Electrochemical performances of HC in different electrolytes. DGM is selected as a model to illustrate the unique feature of Na insertion/extraction in HC. Meanwhile, the conventional solvent, EC/DEC, is also tested under the same conditions as the benchmark. Commercial hard carbon is directly used without further treatment. Scanning electron microscopy (SEM) and transmission electron microscope (TEM) images show that they are irregular microparticles with randomly-oriented and intermittent lattice fringes (Supplementary Fig. S1)34,35. A variety of structure defects in the particle offer numerous active sites for Na storage. Figure 1 presents the electrochemical performances of HC in DGM and in EC/DEC at 0.1 A g-1. The differences between them are very obvious, particularly in terms of Coulombic efficiency and rate performances. As shown in Figs. 1a and 1b, hard carbon in DGM exhibits a higher ICE (~80.0%) than that in EC/DEC (~68.2%), indicating the reduced side reactions. To gain more information about these side reactions, dQ/dV is plotted against the cell voltage for two cases. HC in EC/DEC gives three cathodic peaks at 0.9 V, 0.5 V and 0.1 V in the first cycle (Fig. 1c), which could be assigned to the adsorption of Na+ ions on the edge and/or defects of carbon layers, the electrolyte decomposition and the reduction of Na+ ions within micro-voids36-38. In the following cycles, the two cathodic peaks at 0.9 V and 0.5 V almost disappear, but that at 0.1 V only slightly decreases. This result indicates that the two reactions happened at high voltages are the essential origin for low ICE. Different from the case of EC/DEC, HC in DGM only has two cathodic peaks at 1.0 V and 0.1 V in the first cycle (Fig. 1d). The cathodic peak at 0.5 V is invisible in the profile even in the first cycle, suggesting the absence of electrolyte decomposition. This result is quite interesting and not reported before to the best of our knowledge. More important, the coulombic efficiency of HC in DGM is more stable in the subsequent cycles (Fig. 1e), suggesting the high reaction reversibility. On the other hand, HC in DGM also shows the smaller electrode overpotential (~0.08 V) than that in EC/DEC (~0.15 V) for the redox reaction at 0.1 V. It indicates the fast electrochemical reaction kinetics. This conclusion is also supported by rate performances (Fig. 1f). The capacity difference between DGM and EC/DEC increases as the current density arises. At 2 A g-1, the specific capacity of HC remains ~266 mAh g-1 in DGM, but only ~15 mAh g-1 in EC/ DEC. Even in terms of capacity retention, it still comes to the same conclusion, 78% in DGM vs. 5.9% in EC/DEC. The improvements in reaction kinetics and reaction reversibility do not come at the expense of cycling stability. HC displays a good cycling stability and a large specific capacity in DGM (Fig. 1g). It delivers a capacity of 230 mAh g-1 after 1500 cycles at 1 A g-1, much higher than ~40 mAh g-1 in EC/DEC. Even after 3500 cycles, the capacity is still 224.4 mAh g-1, corresponding to a capacity retention of 88% (Fig. 1h and Supplementary Fig. S2). It is almost the best performance for hard carbon (Tab. S1).
Ex-situ HRTEM images characterization of SEI films. All these data confirm that the electrochemical performances of HC are remarkably improved in DGM, as compared to the case in EC/DEC. Usually, these improvements are attributed to the SEI film in different electrolytes39. However, the cathodic peak related to the electrolyte decomposition is absent in DGM, even in the first cycle (Fig. 1c). Therefore, it is likely without a SEI film on HC in DGM. Then, the previous explanation based on the SEI film does not work for HC. So, it is important to clarify if there is a SEI film on HC cycled in DGM. HRTEM is an intuitive tool to visualize the SEI film. As shown in Fig. 2a, the surface of hard carbon discharged to 0.01 V is clear without an amorphous layer, thus excluding the formation of a SEI film. The result is in good agreement with that observed in dQ/dV. The intermittent lattice near the surface is a characteristic of HC (Fig. 2b). The fringe spacing about 3.94 Å is larger than that before discharge (~3.6 Å, Supplementary Fig. S1), probably caused by the intercalation of Na ions. As hard carbon is charged to 1.5 V, the similar surface is also observed (Fig. 2c), confirming the non-SEI film again. Meanwhile, the fringe spacing is reduced to ~3.74 Å (Fig. 2c), likely due to Na extraction. The similar surface without a SEI film is also identified after 5 cycles in DGM (Supplementary Figs. S3a and S4a), suggesting the good surface stability. In contrast, the surface of HC cycled in EC/DEC is coated by a thin SEI film, as highlighted in Figs. 2d-2f. The lattice fringes of HC do not show up any more. The contrast differences indicate that the spherical inorganics are randomly dispersed in organic species, resulting in this unique composite. This result is well consistent with the well-accepted Mosaic model about SEI film40,41. Unfortunately, the structure information of spherical inorganics is missed, due to the high-energy electron irradiation. In addition, the surface film is not uniform, which affects the local Na-ion flux, increases the stress accumulation and degrades the electrochemical performances. Therefore, after 5 cycles, the surface film on hard carbon becomes much thicker (Supplementary Figs. S3b and S4a), but the internal structure remains as the same.
Ex-situ 1H NMR and GC-MS studies on HC. Although HRTEM gives strong evidences to support the absence of a SEI film on hard carbon in DGM, the detection area by HRTEM is always limited. Moreover, the sample preparation for TEM images may interfere with the identification of the SEI film. So, to avoid the possible misleading by the limitation of TEM, using other techniques to characterize the electrode surface is very necessary. In this context, NMR and GC-MS, which are very good at the detection and characterization of organics, are used42. Once there is electrolyte decomposition, the formation of organics is inevitable in the SEI film. The electrodes discharged to different voltages were soaked in deuterated dimethyl sulfoxide (DMSO-d6) overnight. Due to the strong polarity of DMSO, organics were easily dissolved and the solution was subject to NMR. As illustrated in Fig. 3a, the 1H NMR spectra are kept exactly the same as the standard spectrum of DGM (Supplementary Fig. S5a) throughout the whole discharge process. This result indicates DGM just adsorbs on the surface of hard carbon. This same result is also obtained in 13C NMR, where both the chemical shift and the relative ratio of the signals remain as the same at different voltages (Supplementary Fig. S6). This result excludes the polymerization reaction potentially happened to DGM. In contrast, the 1H spectra of hard carbon
in EC/DEC become totally different (Fig. 3b). The 1H NMR spectrum at OCV is similar to the standard spectrum of EC/DEC (Supplementary Fig. S5b). As the voltage decreases to 0.75 V, the weak chemical shift of sodium ethylene dicarbonate (SEDC) at 3.40 ppm appears in the spectrum43. The signal intensity greatly increases within 0.1-0.01 V, implying the formation of a large amount of SEDC in this voltage range. Beside SEDC, the chemical shift from CH2 of ethylene also shows up at 0.1 V and intensifies at 0.01 V. The formation of SEDC and ethylene can be attributed to two-molecule polymerization of ethylene carbonate (Supplementary Fig. S7, Equ. 1), as supported by their close signal areas. It is noted that the electrolysis products of diethyl carbonate (DEC), i.e. sodium methyl carbonate (SMC) and butane (Supplementary Fig. S7, Equ. 2), also appear at 0.01 V, despite of the low contents. The appearance of these organic species confirms severe electrolyte decomposition in EC/DEC. The similar changes are also identified in the 13C spectra of hard carbon in EC/DEC (Supplementary Fig. S8).
The conclusion on the basis of GC-MS is similar to that from NMR. Different from the electrode processing for NMR, the electrode for GC-MS was treated by CDCl3, instead of DMSO. The difference of CHCl3 and DMSO in polarity allows them to dissolve different organics, thereby providing another perspective to validate the conclusion. The electrode cycled in DGM exhibits the same fragmentation pattern as that of DGM (Supplementary Fig. S9), where the ions at m/z 29,45,59 and 89 can be assigned to the fragments of DGM caused by electron ionization (EI) (Fig. 3c), i.e. [CH3CH2]+, [HOCH2CH2]+, [CH3OCH2CH2]+, and [HOCH2CH2OCH2CH2]+. This result indicates that there are some DGM molecules adsorbed on the electrode surface, which cannot be removed by simple washing. As expected, the fragment pattern remains almost identical at the different voltages, indicating the good stability of the electrolytes and the absence of SEI film on the surface. The same result is even observed in the electrode discharged to 0.01 V after three cycles (Supplementary Fig. S10). On the contrary, the electrode cycled in EC/DEC exhibit the different fragmentation patterns, as the discharging undergoes. A batch of new fragment ions at m/z > 90 appears in the fully-discharged electrode (Fig. 3d), indicating the formation of new organics (e. g. SEDC at m/z 148) in the electrode. This result can be attributed to the electrolyte decomposition at a low voltage, which is also in good agreement with what observed in NMR.
In situ EQCM of HC in DGM. In-situ EQCM allows us to accurately monitor the tiny weight change of the electrodes as a function of the applied voltages. Therefore, it can provide useful information about the formation of the SEI film and the intercalation of Na+ into the electrodes, because both of them involve the mass gain or loss in the electrodes44. In a typical experiment, the electrodes are discharged either in DGM or in EC/DEC at a constant current density in a specially designed cell, using Na as the reference electrode. As shown in Fig. 4a, the voltage quickly falls to -2.65 V (vs. Ag+/Ag) and then gradually approaches -2.8 V (vs. Ag+/Ag) during discharging, which indicates the adsorption and insertion of Na+ into hard carbon. Fig. 4b shows the plot of mass vs. charge, where two stages can be clearly seen. During the first stage, the mass change per electron, ∆m/dq, is 24.6 g mol-1, which is close to the ionic mass of Na+ (23 g mol-1) indicating that the adsorption of Na+ ions on hard carbon, probably at the edges or structure defects of carbon layers. During the second stage, ∆m/dq increases to 160.3 g mol-1 until the end of the discharging. This result is consistent with the mass/charge ratio of [Na(DGM)]+ (157 g mol-1), implying that the solvent molecules are intercalated into hard carbon together with Na+ ions. Interestingly, the carbon layers are not exfoliated upon the intercalation of the organic molecules, because the graphene domains in hard carbon are randomly-oriented and intermittent. Thus, the graphene layer won’t be easily peeled off and detached from the others. As a result, hard carbon exhibits a stable cycling over 3500 cycles, as demonstrated in Fig. 1h. Moreover, the mass/charge ratios related to the SEI components, such as polyDGMs, sodium alkyl alcohol (RCONa), etc. are not observed. It excludes the formation of a SEI film on hard carbon again, consistent with the results from NMR, MS and HRTEM. To our knowledge, these results, especially the identification of solvent molecules in hard carbon, are reported for the first time. The height change with charge as shown in Fig. 4c is consistent with the weight change in Fig. 4b, indicating that the adsorption and the intercalation of Na-related species make the electrode expanded accordingly.
Molecular dynamics simulations. To further understand the co-intercalation of DGM with Na+ into hard carbons, the solvated structures of NaPF6 in DGM are simulated by molecular dynamics (MD) based on first-principle theory45-47. As illustrated in Fig. 4d, a supercell contains 16 DGM molecules, one Na+ ion and one PF6- ion, keeping in line with a low concentration of NaPF6 in DGM. The cell was then heated to 1000 K and 2000 K in turn. At each temperature, MD simulations were conducted for 1 ps with a time step of 0.25 fs to identify the thermodynamically stable configuration. Then, the cell was cooled to room temperature by velocity scaling. After that, the cell was treated by NVT simulation at 300 K for 10 ps. (full name of NVT) Then, it is noted that each Na ions are surrounded by six O atoms. Three of them come from the same DGM molecule, while the other three originate from three different DGM molecules. In this context, the three molecules are easy to dissociate from Na ions, because they are only a monodentate ligand. The DGM molecule binding to Na ions with three O atoms is co-intercalated together in to hard carbon. This configuration is quite stable in DGM, as supported by the trajectories of all O atoms in the cells (Fig. 4f). This result provides the solid basis for the co-intercalation of DGM with Na ions into hard carbon. The same conclusion could be also obtained in the cases of graphite and soft carbon. As shown in HRTEM images (Supplementary Fig. S11), all the particle surfaces after 5 cycles exhibit clear lattice fringes without a sign of a SEI film. These results strongly confirm the absence of a SEI film.