The feedback mode of SECM is frequently used for the interface evolution study in LIBs (Liu et al., 2019a). In this mode, the microprobe monitors the regeneration rate of the redox mediator from the substrate to determine its electronic character. Ferrocene was chosen as the free-diffusion redox mediator and the corresponding cyclic voltammetry between 2.8 and 3.8 V is shown in Figure 1A. The redox potential of the ferrocene/ferrocenium (Fc/Fc⁺) combination is 3.26 V vs. Li⁺/Li, which is suitable for solid electrolyte interface characterization in LIBs (Zampardi et al., 2013; Ventosa et al., 2017). During the feedback measurement (Figure 1B), the microprobe has an applied potential of 3.6 V vs. Li⁺/Li as it moves toward the substrate, and the current at the microprobe increases when the substrate is conductive due to the increase of Fc regeneration rate (positive feedback). The microprobe current decreases when the substrate is passivated because the regeneration and diffusion of Fc are all hindered by the insulating substrate (negative feedback). In Figure 1C, a positive feedback is observed when approaching the pristine electrode surface, indicating the conductive nature of the carbon additive electrode. SEM pictures of Super P powder and electrode are shown in Figure S1, the Super P powder has sub-hundred nanometer spherical shaped particles, and the morphology doesn't change much after being made into an electrode. In addition, the Super P has high surface area of 63.41 m² g⁻¹ from the BET measurement result.

To investigate the effect of conductive carbon on the EEI formation, the approach curve and area scan of SECM were employed to examine the change of substrate conductivity. Figure 2A shows the first five cycles of Super P electrode ranged between 3.0 and 0.0 V vs. Li⁺/Li. Two characteristic reversible capacity regions were observed, one in 1.4–0.6V and the other below 0.6V. This is ascribed to the specific lithium storage mechanism of Super P, i.e., in the vicinity of turbostratic graphene edges (Anothumakkool et al., 2018). In addition, the irreversible capacity is high especially for the first few cycles, which is caused by the electrolyte reduction and SEI formation (Fransson et al., 2001; Gnanamuthu and Lee, 2011). In Figure 2B, the approach curves were recorded over the electrodes at a pristine state and after cycles. The approach curves change from positive feedback to negative feedback, and the normalized current densities decrease with cycles, indicating the kinetics for the Fc⁺ reduction get slower on the electrode with SEI formation. Figure 2C shows the corresponding area scan of an identical 120 × 120 μm region with cycles. For a better comparison, the different sets of area scan were all plotted with the normalized current densities varying <0.03. At pristine state, the normalized current densities are larger than 1.02, indicating the good conductivity of the electrode substrate. In addition, it's commonly assumed that the pristine electrode is uniformly distributed and the variation in the feedback current density is solely caused by topographical difference (Heinz et al., 2014; Liu et al., 2019b,c,d). Thus, the area scan of the pristine electrode in Figure 2C was converted to topographical information via calculation with the approach curve (Figure S2). In Figure 2D, the scanned area has a topographical variation of ~8.5 μm, corresponding to the increase of feedback current with substrate height (Figure 2C).

With the cycling process, SEI evolves and thus the feedback images are not only reflecting the topography but also the lateral heterogeneity of SEI (Heinz et al., 2014). Since the thickness of SEI are in the range of 20~50 nm, the thickness can't induce feedback current change larger than 1 pA, which is negligible compared to the ~nA scale feedback image. So the feedback image in Figure 2C (6 and 10th cycles) are a result of the pristine topographical difference and the heterogeneous passivating properties of SEI for the scanned area. With the pristine topography (Figure 2D), the evolution of passivating SEI can be determined via calculation in combination with their corresponding approach curves (detailed calculation in Supplementary Materials). In Figure 2E, the heterogenous passivating SEI can be seen more clearly with the pristine topography subtracted. For the 6th cycle in Figure 2E, the feedback currents change from 70 to −120 pA as the image changes from red to purple, indicating the decrease of the mediator regeneration rate with SEI passivation. In accordance with the topography (Figure 2D), the passivation layer tends to form in the y <70 μm region. But the SEI passivation sites do not necessarily correlate with the substrate height. This can be seen clearly from the identical height region (Figure 2D: 20 μm < y <110 μm), which has the mixture of positive and negative feedbacks (Figure 2E). Upon the 10th cycle, the feedback current drops to the range of −100 ~ −300 pA, indicating a much slower mediator regeneration rate and more severe SEI passivation effect. Compared with the 6th cycle, the passivation region expands to a larger area with an elliptical shape covering the majority of the scanned area. It demonstrate that the passivating SEI grows with cycles and the growth has lateral heterogeneous properties.

The conductive carbon electrode is cycled between 3.0 and 4.7 V vs. Li⁺/Li to examine the interface evolution in the high-potential region. In Figure 3A of the cyclic voltammetry curves, a pair of reversible redox peaks were observed at 3.28/3.18 V, corresponding to the oxidation and reduction reactions of ferrocene. In addition, the redox peaks at 4.56/4.10 V have obvious larger anodic peak than the cathodic peak, demonstrating the existence of an irreversible oxidation process. The anodic peak can be induced by anion intercalation, oxidative decomposition of electrolyte and conductive carbon oxidation (itself or its functional groups) (Qi et al., 2014, 2015; Soon et al., 2018). With these three contributions, the reversible part could be the (de)intercalation of PF6- anions to/from the electrodes (Qi et al., 2014). The oxidation of electrolyte on conductive carbon at potential larger than 4.6 V is irreversible once the cathode electrolyte interface has formed (La Mantia et al., 2013; Li et al., 2013, 2014). The interface layer prefers to form on the carbon additive of the electrodes (Nordh et al., 2015). In addition, carbon black degrades with cycling via agglomeration and loss of crystallinity to amorphous structure. This leads to the decrease of the electrical conductivity, increase in the heterogeneity of the conductive carbon network, and reduction in the electron percolation of the composite electrode (Ngo et al., 2016; Scipioni et al., 2016).

The change of electronic character of the electrode substrate with cycles was examined by the approach curve and area scan of SECM (Figures 3B,C). In accordance with the result of low-potential sweep (Figure 2B), the approach curves change from positive feedback to negative feedback with cycles, indicating the decrease of electrode substrate conductivity induced by electrolyte decomposition and conductive carbon degradation. At the same sweep rate of CV, the passivation rate of the interface layer in the high-potential region (3.0–4.7V) is slower than that in the low-potential region (3.0–0.0V), which is caused by the different interface components. SECM area scan was applied to visualize the areal heterogeneity of interface evolution (Figure 3C), here we also converted the pristine feedback image to topographical information (Figure 3D), and the area scan after cycles having the topography feedback subtracted, i.e., passivation property feedback (Figure 3E). In Figure 3D, the scanned area has a topographical variation of ~12 μm with higher upper region (60 < y <100). After the 10th cycle, the passivation layer covers the majority of the electrode surface with negative feedback current densities, which is induced by electrolyte decomposition and conductive carbon degradation. Upon the 20th cycle, the magnitude of areal feedback current is not decreased, while the areal heterogeneity of the passivation layer is becoming more obvious. The nearly unchanged areal feedback current magnitude from the 10 to 20th cycles is unexpected. This could because the surface of conductive carbon has been covered by the electrolyte degradation products in the first ten cycles; for the second ten cycles, when a large amount of the surface area and active functional groups of conductive carbon has been covered, the oxidative decomposition of electrolytes is less severe. While the structural degradation, aggregation, and agglomeration of conductive carbon last throughout the whole cycling process, which may leads to the evolution of passivation layer and increase of areal heterogeneity.

TEM was applied to investigate the morphology and microstructure of the pristine and cycled conductive carbon electrodes at nanoscale. At pristine state in Figures 4a,b, the Super P particles have spherical shape with diameter of 30–50 nm. The particles have quasi-crystal structure and the graphitic structures are oriented concentrically tangent toward the surface of the primary particles (Gnanamuthu and Lee, 2011). After lithiation, the internal structure of the Super P particles were preserved while ~5 nm thick SEI layer is formed on the particle surface (Figures 4c,d) (Anothumakkool et al., 2018). In Figures 4e,f, the particles of the charged sample swelled and lost crystallinity. The structural degradation can be partially explained by the anion intercalation and co-intercalation of solvent molecules to expand the interlayer and electrolyte decomposition inside the carbon (Qi et al., 2014).

The electrodes after cycles in the low-potential and high-potential regions were analyzed by XPS to provide information about the chemical composition on the surface. The elemental concentration of major components are listed in Table 1 with the corresponding C1s, O1s, and F1s spectra shown in Figure 5. The chemical composition of the solid electrolyte interfaces cycled in the low-potential and high-potential windows are similar except for the difference in the composition amount. This might start from the spontaneous reaction between the carbonate electrode and electrolyte after immersion in the electrolyte. Membreño et al. revealed the spontaneous decomposition of electrolyte on the carbon additive in a corrosion-like reaction with spontaneous polymerization reactions on the carbon surface (Membreño et al., 2015). The functional groups include alkanes (C-C), alkoxides and ethers (RCO), lithium carboxylate and esters (ROCO), carbonates (RCO₃), fluoroalkanes (RCF_n), lithium fluoride (LiF) and degraded lithium salt products (Li_xPF_y and Li_xPO_yF_z) (Qi et al., 2015; Younesi et al., 2015). The spontaneously formed SEI evolves during the electrochemical cycling in the low and high potential regions. After cycles in the 0–3 V vs. Li⁺/Li, the electrode is covered by a large fraction of inorganic compounds, LiF (mostly), Li₂CO₃, and Li_xPF_yO_z, and fewer organic species (RCO, ROCO, ROCO₂, and RCO₃) formed via the degradation of salt and solvents. In comparison, the electrode cycled in the high-potential region has fewer inorganic species, which can be told from the lower amount of Li1s, F1s, and much higher C1s element concentration (Table 1). In addition, the much higher ratio between C-C bonds and carbon oxygen bonds demonstrates that the electrode surface has less oxygen-containing species and less electrolyte decomposition (Soon et al., 2018; Tatara et al., 2019).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.