu-N-C-800 resulted in 28.07% and 46.00% decrease of desulfurization rate and adsorption kinetics (i.e., the rate constant (K2)), respectively (Figure 3c and 3d). In order to explore whether the decreased desulfurization performance only originated from the decrease of adsorption sites, i.e., the disappearance of protonated pyridinic-NH+ species, which could behave as adsorption sites to remove DBT via the cation-π interaction, we reanalyzed the EXAFS data. The corresponding R space spectrum of EXAFS revealed that the main peak position of Cu-N-C-800* showed a positive shift compared with Cu-N-C-800 (Figure 4a), indicating that the protonation could alter the local ligand field around the single-atom Cu-N sites to increase the Cu-N bond length. Combining the above results, it implied that in addition to eliminating the pyridinic-NH+ active sites, the deprotonation modulated the electronic structure of single-atom Cu-N sites to probably weaken the bonding strength for DBT, contributing to the decreased desulfurization performance. In other words, the protonated pyridinic-NH+ species played the roles of both adsorption sites and regulator for modifying the electronic structure of single-atom Cu-N sites.
By further comparing the N K-edge XANES spectra and its 1st derivative of Cu-N-C-800 and Cu-N-C-800*, we found that the pyridinic and graphitic N in Cu-N-800* positively shifted to the higher binding energies (Figure 3b and Figure S17). Meanwhile, the Cu K-edge XANES spectra displayed that the rising absorption edge energy of Cu-N-C-800* shifted toward lower energy compared with Cu-N-800 (Figure 4b). These data directly suggested that the protonation reconstructed the ligand-field around single-atom Cu sites, lessening the electron transfer from the N-doped carbon matrix to the isolated Cu atoms via Cu-N coordination bonds in Cu-N-C-800, which not only weakened the interaction between the single-atom Cu sites and N-doped carbon support reflected by the slightly increased Cu-N bond length (Figure 4a and Table S1), but also resulted in the more electron-deficient single-atom Cu centers than that in Cu-N-C-800*.
After adsorbing DBT, the corresponding Cu 2p XPS spectra showed that the binding energy of Cu in Cu-N-C-800 displayed a more significant shift compared with Cu-N-C-800* (Figure 4c and Figure S18), indicating a stronger interaction between the Cu-N sites and DBT in Cu-N-C-800. This demonstrated that the more electron-deficient feature induced by protonation was beneficial to enhance the bonding strength of single-atom Cu centers for DBT. In addition, we also chose ultraviolet photoelectron spectroscopy (UPS) to evaluate the d-band center position of Cu-N-C-800,[21] which was also used to decipher the bonding strength between active site and adsorbate. Figure 4d revealed that the protonation obviously upshifted the d-band center position of Cu-N-C-800 toward the Fermi level. By further linking the d-band center with the corresponding adsorption capacity and desorption rate, we uncovered that the upshifted d-band center made Cu-N-C-800 exhibit a higher adsorption rate and a lower desorption rate compared with Cu-N-C-800* (Figure S19), further implying the crucial role in strengthening the bonding strength between Cu-N-C-800 and DBT. These results jointly validated the protonated pyridinic-NH+ species not only behave as the desulfurization sites, but also modulated the electronic structure of Cu-N-C-800 via reconstruction of ligand field to thermodynamically strengthen the bonding strength for DBT, which was beneficial to rapidly anchor and stabilize DBT once they contacted with each other.
In addition to manipulating the thermodynamic of adsorption via reconstruction of ligand field, our result disclosed that the protonation also improved the adsorption kinetics, which was reflected by the rate constant (K2). Considering that the protonation was able to modify the surface charge of Cu-N-C-800, we thus performed the theoretical calculation to analyze the change of electric field around adsorption sites, because the charged surface could induce a local electric field [22] to modulate adsorption kinetics by altering the mass transfer. As expected, the protonated pyridinic-NH+ obviously altered the electric field around adsorption sites in Cu-N-C-800 compared with that in the deprotonated Cu-N-C-800* (Figure 4e). To provide the direct experimental evidence, the surface electric fields of Cu-N-C-800 and Cu-N-C-800* were further determined using Zeta potential meter. Figure 4f indicated that the negative surface potential of Cu-N-C-800 notably increased from -28.17 to -18.46 after the protonation, leading to a weaker negative local electric field. Due to the electron-rich feature of DBT, this weaker negative local electric field induced by the protonation would contribute to the faster transport and accumulation of DBT by significantly weakening the repulsion effect between DBT and Cu-N-C-800. It would help to maintain the notable concentration gradient of DBT between the surface and porous channel of Cu-N-C-800, accelerating the subsequent mass transfer of DBT from pores to the adsorption sites. As a result, Cu-N-C-800 presented the faster the adsorption kinetics during desulfurization compared with the deprotonated counterpart. Furthermore, once the electron-rich DBT rapidly approached the adsorption sites of Cu-N-C-800 via this field-induced acceleration effect, the more electron-deficient single-atom Cu-N along with the positively charged pyridinic-NH+ and graphitic N sites would quickly capture and stabilize DBT, ensuring the significantly enhanced desulfurization rate of Cu-N-C-800.
The excellent adsorption capabilities of single-atom Cu-N, charged pyridinic-NH+ and graphitic N sites in Cu-N-C-800 were also demonstrated by the theoretical calculation. Figure 5 showed the calculated adsorption energies of typical DBT molecule over various active sites. It was observed that the protonation of pyridinic N could lead to an obvious increase of adsorption energy of single-atom CuN4 site. This agreed with the experimental results, implying that the protonated pyridinic-NH+ was able to modulate the electronic structure of Cu-N site via reconstruction of ligand field to thermodynamically raise the adsorption energy of isolated Cu-N site for DBT, leading to an enhanced adsorption. Moreover, such an adsorption energy of isolated Cu-N site would be further improved with the increase of the protonated pyridinic-NH+ species, which could approach that of the positive-charged graphitic N and the protonated pyridinic N and obviously exceeded that of the pyridinic-N and π-π interaction. These supported the above experiment result that the protonated pyridinic-NH+, positive-charged graphitic N and single-atom Cu-N species were the mainly desulfurization sites.