Investigation of the Electron Transfer at Si Electrodes: Impact and Removal of the Native SiO2 Layer

Silicon is considered as one of the promising alternatives to graphite as negative electrode material in lithium-ion batteries. The electron transfer at uncharged microstructured and planar Si was characterized using the feedback mode of scanning electrochemical microscopy (SECM) and 2,5-di-tert-butyl-1,4-dimethoxybenzene as redox mediator. Approach curves and images demonstrate that the electron transfer rate constants at pristine Si are relatively small due to the native SiO2 surface layer. In addition, the electron transfer rate constants show local variations because of the heterogeneous coverage of SiO2. The SiO2 layer is at least partially removed by mechanical contact and abrasion with the microelectrode probe. After SiO2 removal by the microelectrode or by a hydrofluoric acid dip, the electron transfer rate constants increase strongly and remain heterogeneous. Moreover, the surface of the Si electrodes is at least stable over hours after SiO2 removal. The consequences for investigating the formation of the solid electrolyte interphase (SEI) on Si are discussed. © The Author(s) 2015. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0731603jes] All rights reserved.

Silicon is considered as one of the promising alternatives to graphite as negative electrode material in lithium-ion batteries (LIBs). This is due to its wide abundance, low voltage and high theoretical gravimetric capacity of 3579 mAh g −1 . 1 During the lithiation, the electrochemical potential of the Si electrode exceeds the stability window of the electrolyte and, consequently, the electrolyte is reductively decomposed. 2 The decomposition products form a solid electrolyte interphase (SEI) covering the Si material lying beneath. 3 A major challenge of Si as negative battery electrodes is the large volume change of ca. 270% 4 upon lithiation. Recurring lithiation/delithiation causes the electrode material to crack leading, in the worst case, to irreversible loss of active material. During the volume expansion non-passivated Si surfaces are expected to be exposed which are immediately covered by a SEI layer because of ongoing electrolyte decomposition. 5 Thus, similar to metallic Li electrodes the SEI layer is continuously re-formed and remains instable during each lithiation process. 6 The properties of the SEI are important for the performance of Si negative electrodes. 5,7 In order to evaluate the Si electrode performance, reliable in situ characterization of the SEI is definitely a key issue for the progress in this field. In general, scanning probe techniques are frequently applied for various aspects of battery research. 8,9 Especially atomic force microscopy (AFM) has been used to characterize in situ and ex situ the SEI at amorphous Si (a-Si) thin films, [10][11][12][13][14] Si-based thin films, 11,15 a-Si nanopillars 16 and Si nanowires. 17,18 AFM provides information about the morphology evolution and physical properties of the SEI. The present study aims at characterizing the functional properties of Si electrodes in situ by the feedback mode of scanning electrochemical microscopy (SECM), which probes the local electron transfer rate constants at the surface of interest in the electrolyte environment.
In the last years significant progress was made by using SECM to study electron transport across SEI-covered negative electrodes. Zampardi et al. investigated the SEI at TiO 2 paste negative electrodes 19 and glassy carbon model electrodes 20 using ferrocene as redox mediator. Our studies at uncharged 21 as well as charged graphite composite 22 and metallic Li electrodes 23 relied on 2,5-di-tert-butyl-1,4-dimethoxybenzene (DBDMB) as a redox mediator (Figure 1a). DBDMB was introduced by Dahn et al. 24 as overcharge protection agent for LIBs and turned out to be an excellent choice as SECM mediator in organic solvents, too. 22 In this study the electron transport at pristine Si electrodes is characterized by SECM using DBDMB as redox mediator in organic electrolyte (Figure 1b). 25 The investigation of pristine Si electrodes by SECM is an indispensable step toward the understanding and characterization of the SEI at Si electrodes in general. Although only Li + electrolyte is used in this study, the lithiation is not investigated, since the potential of all Si electrodes was kept at open circuit (E S = 3.2 V) and the first lithiation requires E S ≤ 0.2 V. 26 In contrast to pristine graphite electrodes, pristine Si electrodes are covered by a native SiO 2 surface layer, which is at least 1-3 nm in thickness and possess passivating properties for many charge transfer reactions. [27][28][29] Undoped Si/SiO 2 has been used as insulating substrate for the characterization of single-walled carbon nanotubes 30,31 and gapped Au nanobands 32 by SECM. For these examples the insulating property of undoped Si/SiO 2 substrates is beneficial. However, as discussed below, the insulating SiO 2 is disadvantageous for the characterization of the electron transfer kinetics at Si electrodes before and after lithiation because the electron transport across SiO 2 -covered Si is rather slow and hardly distinguishable from insulating behavior by SECM. Since the SEI on Si will further reduce the electron transfer rate, the electron transfer rate at SEI and SiO 2 -covered Si cannot be distinguished clearly from each other. Therefore, the passivating properties of the native SiO 2 layer, local differences in their properties as well as the mechanisms for damage and removal must be understood before studying the additional passivation or local formation of the SEI on Si electrodes. Here, we discuss the impact of the native SiO 2 surface layer on the electron transfer rate constants at boron-doped Si.

Experimental
Silicon electrodes.-A 1 μm thick layer of Cu as current collector was sputtered onto a 200 μm thick highly boron-doped monocrystalline Si wafer (100 orientation, 8 m cm specific resistivity, Shin-Etsu, Chiyoda, Japan). The Cu-covered Si wafer was cut into pieces of 4 × 4 mm 2 , which were used as planar Si electrode for the SECM experiments.
Another Cu-covered Si wafer with same specifications was etched by deep reactive-ion etching to form columns with a 50 × 50 μm 2 ground plot, a column height of 50 μm and a column-column separation of 50 μm. The microstructured Cu-covered Si wafer was cut into pieces of 4 × 4 mm 2 and was used for the SECM experiments.
Mechanical damage to the Si wafer was introduced by abrading with the microelectrode (ME) of the SECM. The resulting surfaces as well as the effect of the HF dip were investigated by AFM and X-ray photoelectron (XP) spectroscopy (Supporting information SI-3 to SI-6).
Hydrofluoric acid dip.-The planar Si electrodes of 4 × 4 mm 2 were washed with pure ethanol in order to remove organic residues from the surface and were subsequently dried in the drying oven at 60 • C for half an hour. For the removal of the SiO 2 layer, 100 μL of 5 mass-% hydrofluoric (HF) acid solution was placed on the Si surface by a soft Pasteur pipette and was kept there for 90 s. Care was taken to avoid the contact of the HF solution with the Cu current collector. Caution: HF solution is highly toxic and should be handled with extreme care. Afterwards, the Si electrodes were rinsed three times with deionized water and the water was immediately removed under high-pressure Ar flow to minimize the reoxidation of the Si surface. This rinsing step removed the hexafluorosilicic acid formed during HF etching. This procedure was reported elsewhere 33,34 for the investigation of the SiO 2 -free Si surface. The Si electrodes were placed directly under vacuum for a few hours before they were transferred to an Ar-filled glove box (Uni-Lab, M. Braun GmbH, Garching, Gemany).
Scanning electron microscopy.-The physical shape of the Si electrodes was evaluated by scanning electron microscopy (SEM) us-ing a Zeiss Gemini DSM 982 (Leo Elektronenmikroskopie GmbH, Oberkochen, Germany) with a beam current of 0.1 nA and an accelerating voltage of 5 kV. The secondary electron images were recorded using an Everhart-Thornley detector (ETD) with dwell times of 100 ms, a resolution of 1280 × 1024 and a tilt of 41 • .
Scanning electrochemical microscopy.-The scanning electrochemical microscope (SECM) was operated under the SECMx control software, 35 a 3-axes micropositioning system (MS30 precision actuator and PS30 distance measurement system, CU30 controller, mechOnics AG, Munich, Germany) and a bipotentiostat (Compactstat, Ivium Technologies, Eindhoven, The Netherlands). The positioning system was placed under a custom-made plexiglas bell (Figure 1c). 36 The bell and the controller for the micropositioning system were placed inside an Ar-filled glove box (Uni-Lab, M. Braun GmbH, Garching, Gemany). The cables for the 4 electrodes and the USB cable for the CU30 controller were fed through ports on the back side of the glove box.
The SECM was operated with MEs of radius r T ≈ 13 μm (Figure 1b), microstructured or planar Si as sample and two separate Li foils as auxiliary and reference electrodes. MEs were prepared by sealing a Pt wire of 25 μm specified diameter (Goodfellow GmbH, Bad Nauheim, Germany) into borosilicate glass capillaries (Hilgenberg GmbH, Malsfeld, Germany). The MEs were grinded using a Micro Grinder EG-400 (Narishige, Tokyo, Japan) and polished using rotating wheels with micropolishing cloth with a suspension of 0.05 μm alumina particles to a mirror finish and a RG ratio of ≈ 5-10. RG is the ratio of the thickness of the insulating glass sheath around the Pt wire and the radius r T of the active electrode area.
The cylindrical opening of the SECM cell served as reservoir for 0. Scanning electrochemical microscopy measurements.-Before approach curves and images were recorded, the ME was polarized to 4.1 V vs. Li/Li + , where DBDMB is oxidized at the ME under steady state conditions (Figures 1a and 1b). The timer was set to zero, when the working solution was filled into the cell. The approach curves of Figures 3, 4 and 6 were recorded with a step size of 0.5 μm, a delay of 1.0 s between the translation and the current recording giving an average translation rate of 0.25 μm s −1 . Prior to fitting, the radius of the electroactive area r T and the RG were determined by confocal laser scanning microscopy (CLSM) using a TCS SP2 AOBS (Leica Microsystems GmbH, Wetzlar, Germany). For fitting approach curves by the analytical approximations by Cornut and Lefrou, 38 the uncertainties of 0.5 μm and 1.5 for r T and RG with respect to the CLSM results were considered. Approach curves were fitted by adjusting κ, i T,∞ and d 0 (within reasonable range) using a least square approach. κ is the heterogeneous normalized first-order rate constant, i T,∞ the diffusion-limited steady state current in the bulk solution (at quasiinfinite distance to the surface) and d 0 is the smallest ME-sample distance of the curve where a valid data point was recorded before the mechanical touch between ME body and sample. The apparent heterogeneous rate constant k eff was calculated by Equation 1 using a diffusion coefficient of D = 8.15 × 10 -7 cm 2 s −1 for DBDMB (see above for the definition of κ and r T ). 23 The diffusion coefficient was determined from the steady-state diffusion limited current of the ME i T,∞ .
The parameters of the SECM images are described elsewhere. 25 The distances between ME and sample electrode during imaging are given for the position (x/μm, y/μm) = (0, 0) of the image.

Results and Discussion
Imaging of microstructured Si electrodes.-Several cyclic voltammograms of the monocrystalline Si electrodes were recorded and the results confirm earlier results on polycrystalline Si presented by Baggetto et al. 39 Figure 2a depicts the SEM image of the regular structure of the microstructured Si electrode. The dimensions of each square-based pillar were 50 × 50 μm 2 , the column height was 50 μm and the pillar-pillar distance was equal to 50 μm. Figure 2b shows a SECM image of an uncharged microstructured Si electrode at E OCP = 3.21 V vs. Li/Li + at a ME-Si electrode separation of d ≈ 7 μm with the same periodic arrangement of pillars as in the SEM image in Figure 2a by a stronger negative feedback above the protruding pillars. The decrease of i T with decreasing d indicated a rather small electron transfer rate constant and was also observed for an undoped mesoporous SiO 2 /Si substrate with ferrocenemethanol in aqueous solution. 40 The reason for the rather small electron transfer rate constant was the native SiO 2 surface layer of the Si electrode ( Figure 1b) which causes high impedance also evident from electrochemical impedance spectroscopy (EIS). 41 The ME current i T was locally increased at the edge of the upper left pillar in Figure 2a at x = 20 μm/y = 220 μm (circle symbol), because i T amounted to 2.8 nA and was even 0.5 nA larger than i T,∞ ≈ 2.3 nA. Therefore, it can be concluded that the local electron transfer rate constant at this position was much larger than at all other positions of the Si pillars. The increased local electron transfer rate constant was caused by a gentle collision of the probe with the Si pillar during a preceding imaging experiment. Thus, a part of the insulating SiO 2 was removed by this collision leading to an increased electron transfer rate constant at the underlying boron-doped Si. It was shown before that the surface layer of the Si electrode can be damaged and removed by an AFM probe. This was either intended 17,42-44 or occurred accidentally due to the interaction of AFM probes with the sample using the contact mode. 16 Approach curves at planar Si electrodes.-In order to quantify the electron transfer rate constants, approach curves were recorded at planar Si electrodes. This allows for the application of the analytical approximations by Cornut and Lefrou 38 describing the steady state i T (d, κ) for infinitely large sample electrodes and finite sample kinetics characterized by κ. The Si electrodes provide advantages for the approach curve characterization because of the smoothness and rigid structure compared to composite electrodes that are rough, flexible and do swell when contacted by solvents. Thus, the absolute position of the Si electrode remained constant during the experiments, because the Si was kept at E OCP = 3.2 V and thus was not lithiated. These conditions supported the accurate quantification of κ. The accuracy is limited by the uncertainty in the determination of the distance d of the shortest approach ( d ± 1.5 μm). Consequently, the method is suitable to differentiate significant changes of κ rather then subtle changes in d. Approach curves are plotted as normalized current I = i T /i T,∞ vs. the normalized distance L = d/r T , where r T is the radius of the active ME area (Figure 3). The ME current i T depends strongly on the type of sample (κ) and d. When the ME approaches a sample (Figure 3, curve 1), where the mediator regeneration at the sample is diffusion-controlled (positive feedback), the fast reduction  2.2 × 10 −6 -3.6 × 10 −5 a Please note that the lower limit for k eff might be orders of magnitude smaller.
of DBDMB + to DBDMB at the sample yields an additional flux of DBDMB to the ME and thus increases i T with decreasing distance (Figure 3, inset i). Figure 3 shows that the highest quantified electron transfer rate constant at the native SiO 2 -covered Si electrodes was rather small (curve 2) compared to the situation of positive feedback without kinetic limitation (curve 1). The curves 2 and 3 were still distinguishable from the approach to an inert and insulating surface, where no reduction of DBDMB + occurs and the presence of the sample hinders the diffusion of DBDMB from the solution bulk to the ME (negative feedback, curve 4). However, curves with a significantly smaller electron transfer rate than curve 3 cannot clearly be separated from the situation of hindered diffusion in curve 4. The dark gray ribbon between curves 2 and 4 depicts the possible range of observed electron transfer rates at SiO 2 -covered Si electrodes. Thus, the electron transport across SiO 2 -covered Si electrodes is generally possible, i.e. the SiO 2 layer does not completely inhibit the electron transfer to DBDMB + . However, there might be locations where the electron transfer rate constant is orders of magnitude smaller and negligible. Table I shows the obtained k eff values of 2.0 × 10 −6 -2.6 × 10 −5 cm s −1 for SiO 2 -covered Si electrodes in 1 M LiClO 4 PC electrolyte. Although the lower limit of k eff was unknown, the determined range of k eff values indicated a local heterogeneity of electron transfer rate constants. The local heterogeneity of k eff was expected because the SiO 2 coverage is heterogeneous at planar Si electrodes [45][46][47] and Si nanoparticles. 48 The k eff range for SiO 2 -covered Si electrodes was similar to that of SEI-covered metallic Li in the same electrolyte (Table I). The data in Table I were obtained from approach curves at specific locations preselected from two-dimensional imaging in Figures 4 and 6.
After the SiO 2 layer had been damaged and partially removed by the ME probe, the k eff values increased significantly (Figure 3, curves 5 and 6). The k eff range of 1.2 × 10 −4 -6.3 × 10 −3 cm s −1 in Table I corresponds to the curves 6 and 5, which differ by a factor of 53. In contrast to the monotonic curves 6 and 5, curves with a maximum are shown in the SI in Figure S1f-h. Please note that the smallest k eff value after damage (curve 6) was still five times larger than the largest k eff value for the SiO 2 -covered Si electrode (curve 2). Considering the largest k eff value after damage (curve 5), the ratio amounted to 240. A similar ratio of 200 was estimated for the increase of k eff several seconds after damaging a SEI-covered metallic Li electrode by the ME probe. 23 Thus, the SiO 2 layer on Si electrodes inhibited electron transfer with approximately the same efficiency as the SEI on metallic Li electrodes. However, the k eff at damaged SiO 2 layers on uncharged Si was stable within the recoding time of the approach curve (1-2 min, curves 5 and 6) in contrast to the damaged SEI on metallic Li electrodes. 23 SiO 2 removal by the probe.- Figure 4a shows a SECM image of a planar SiO 2 -covered Si electrode. As expected from the approach curves, i T was smaller than i T,∞ ≈ 2.3 nA over the entire image. In addition, the variation of i T was rather small because of the planar geometry of the sample. AFM imaging over an area of 95 × 95 μm 2 revealed topographic differences below 20 nm ( Figure S6a and b). After the image, an approach curve at x = 0 μm/y = 0 μm (circle symbol in Figure 4a) was recorded and the obtained k eff fitted to the k eff range of SiO 2 -covered Si electrodes (Figure 4h), although the k eff value was relatively large for SiO 2 -covered Si electrodes. An average SiO 2 layer thickness of 2.3 nm was calculated based on angle resolved XP spectroscopy for pristine Si ( Figure S12).
After t = 55 h, the ME was slightly pushed against the Si electrode by an approach curve close to the circle position (schematic in Figure 4b). Figure 4c depicts the SECM image of the same region as in Figure 4a after the mechanical contact during the approach experiment. The ME current i T strongly increased close to the position of the active electrode area (circle symbol in Figure 4c). The touch was likely to occur between the insulating glass sheath of the microelectrode rather than the active electrode area. Therefore the damaged region was somewhat offset to the position of the active electrode area. Thus, even a slight vertical push of the ME probe toward the Si electrode is sufficient to damage the insulating SiO 2 layer and to increase k eff significantly.
After t = 73 h, the ME approached the Si electrode until the surface was reached (schematic in Figure 4d). The ME was further pushed toward the sample. The resulting force F and pressure p was calibrated with an electronic balance (SI-4). The displacement for Figure 4e was 9 μm (p = 6.7 ± 0.2 MPa) and 2 μm (p = 1.7 ± 0.1 MPa) in Figure S4b. In contrast to the movement depicted in Figure 4b, the ME recorded an image at d = 0, i.e. abrading the SiO 2 layer during the lateral movement of the ME. The corresponding SECM image (Figure 4e) shows an overflow current of ca. 4 nA at several positions. The overflow currents were caused by a direct contact of the active Pt area of the ME and the Si below the SiO 2 layer. They proved indirectly that the upper insulating SiO 2 layer was removed. Figure 4f depicts the positions where the overflow currents were realized as dark squares. Please note that the abrasion was only conducted within 0 < y < 60 μm. After the abrasion in Figure 4e, another SECM image of the same region was recorded (Figure 4g). Within the abraded region as shown in Figure 4e, a strong increase of i T was observed in the subsequent image in Figure 4g. Consequently, the abrasion removed effectively the passivating properties of the SiO 2 layer. In addition, the maximum i T in Figure 4g was larger compared to Figure 4c, although the ME Si separation in Figure 4g was slightly increased. As a consequence, the SiO 2 was more effectively removed by horizontally abrading (Figure 4d) compared to the slight contact at the end of an approach curve (Figure 4b).
The abraded area was also analyzed by AFM and XP spectroscopy. To this end, the abraded area was marked by macroscopic scratches in the shape of a crosshair made by a diamond glass cutter facilitating the relocation of the abraded area in XP spectroscopy and AFM. The AFM analysis of the glass cutter scratches revealed a deformation of 131 nm ( Figure S6d, e), compared to 20 nm height variations of pristine Si (Figure S6a and b). There was no clear evidence for  topography change by abrasion of the ME in AFM imaging (SI 5, Figure S6g-j). The height variations were similar for both pristine Si ( Figure S6b) and ME abraded Si ( Figure S6h). This finding was also corroborated by XP spectroscopy and imaging. XP imaging revealed a significant decrease of O 1s and Si 2p (SiO 2 ) signals ( Figure S9a and b) at the positions of macroscopic scratches. However, the Si 2p (SiO 2 ) signal did not change ( Figure S9c). Hence, the glass cutter scratch removed significantly the SiO 2 layer and this change was detectable after sample transfer. In contrast, there was no decrease of O 1s ( Figure S9d) and Si 2p (SiO 2 , Figure S9f) signals at the ME-abraded region. Thus, the decrease of the passivating properties of the native SiO 2 layer was not associated with significant changes of the elemental composition as detectable by XP imaging. Angleresolved XP spectroscopy was used to determine the thickness of the SiO 2 layer which was 2.3 nm for the pristine sample and 2.6 nm after exposure to the electrolyte ( Figure S12). In conclusion, the AFM and XP spectroscopy results suggest that the ME abrades partially the SiO 2 but does not fully remove it, yet a very significant decrease of the passivating property is observed. This is understandable because SECM feedback can be well maintained at partially blocked surfaces.
Approach curves were recorded after Figure Figure 4a. However, the value was located at the lower limit of the k eff values for the damaged Si electrode. It is likely that the removal of the SiO 2 layer at position i during the abrasion was not very effective because i T is in general relatively small at the line y = 0 in Figure 4e. The k eff value of the smallest current position ii fits to the range of SiO 2 -covered Si in Figures 4g  and 4h. This is expected because the SiO 2 layer at the position ii was neither damaged by abrasion nor by a touch after an approach curve. The opposite holds for the position iii in Figure 4h. Because the SiO 2 was removed locally and extensively, the large k eff value was found.
Because of the large k eff values of the abraded region in Figure 4g, this region was very suitable to study the SEI formation on bare Si. Since the SEI will cause a decrease of the electron transfer rate using DBDMB, 22,49,50 the k eff values in these regions will still be large enough in order to distinguish locally the electron transfer rate constants. On the contrary, the k eff values of the region defined by 0 < x/μm < 240 and 130 < y/μm < 240 were so small that a further decrease of k eff cannot be quantified reliably. Figure 5 show the temporal development of the passivating property for the same sample region from which Figures 4a, 4c and 4g were obtained. The shape of the reactive region in all panels of Figure 5 was similar to that in Figure 4g, which had been recorded 15.3 h before Figure 5a at the same working distance. In order to identify possible artifacts due to the mechanical interaction between the ME body and the Si surface, the working distance was decreased by 1 μm in Figure 5d. In case of mechanical interaction this would increase the intensity of abrading or enlarge the damaged area in case of a slight sample tilt. However, those effects were not observed proving the unrestricted motion of the ME in all panels of Figure 5. Consequently, the local reactivity of both the mechanically damaged and undamaged SiO 2 -covered regions remained stable over hours in 5 mM DBDMB 1 M LiClO 4 in PC solution. This behavior of uncharged Si electrodes differed clearly from SEI-covered metallic Li in the same electrolyte, 23 charged SEI-covered graphite composite 22 and uncharged pristine graphite composite electrodes 21 in 5 mM DB-DMB 1 M LiPF 6 ethylene carbonate (EC):diethyl carbonate (DEC) 1:1 solution. In case of graphite composite electrodes, local swelling of the uncharged graphite composite electrodes due to binder-solvent interactions complicates the image interpretation. This is not an issue for SiO 2 -covered and SiO 2 -free Si electrodes. The extended scale of Figure 5d in comparison with Figure 5c was caused by the decrease of d by 1 μm, which increased or decreased i T depending on the local electron transfer rate constant (c.f. approach curves in Figure 3).

Long-term stability of SiO 2 removal.-The SECM images in
A small decay of i T (x, y) occurred during the imaging experiments (Figures 4g and 5) and can be explained by a slow change of the solution composition and slow passivation of the ME during prolonged imaging. Therefore, the freshly exposed Si surfaces are at least stable over hours after the removal of SiO 2 and can be used to study the impact of SEI formation at the Si electrodes. In fact, the continuous i T decrease was rather small compared to EC-based solutions. 21,23 The complete local removal of the SiO 2 would expose the underlying Si, which reacts with H 2 O to form SiO 2 . 51 However, this reaction is rather slow even in pure H 2 O solvent, where oxide formation was not detected after 5 min of immersion by Fourier transformed infrared attenuated total reflection spectroscopy. 34 In fact, the SiO 2 layer grew continuously and reached a thickness of 1.5 nm after 69 d. Thus, the quick H 2 O rinsing steps after the HF etching do not trigger significant SiO 2 formation.
Moreover, Si oxidation occurs also in dry nonaqueous electrolytes, where oxygen for the SiO 2 formation stems either from trace amounts of H 2 O or from the oxidative decomposition of the electrolyte components. 51 According to Chazalviel et al., 52,53 Si gradually oxidizes in several nonaqueous electrolytes including PC due to χ(H 2 O) = 10 ppm. The oxidation forms SiO 2 islands (rather than continuous layers) that grow in thickness and lateral extension during further oxidation. 33,34,52,53 After 7d of immersion in several nonaqueous electrolytes, the SiO 2 islands are ca. 0.6 nm thick and cover about 60% of the surface area. 53 Consequently, the long-term stability after mechanical damage of the SiO 2 layer within 1 d in Figure 5 was due to the slow Si oxidation process and the insular growth, i.e. the SiO 2 layer was not continuous and too thin in order to affect k eff significantly. In addition, approach curve fits revealed no significant changes of k eff during 55 h ( Figure S2). SiO 2 removal by hydrofluoric acid dip.-Although the use of the ME probe to abrade the SiO 2 layer is useful for planar Si electrodes, the ME probe would damage the micro-and nanostructure of Si electrodes. Consequently, the SiO 2 should be removed specifically by chemical means and the method of choice is the hydrofluoric acid dip. 46,48,54 XP spectra (SI-6) showed that there was considerably less SiO 2 after HF dip than on pristine, oxide-covered Si. Figure 6a shows a SECM image of location 1 of a planar etched Si electrode. Due to the large difference between the maximum and minimum i T of 6.19 nA, the difference in electron transfer rate constants was very large. Approach curves were recorded at the positions i (x/μm, y/μm) = (0, 0) and vi (0, 240). The local k eff values of 1.3 × 10 −4 and 4.7 × 10 −4 cm s −1 are graphically shown in Figure 6f, column 4. They demonstrate that the electron transfer was clearly larger compared to the SiO 2 -covered Si electrode (Figure 6f, column 1; Table I). Thus, the electron transfer rates were increased at all area of Figure 6a (Figure 6f, column 4, lower circle symbol) was obtained for the assumption of d = 0. Although the latter scenario was rather unrealistic due the known distances of positions i and vi and the fact that no shortcircuit current was observed, the calculated k eff demonstrated that the electron transfer rate constants were very heterogeneous within this image frame. The overall range of electron transfer rates was similar for location 1 of sample A (Figure 6f, column 4; Table I) compared to the mechanically damaged Si electrode (Figure 6f, column 1; Table I).
There was no experimental evidence of a mechanical damage by the probe before the recording of Figure 6a. At t = 11.1 h the ME was moved toward the planar Si electrode until the ME touched the surface (Figure 6d). Subsequently, an image was recorded and the ME abraded the SiO 2 surface layer by the lateral movement. Afterwards, the ME was retracted to the same height as used in Figure 6a and the recorded image is shown in Figure 6c. Although Figures 6a and  6c show the same location of an etched Si electrode, the i T values of the SECM image after mechanical damage (Figure 6c) were more homogeneous. The homogeneity of i T was in agreement with the rather small range of k eff values (Figure 6f, column 5; Table I) of positions i, ii, iii, v and vii, which were distributed over the whole SECM image (Figure 6c). Consequently, the mechanical damage by the ME probe after the HF dip further increased the k eff values by additional removal of the SiO 2 . For instance, k eff at position i (x = 0 μm, y = 0 μm) increased from 1.3 × 10 −4 cm s −1 (Figure 6f, column 4) to 4.6 × 10 −3 cm s −1 (Figure 6f, column 5). Thus, the additional removal of SiO 2 by the probe confirmed that the SiO 2 layer was not completely removed by the HF dip. Figure 6b shows the SECM image of the neighboring region of Figure 6a as indicated by the x axis. In addition to position vi (240, 0), also the k eff values at position ix with minimum i T (465, 30) and position viii with maximum i T (370, 0) were characterized in more detail (Figure 6f, column 3). The range of k eff values was significantly smaller in Figure 6b compared to location 1. A clear increase of electron transfer rate constants compared to the SiO 2 -covered Si was detected after the preceding HF dip. The SECM images and k eff values of sample A demonstrate that significant local differences in k eff values can be characterized on the μm scale using the SECM with DBDMB as redox mediator. Therefore, sample A was chosen for the discussion. Furthermore, Si electrodes with inhomogeneous coverage by SiO 2 such as sample A are interesting for the investigation of the SEI formation on Si electrodes, because the highly debated conversion of SiO 2 41,55,56 during lithiation can be addressed. Figure 6e shows a SECM image of another etched Si electrode. Although the current difference of 3.0 nA between the minimum and maximum i T was still significant, the quantified range of k eff values at positions i-iii was very small (Figure 6f, column 2; Table I) compared to sample A. Local differences in k eff values after the HF dip were expected, because the native SiO 2 coverage before the HF dip is heterogeneous [45][46][47] and the local etch rate depends on the HF concentration, 57 which changes locally during the etch process depending on the reaction rate and mass transport. Moreover, the etch rate depends on the local boron dopant concentration. 58 In addition, the quantified k eff values were close to the maximum k eff values, which were obtained after mechanical damage (Figure 6f, column 1 and 5). Therefore, the SiO 2 layer was effectively removed over the whole area of the SECM image in Figure 6e by the HF dip.
The reason for the outstanding behavior of sample A remained unknown. Due to the fact that all k eff values increased significantly above the upper limit for the native SiO 2 -covered Si for locations 1 and 2 of sample A (Figure 6f, column 3 and 4), both locations were wetted by the HF solution excluding the option that enclosed air bubbles prevented the contact between the SiO 2 surface and the HF on the μm scale.

Conclusions
In this study the electron transfer at microstructured and planar Si was characterized using the feedback mode of SECM and 2,5-di-tertbutyl-1,4-dimethoxybenzene (DBDMB) as redox mediator. Approach curves and SECM images demonstrate that the electron transfer rate constants at pristine Si are relatively small due to the native SiO 2 surface layer. In addition, the electron transfer rates are locally het-  Figure 4), for the HF treated sample B (column 2) and locations 2 (column 3) and 1 (column 4) of sample A. Column 5 shows the k eff for the HF treated location 2 of sample A after additional mechanical damage by the ME. Approach curves in Figure S1. erogeneous because of the heterogeneous coverage of SiO 2 . The investigation of SEI formation will require a removal of the SiO 2 layer prior to recording SECM approach curves because the k eff values at oxide-covered Si are close to the quantification limit. Otherwise it would be impossible to detect reliably a further decrease due to the formation of the solid electrolyte interphase (SEI) upon charging of the Si electrode. The SiO 2 layer is at least partially removed by a mechanical contact during a gentle vertical touch, e.g. at the end of an approach curve or by abrading with the microelectrode probe body. AFM and XP spectroscopy results indicated that the SiO 2 layer was abraded. After mechanical damage or chemical SiO 2 removal, the electron transfer rate constants increase strongly, but remain hetero-geneous and stable over hours. Thus, the fully and partially etched Si electrodes can be used to study the SEI formation and to address the debated conversion of SiO 2 during lithiation.
was funded by Deutsche Forschungsgemeinschaft (INST 184/144-1 FUGG) and the State of Lower Saxony. We thank Infineon Technology Austria AG and the Institute for Electron Microscopy and Nanoanalysis, Graz University of Technology and Graz Centre for Electron Microscopy. M. St. and M. W. thank the Austrian Federal Ministry of Science, Research and Economy, and the Austrian National Foundation for Research, Technology and Development for financial support. In addition, the authors are grateful to Sebastian Matz (Carl von Ossietzky University of Oldenburg) for the hydrofluoric acid dip.