Comparison of Electron Transfer Properties of the SEI on Graphite Composite and Metallic Lithium Electrodes by SECM at OCP

The passivating properties of solid electrolyte interphases (SEI) at metallic lithium were characterized using the feedback mode of scanning electrochemical microscopy (SECM) and 2,5-di- tert -butyl-1,4-dimethoxybenzene (DBDMB) as redox mediator at OCP. The SEI at Li allows electron transfer toward DBDMB with ﬁnite rate. In comparison to charged graphite composite electrodes, the electron transfer rate tends to be smaller at Li. Both, graphite composite and Li electrodes, show a local variation of electron transfer rates and temporal changes within a time span of hours. The long-term changes of SEI passivity at metallic Li are dependent on the solvents in the liquid electrolyte. In addition, signiﬁcant short-term changes of SEI passivity occur at both electrodes. However, the frequency of such events is smaller for metallic Li compared to graphite. A strong decrease of SEI passivity and a strong increase of ﬂuctuations in the passivating properties are observed when the microelectrode mechanically touches the metallic Li and damages the SEI. The changes of SEI passivity by a mechanical touch are orders of magnitude larger compared to spontaneous changes. A local SEI damage by the microelectrode decreases not only the SEI passivity locally, but also a few hundreds of μ m apart.

Li metal is currently an electrode material of interest for rechargeable lithium-air and lithium-sulfur batteries. 1,2 It is promising because of its high theoretical specific capacity 3 of 3860 mA h g −1 and smallest electrochemical potential of −3.040 V vs. SHE. Li metal reductively decomposes electrolyte molecules upon contact, because the potential of Li exceeds the stability window of the electrolytes. 4 The decomposition products form a solid electrolyte interphase (SEI) on top of the metallic Li. 5 The properties of the SEI are very important for the performance of the Li negative electrode, because the SEI affects Li dendrite growth. 4 The tendency of Li to form dendrites or high surface area lithium during galvanic deposition is one major drawback of Li metal and causes safety concerns regarding this electrode. The second major drawback is the low coulombic efficiency caused by the ongoing lithium corrosion and sensitivity to SEI passivity. 6 Thus, SEI passivity is a relevant parameter for practical applications.
Since the SEI properties are significant for battery performance, substantial ex situ, in situ and in operando techniques were applied for SEI investigation in general. 7 SEI formation on graphite occurs mainly in the first cycle because of the rather stable graphite host structure. 4 In contrast, the SEI on metallic Li is subject to continuous reformation upon cycling. Despite this significant difference, both graphite and Li metal are covered by similar SEIs. 8 In this study the SEI passivity is characterized by in situ scanning electrochemical microscopy (SECM) 9 using 2,5-di-tert-butyl-1,4-dimethoxybenzene (DBDMB) as a redox mediator. DBDMB was introduced by Dahn et al. 10 as overcharge protection agent for lithium ion batteries (LIB) and turned out to be an excellent choice as SECM mediator in organic solvents as well. 11 Scanning probe techniques are frequently applied for battery research. 12,13 Among them are in situ atomic force microscopy (AFM) 14,15 and in situ scanning vibrating electrode technique (SVET) 16,17 for the investigation of Li metal. SVET provides information about the local Li + ion transfer, whereas SECM is capable of selectively probing the local electron transfer rates in the electrolyte environment. * Electrochemical Society Active Member. z E-mail: gunther.wittstock@uni-oldenburg.de In the last years significant progress was made by using the SECM and related techniques to study LIB electrodes and related processes. Takahashi et al. 18 mapped redox activity of LiFePO 4 positive electrodes with a resolution of 100 nm using the scanning electrochemical cell microscopy (SECCM). Co 2+ dissolution and O 2 release from LiCoO 2 positive electrodes were analyzed in ionic liquids by the sample generation/tip collection mode of SECM. 19 Xu et al. 20 characterized the Li + ion transport from LiCoO 2 positive electrodes by reductive currents at the probe. Using ferrocene as a redox mediator Zampardi et al. investigated the electron transport at SEI covered TiO 2 paste negative electrodes 21 and SEI covered glassy carbon model negative electrodes. 22 In addition, a combined SECM/AFM setup was developed to study the SEI on glassy carbon model electrodes. 23 Bülter et al. 11 characterized spatiotemporal changes of the SEI passivity of charged graphite composite negative electrodes.
There are two major reasons for our motivation to study the passivation films on battery electrodes. (i) It is generally assumed that the SEI both on graphite 24,25 and metallic Li 3,8 is electronically insulating. Thus, the general operation principle of overcharge protection agents is in question, 26 because the oxidized species must be reduced at the SEI covered negative electrode. Consequently, one aim is to characterize the electron transfer of redox shuttles at the SEI covered negative electrode, which may be locally different due to heterogeneities in the SEI composition according to the model of Besenhard and Winter. 27 (ii) The long-term goal is to understand the reduction of electrolyte components at negative electrodes. Electrolyte reduction is essential for SEI formation, however, if occurring continuously it is detrimental for battery stability and the understanding of its spatiotemporal behavior would guide steps to improve battery performance by improved electrode and/or electrolyte components.
DBDMB is a mediator of choice to study electrolyte reduction because its functional groups are rather similar to typical electrolyte solvents like ethylene carbonate (EC) and diethyl carbonate (DEC) etc. This is in contrast to other commonly used SECM mediators such as ferrocene or ferrocene derivatives. 26,[28][29][30] Experimental Electrodes preparation.-The preparation and characterization of the graphite composite electrodes (Figure 1a) was described in detail elsewhere. 11 Briefly, electrodes were prepared with a composition of 81 mass-% graphite, 6 mass-% carbon black and 13 mass-% polyvinylidene fluoride (PVDF). All three solid components were thoroughly mixed in the dry state under dynamic vacuum (<100 mbar, R02Vac intensive mixer, Eirich, Hardheim, Germany) for 3-5 min. N-methylpyrrolidone (AppliChem GmbH, Darmstadt, Germany) was slowly added under continuous stirring and vacuum until a solid mass concentration of 0.44 g/cm 3 was reached. Stirring was continued for another 2 min before opening of the vacuum vessel.
Graphite composite electrodes were produced from the slurry by doctor blade coating on a continuous coating machine (Werner Mathis AG, Oberhasli, Switzerland) on a 20 μm thick, electrochemically roughened copper foil (Carl Schlenk AG, Roth, Germany). The solvent was then removed from the wet film (approximately 150 μm initial thickness) in a two-step process by heating using infrared radiation and subsequent hot air drying. Finally, calendaring of the electrodes was carried out at 100 N mm −1 line pressure. Final loading of the electrodes was 8.5 mg cm −2 with a final thickness of 80 μm (Figure 1a). Layer porosity amounted to 50%. 11 The graphite particles had a specified average size of 32 μm. 31 The R a = 2.5 μm value for quantifying the overall roughness was calculated according to DIN EN ISO 4287:1997. 11 Formation of electrodes.-Pouch cells were constructed using the graphite composite as working electrode and a lithium foil (BASF SE, Ludwigshafen, Germany) as counter electrode for electrochemical conditioning in an Ar-filled glove box. The graphite electrodes and Li metal were cut to identical size and stacked with a layer of glass filter (Whatman GF/A, GE Healthcare, Little Chalfont, UK) as separator. The electrolyte was 1 M LiPF 6 in EC:DEC 1:1 (LP40, Merck KGaA, Darmstadt, Germany).
Electrochemical cycling was carried out between 0.01 V to 1.5 V vs. Li/Li + at galvanostatic conditions (corresponding to 0.5 C). In the first half cycle, the graphite electrode was lithiated and its voltage was 0.01 V vs. Li/Li + . In the second half cycle, the de-lithiation took place and was interrupted when the electrode reached 1.5 V (vs. Li/Li + ). For SECM investigations, cycling was finished with a fully lithiated graphite electrode (i.e. after 5.5 cycle), followed by a transfer to an Ar-filled glove box and dissecting immediately prior to SECM investigation. We also investigated samples after 0.5 cycles in the pouch bag cell, without finding pronounced differences in the behavior. During this step, the graphite electrode was always covered by a layer of the liquid electrolyte LP40.
Scanning electrochemical microscopy setup.-The scanning electrochemical microscope (SECM) was operated under the SECMx control software developed in house. 32 It used 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 custommade plexiglas bell (Figure 2a). 33 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 microelectrodes (MEs) of radius r T ≈ 13 μm (Figure 2b), the graphite composite or Li metal electrodes as samples ( Figure 1) and a platinum wire and a silver wire as auxiliary and reference electrodes, respectively. 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 electrodes were grinded using a  Micro Grinder EG-400 (Narishige, Tokyo, Japan) and polished using rotating wheels with mircropolishing 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 0.4 mL electrolyte solution. The basic electrolyte was LP40 or 1 M LiClO 4 in propylene carbonate (PC, BASF SE, Ludwigshafen, Germany) with 5 mM DBDMB as redox mediator.
Scanning electrochemical microscopy measurements.-The ME was polarized to 4.1 V vs. Li/Li + , where DBDMB is oxidized under a diffusion-controlled condition (Figures 2b/2c). The timer was set to zero, when the working solution was filled into the SECM cell. The approach curves of Figure 3 were recorded with a step size of 0.5 μm and a delay of 1.0 s between the translation and the data acquisition giving an average translation rate of 0.25 μm s −1 . Approach curves were fitted to the analytical approximations recommended by Cornut and Lefrou 34 and are shown in normalized current I T = i T /i T,∞ vs. normalized distance L = d/r T (i T , ME current [nA]; i T,∞ , ME current in the bulk [nA]; d, working distance [μm]; r T , radius of the active electrode area [μm]. Prior to fitting, the radius of the electroactive area r T and the RG were determined by confocal laser scanning microscopy (CLSM) using CSLM TCS SP2 AOBS (Leica Microsystems GmbH, Wetzlar, Germany). For fitting, the uncertainties of 0.5 μm for r T and 1.5 for RG with respect to the CLSM results were considered. κ, i T,∞ and d 0 (within reasonable range) were varied in order to fit the experimental approach curves. κ is the normalized first-order heterogeneous rate constant and d 0 is the point of closest approach of the ME to the sample. The apparent heterogeneous rate constant k eff was calculated by Equation 1 using a diffusion coefficient of D = 2.15 × 10 −6 cm 2 s −1 for DBDMB in LP40 electrolyte and D = 8.15 × 10 −7 cm 2 s −1 for DBDMB in 1 M LiClO 4 in PC. The diffusion coefficients were determined by the steady-state diffusion limited current of the ME.

Parameter Value
Step size x/μm 5 . 0 0 Scan length l x /μm 240.00 / 480.00 Average translation rate v x a /μm s −1 3.85 Delay between translation and current recording/s 0.3 Step size y/μm 15.00 Scan length l y /μm 240.00/480.00 Average translation rate v y a /μm s −1 11.54 a Average translation includes delay between translation and current recording. Table I summarizes all parameters of the SECM image acquisition. The distance between ME and sample electrode d is given for the horizontal position (x/μm, y/μm) = (0, 0) of the image.

Results and Discussion
Approach curves.-Approach curves in normalized coordinates are plotted in Figure 3. The ME current depends strongly on the type of sample and on d: When the ME approaches the uncharged graphite composite electrode ( Figure 3, curve 1), the oxidized DBDMB + is reduced at the graphite composite electrode to DBDMB (Figure 3, inset). The reduction of DBDMB + at the graphite composite electrode yields an additional flux of DBDMB at the ME and thus increases i T with decreasing distance. Compared to the case of diffusion-controlled reactions at the ME and the sample ( Figure 3, curve 6), the electron transfer is significantly slower at graphite composite electrodes. SEI formation ( Figure 3, curves 2/3) decreases the electron transfer rate significantly compared to uncharged graphite composite electrodes. 11 Curve 2 and 3 indicate the curves with the largest and smallest electron transfer rate out of ca. 30 similar curves indicating that the electron transfer rate might vary locally on charged graphite composite electrodes. 11 The reasons for the local variation in electron transfer rates could be differences in local composition of the graphite composite electrode and local varying properties of the SEI. Figure 3, curves 4/5 show that the electron transfer rate varies locally on metallic Li, too. The local variation of electron transfer rate for Li is reasonable, since the SEI is inhomogeneous with respect to composition and thickness also on smooth metallic Li. 3,4,14 The expected thickness variations of the SEI are within a few nanometers which is below the resolution of SECM approach curves (about ± 1 μm resolution with r T = 12.5 μm). In principle identical location AFM-SECM studies or studies with combined AFM-SECM probes could provide such information but the mechanical interaction between the probe and the SEI may influence their dynamic behavior. The calculated k eff values of 3.6 × 10 −5 -3.1 × 10 −4 cm s −1 for Li (Table II) correspond to the curve 5 (minimum value) and curve 4 (maximum value) in LP 40 electrolyte. The approach curves toward Li tend to provide smaller electron transfer rate constants compared to charged graphite composite electrodes ( Figure 3, Table II) although the ranges of the observed electron transfer rate constants for charged graphite and Li overlap. Since the Li foil was not a composite (in contrast to the graphite electrode), the recorded differences can be clearly assigned to the differences in local SEI passivity. Topographic roughness which might be an interfering factor in graphite composite electrodes is not important on Li metal foils. On average k eff of SEI covered charged graphite is smaller than k eff of pristine graphite by a factor of 6 × 10 −2 while k eff of Li is smaller by a factor of 2 × 10 −2 compared to pristine graphite (Table II). The trend of increased passivity of Li metal compared to charged graphite can be explained by its lower potential. In addition, surface functional groups on graphite (depending on the preparation history) may change the way the initial SEI is formed or reformed. The potential of graphite at the end of a charging cycle is set to 0.01 V vs. Li/Li + in order to avoid Li plating. 35 Consequently, the reducing power of metallic Li is higher and this might result in a SEI with higher passivity. Although the approach curve on Li ( Figure 3, curve 5) provides the smallest electron transfer rate, the curve is clearly different from the approach to an inert surface, where no reduction of DBDMB + occurs (curve 7). Thus, electron transfer at SEI-covered metallic Li takes place continuously when using DBDMB as a redox mediator. As a consequence, DBDMB is useful to study electron transport at Li. The usefulness of DBDMB to investigate the electron transport at SEI covered graphite was shown before by different techniques. 11,36,37 The electron transfer rate at SEI-covered metallic Li depends on the electrolyte composition (Table II). For 1 M LiClO 4 in PC the k eff range with 3.6 × 10 −5 -2.2 × 10 6 cm s −1 is below the range of k eff for 1 M LiPF 6 in EC:DEC (1:1). Thus, the SEI passivity is higher for 1 M LiClO 4 in PC compared to 1 M LiPF 6 in EC:DEC (1:1). Figure S1 shows the κ range of approach curves in 1 M LiClO 4 PC electrolyte solution.
Long-term stability of SEI passivity in LP40 electrolyte.-As reported before, the SEI passivity changed locally at some positions over a time span of hours in LP40 electrolyte after the transfer of lithiated graphite electrodes from the pouch bag cells to the SECM cell and addition of further electrolyte with DBDMB. 11 In order to compare the behavior of charged graphite electrodes with metallic Li electrodes, a sequence of images of an identical region on Li was recorded over a time span of 9.8 h at a constant height ( Figure 4). Within this sequence from Figure 4a to Figure 4e one observes regions where the current increases relative to the average current of the image frame (marked by upwards oriented arrows). Since swelling is not expected for a metallic electrode and SEI formation may lead only to thickness variations well below a micrometer at the open circuit potential (OCP), the partially strong changes cannot be caused by topographic changes, but must result from local changes of the electron transfer rate. Thus, the upwards oriented arrows in Figure 4 show an increase of the electron transfer rate, i.e. the SEI passivity decreases. At the same time the downwards oriented arrows indicate an increase of SEI passivity on other regions of the same image frame. There are also regions in the image frames, in which no changes are observed over the whole time span ( = marker). In summary, the passivity of the SEI on Li shows a dynamic behavior on the micrometer scale over a time span of hours after electrolyte addition similar to the already reported behavior of charged graphite composite electrodes in LP40 electrolyte. 11 Cohen et al. 14 demonstrated by in situ AFM at OCP morphological changes of the surface film on metallic Li within 50 minutes in EC/DEC solution. The morphological changes were as large as 266 nm within an area of 1 μm × 1 μm. This corresponds to changes of local composition and might account for the local change of electron transfer rate. On the other hand the topographic variation of 266 nm corresponds to only 0.02 r T and therefore it is negligible for the SECM imaging process. Changes of the Li electrode surface resistance determined by electrochemical impedance spectroscopy 6 at OCP in EC/DEC over a time span of several days are in line with changes of the electron transfer rate constant. Both charged graphite composite 11 and Li metal reveal changes of the SEI passivity over a time span of hours. This might be explained by the similarity of the formed SEI in the same solution. 8 The maximum of i T decreases from 4.0 nA in Figure 4a to 3.5 nA in Figure 4b along with an decrease of the average current in the SECM images. The tendency of decreasing average currents with time ( Figure 4) is related to impurities in the electrolyte which adsorb on the Pt surface of the ME probe. Since each image is built up of consecutively measured data points and the relative time difference between the measured locations of one image frame is constant, the continuous decrease of i T due to impurities does not affect the analysis of relative current changes within one image frame of the sequence shown in Figure 4.

Long-Term stability of SEI passivity in 1 M LiClO 4 PC
electrolyte.- Figure 5 shows an image sequence recorded above a metallic Li foil in 1 M LiClO 4 PC solution. The current difference between y = 0 and y = 240 μm is mainly caused by a height difference. Fits of approach curves at x = 0/y = 0 and x = 240 μm/y = 240 μm (arrows in Figure 5a) reveal a height difference of 10 μm, whereas the ME Li foil separation is the smallest at x = 0/y = 0 with 12 μm. The distance was decreased between Figures 5a/5b by 6 μm. On one hand, the overall current decreased because of the smaller d (compare Figure 3, curves 3/4) and on the other hand the overall shape remained similar because of the relatively small range of local variation of k eff (Table II). In contrast to LP40 solution (see above), there is not a significant decay of i T with t in 1 M LiClO 4 PC solution.
Consequently, the absolute i T values can be analyzed in 1 M LiClO 4 PC solution.
The Figures 5b-5h are consecutive images recorded at the same d of an identical region. Since the distance was kept constant, no changes are expected from topographic influences. 19.5 h after addition of the electrolyte to the cell i T increased strongly at a specific region in Figure 5c (square symbol). Afterwards, i T of this specific region (square symbol) decreased again and in Figure 5f at 22.0 h i T is again almost identical to i T in Figure 5b, which were recorded 18.2 h ago. Thus, the original SEI passivity at x = 240 μm/y = 105 μm in Figure 5b is reestablished within 2.5 h after the detection of a significant decrease of SEI passivity in Figure 5c. According to Figures 5a-5f only the region around the square symbol shows long-term changes at OCP without a mechanical touch by the ME.
This observation is shown more quantitatively in Figure 6 where the average current i T,av (x,y,t) between the forward and reverse scan is plotted for the selected location with the largest current (the square symbol in Figure 5 at x = 240 μm/y = 105 μm). This plot was chosen in order to minimize the effect of short-term fluctuations of SEI passivity. The error bar in Figure 6 is based on a 30 pA noise for each scan. i T,av increases between 1.8 nA ( Figure 6, label b) and 2.9 nA (Figure 6, label c) by 57%. Since i T,av increase relative largely, the SEI passivity decreases significantly. Assuming an average k eff of 1.9 × 10 −5 cm s −1 (Table II) for the position x = 240 μm/y = 105 μm (square symbol) in Figure 6, label b, d amounts to 12 μm for i T,av = 1.8 nA. When i T,av increases to 2.9 nA in Figure 6, label c and d is constant, k eff amounts to 2.3 × 10 −4 cm s −1 . Thus, k eff for the square symbol in Figure 6, label c increases by a factor of 12 (Table III). Please note that the estimated k eff ≈ 2.3 × 10 −4 cm s −1 for Figure 6, label c exceeds the k eff range for metallic Li in 1 M LiClO 4 and is within the range for both metallic Li and charged graphite composite for LP40 electrolyte solution. The ratio between the estimated k eff for Figure 6, labels c-f and the estimated k eff for Figure 6, label b decreases from 12 to 1 (Table III). The average time difference between the labels c-f of Figure 6 (i.e. the time difference between the images of Figures 5c-5f) amounts to 0.8 h. Therefore, the ratios of the k eff values in Table III are halved between 0.8 h of two consecutive images starting from Figure 6, label c.    Figure 6. Considering the noise of 30 pA for each single scan, i T,av was considered as stable.
In comparison, there were significant differences of the detected long-term SEI passivity developments between LP40 and 1 M LiClO 4 PC solutions: A larger fraction of the 240 × 240 μm 2 image frame showed long-term changes in LP40 compared to 1 M LiClO 4 PC electrolyte, where the SEI passivity of most regions did not change signif-icantly. For LP40 electrolyte solutions only unidirectional long-term changes were detected. Repassivation was not observed. In contrast, i T of the square symbol region in the case of 1 M LiClO 4 PC electrolyte solution increased first and decreased later, i. e. the long-term change was bidirectional.
In contrast to EC/DEC solutions, no morphological changes upon storage were detected by in situ AFM in PC solutions. 14 However, the investigated PC solutions contained LiPF 6 and LiAsF 6 as conducting salt in contrast to LiClO 4 used in this study. Therefore, the conditions are not comparable. The fact that no morphological changes were observed for PC solutions 14 could account for the large regions without significant changes of long-term SEI passivity.
Short-term stability of SEI passivity.- Figure 7a shows the forward image of a Li foil with dimensions of 480 μm × 480 μm. The  measured reactivity was heterogeneous. Since i T decreased with decreasing vertical distance to the Li sample ( Figure 3), a protruding spot on the Li foil resulted in a smaller i T value. Thus, Figure 7a would represent the inverse of topography. However, since also the local electron transfer rate differed (Figure 3), the local i T value in Figure 7a is a result of differences in electron transfer rate and in topography. Consequently, it cannot be decided based on Figure 7a alone, if the image shows topographic and/or reactivity features.  Figure 7c presents the absolute current difference between the images of Figures 7a and 7b. At most positions the current difference i T remains below 0.1 nA, whereas i T amounts to 0.9 nA at positions marked by arrows. Since the working distance is identical in forward and backward images of the same run, the changes of i T are related to changes of the electron transfer rate only. Thus, short-term changes of the SEI passivity occur on Li.
A major observation on SECM feedback images of charged graphite composite electrodes was the occurrence of peaks. 11 Figure 8 depicts our interpretation of these peaks. When the reverse line scan is recorded (Figure 8b), a strong increase of i T is locally observed. An increase of i T is equivalent to a decrease of SEI passivity. Because the increase of i T can be as large as i T of a pristine graphite electrode without SEI, the SEI must had been at least partially destroyed. Because i T during the 3 rd line scan is identical to i T of the 1 st line scan (Figure 8c), the local SEI passivity is recovered and thus the SEI is reformed.
The occurrence of such peaks in SECM experiments has important implications for battery applications: The local destruction of the SEI leads to further reaction of electrolyte with the lithiated graphite (or metallic Li), i.e. a loss of electrolyte and reversible capacity. Thus, Table III. Estimated k eff increase at x = 240 μm/y = 105 μm during imaging in Figure 5.

Image
Label Figure 6 k eff / k eff (Figure 5b an ideal SEI of a high performance battery would show none of the short-term events. Consequently, the characterization of these peak events and the subsequent modification of the electrodes in order to minimize their occurrence represents a route to improved LIBs. 11 In order to count the peak events, a criteria was defined: 11 A deviation between forward and reverse scan is considered as a significant "event" if i T of the forward line scan at a defined position deviates from the current in the reverse scan at the same position by more than 21% of the average image current i T . Figure 9a indicates the 21% threshold by the gray shaded band. Because i T of the solid curve exceeds the threshold at x ≈ 210 μm, four events are counted as indicated by black arrows in Figure 9a. In Figure 9b the reverse line scan follows the forward line scan without exceeding the threshold. This situation is found for most line scans within the recorded image frames. The threshold of 21% i T was chosen in order to overcome experimental imperfections and to count only strong changes of SEI passivity. Figure 10a shows a 2D histogram plot where the local events on Li are summarized over a sequence of nine forward and nine reverse images using a threshold of 21% i T in LP40 solution. Each square represents a measurement position and the gray filling of a square indicates the occurrence of an event during the image sequence. Seven positions showed a single event over the whole images sequence. Consequently, significant short-term changes of SEI passivity take place on metallic Li similar to charged graphite composite electrodes. 11 Recording one event on Li did not indicate a probability to find further events in subsequent image frames. This is in contrast to lithiated graphite (vide infra).
For comparison, Figure 10b shows a histogram plot for a charged graphite composite electrode in LP40 electrolyte. Here the number of events is indicated by the darkness of the circle filling. Similar to Li, there are many positions with no event at all. However, in contrast to Li, there are also positions with multiple events (up to three) which are also close to each other, indicating hot spots of strong shortterm changes of SEI passivity. Despite the fact that the example in Figure 10b was rather inactive (compared to the sample reported in reference 11), the event density is much larger for the charged graphite composite electrode. Figure 11a depicts a plot where the events are summarized over a sequence of eight forward and eight reverse images using a threshold of 21% i T in 1 M LiClO 4 PC electrolyte solution in contrast to the plot for the LP40 electrolyte in Figure 10a. Only three single events are counted and thus the result is similar as for the LP40 electrolyte. Figure 11a is based on the image sequence between 3.8-25.1 h of Figure 5. As demonstrated in the previous chapter, bidirectional longterm changes of SEI passivity took place at SEI covered metallic Li in 1 M LiClO 4 PC electrolyte solution. Since only a few single events are located within the square symbol region of the long-term change in Figures 5 and 6, the long-term change over at least 2.5 h is not accompanied by significant short-term changes of SEI passivity. Figure 12a shows a SECM feedback image of a Li foil. i T is decreasing in general from left to right, i.e. with increasing x values, because of a small tilt of the Li foil in x direction. i T of Figure 12a was generally relatively small, because d is very small (Figure 3, curves 4/5). Assuming an average k eff of 1.74 × 10 −4 cm s −1 (Table II) 34 Thus, the tilt is about 2 μm in z for a change of 240 μm in x. The movement of the ME relative to the Li foil in z direction is indicated by arrow I (Figure 12a).The i T values of all linescans in Figure 12a were continuously decreasing with increasing x because of the tilt, except lines y = 0, 15 and 30 μm. At y = 0 there is a significant increase of i T after x = 180 μm. Since all other lines are decreasing with x except y = 15 and 30 μm and d amounts to 0, the strong increase of i T is explained by a mechanical touch of the ME at the SEI covered Li surface at these positions. Figure 12b depicts the reverse image. Similar to the forward image in Figure 12a, i T increased with decreasing x starting at x = 240 μm. However, the increase is much stronger in every line. Especially at the local maxima x = 185 μm/y = 0 μm (Figure 12b, arrow VI) and x = 185 μm/y = 240 μm (Figure 12b, arrow V) the increase of i T amounts to 244% and 329% compared to i T of the forward image at the same positions (Figure 12a). Since the distance was not changed between forward and reverse image, the strong increase of i T reflects a change of electron transfer rate, i.e. a significant decrease of SEI passivity. Figure 12c shows the current difference, i T , between the forward and backward images. The shape of Figure 12c with respect to its main features is similar to the reverse image in Figure 12b. Consequently, strong changes of i T are characteristic for the reverse image. According to Figure 12c, most positions provide i T ≥ 0.2 nA, except a few at ca. x = 240 μm and x = 0. This observation become more evident in Figure 13a where the events are shown using a threshold of 21% i T . In Figure 13a 88.84% of all positions show at least one event. In contrast the results from Figure 10a showed only 7 events over a sequence of 9 images under conditions without mechanical touch between ME and sample. Consequently, only 0.09% of all positions over the whole sequence provided events. Since i T of Figures 12a/12b is relatively small due to the small distance, the threshold was doubled to 42% i T (Figure 13b) in order to decrease sensitivity. Although the event density decreased from 88.84% in Figure 13a to 21.57% in Figure  13b, the event density is still larger by a factor of 2 × 10 2 compared to the condition without mechanical touch by the ME. In conclusion, the event density is significantly affected by mechanical stress by the ME. According to Figures 12c and 13a there are no large i T values or events close to x = 0 and up to x = 25 μm. This is expected since the ME Li foil separation was estimated to be 2 μm at x = 0 and the SEI was not damaged by the ME in this region. The arrows III and IV in Figures 12b/12c and 13a/13b indicate local maxima of i T . These maxima are separated from the region with no event at x = 0 by only a few tenth of μm. Based on the relatively large distance between ME and Li foil at x = 0 of ca. 2 μm, the unavailability of events close to x = 0 and the vicinity of the local maxima to x = 0, it is supposed that the ME did not destroy the SEI at the local maxima of arrows III/IV. Consequently, the damage of the SEI at d = 0, i. e. x ≈ 200 μm, affected not only the passivity locally, but also the SEI passivity ca. 150 μm apart at the positions of arrows III/IV.

Impact of mechanical stress on SEI passivity.-
Further evidence for the assumption that the ME did not touch the Li foil at x ≈ 50 μm is provided by Figure 13b: Using a threshold of 42% i T , the neighbored events in the regions of arrows III/IV are separated from the region with a high frequency of events between 145 ≤ x/μm ≤ 225 of arrows V/VI by a large region without events between 60 ≤ x/μm ≤ 140. Because the threshold 42% i T is relatively large and indicates only significant changes of SEI passivity, the region with a high frequency of events between 145 ≤ x/μm ≤ 225 is caused by a local contact of the ME to the SEI covered Li foil. Since the large region without events between 60 ≤ x/μm ≤ 140 is located in an area with relatively large working distance between ME and Li foil, a contact to the Li foil was unlikely compared to the region with a high frequency of events between 145 ≤ x/μm ≤ 225. Thus, a contact at the arrow III/IV positions in Figures 12b/12c and 13a/13b is not reasonable, since the ME Li foil separation is even larger compared to the region with no events between 60 ≤ x/μm ≤ 140.
Although there is significant evidence for a touch of the ME to the Li foil at x ≈ 200 μm, a current overflow is not observed in contrast to charged graphite composite electrodes 11 for which there might be the following reasons: (i) The overflow current is caused by a direct contact of the sample and the active area of the ME. However, the active area of Pt is surrounded by an insulating glass sheath, which is between five and ten times larger than the active Pt in diameter. Since the surface of the ME is not likely to be perfectly planar oriented to the Li foil ( Figure 13, insets), contact between ME and Li foil is more likely to occur between the insulating glass sheath and the Li foil. (ii) The average roughness of the investigated charged graphite electrodes was 2.5 μm. 11 In contrast, the Li foil is smoother and the roughness is less than a few hundred nm. 14 In addition, the graphite particle arrangement provides higher flexibility to the ME touch and might therefore still touch the active Pt disk at the ME. (iii) Furthermore, a possible tilt of the Li foil will shift the contact positions of the ME to the outer insulating glass sheath.
It is questionable why i T is significantly smaller between 225 ≤ x/μm ≤ 240 compared to the region with a high event frequency between 145 ≤ x/μm ≤ 225 (Figures 12c and 13a/13b) because the ME Li foil separation was supposed to be even smaller for the region 225 ≤ x/μm ≤ 240. Thus, the damage of the SEI by the ME should be at least as large as for the region with a high frequency of events at 145 ≤ x/μm ≤ 225. A possible explanation is provided by the insets of Figures 13a/13b: When the ME is not perfectly parallel to the Li foil and the ME Li foil separation is smaller for the left part of the insulating glass sheath than for the right part, then is the damage to the SEI made by the left part of the glass sheath. Consequently, damage is only made within the region which is left from the turning point of the ME at x = 240 μm, i. e. there will be no damage for x ≈ 240 μm.
Curve 1a in Figure 14a shows an approach curve to metallic Li in 1 M LiClO 4 PC electrolyte solution. The approach curve is rather close to the calculated curve for an inert surface (curve 8) because of the rather small electron transfer kinetics (Table II). At a very small distance between ME and Li foil of L ≈ 0, the dimensionless current I approaches the detection limit, because of contact between the ME and Li foil. After 17 s (Figure 14d) the ME is retracted from the Li foil and the retraction curve is shown in curve 1b of Figure 14a. Curve 1b demonstrates a strong increase of I compared to curve 1a before. For comparison calculated curves for different κ values are shown in Figure 14a. I of the retraction curve 1b at L ≈ 0.05 matches the calculated I for κ = 0.75 (curve 4). Since curve 1a is characterized by κ = 4.0 × 10 −3 , κ increased by a factor of 2 × 10 2 . Thus, the SEI passivity decreased significantly by the mechanical touch of the ME.
In addition, the overall shape of curve 1b in Figure 14a is very different from the calculated curves and consequently it is impossible to fit the curve to the theory. 34 The outstanding shape of curve 1b is a further indication of significant short-term changes caused by the mechanical contact before.
After 18 minutes (Figure 14d) an approach curve is again recorded at the same position (Figure 14b, curve 1c). In contrast to curve 1b,  curve 1c was again matching the theory and the resultant κ of 1.1 × 10 −2 was three times larger than the initial κ of curve 1a. As a conclusion, the SEI passivity was almost recovered 20 minutes after the mechanical touch by the ME. Curve 1d in Figure 14c was recorded 1.3 h after curve 1c at the same position (Figure 14d). κ of curve 1d amounted to 4.9 × 10 −3 and was identical to κ of the initial curve 1a (Figure 14a) considering the accuracy of the fit. Thus, SEI passivity was fully recovered 2 h after mechanical touch by the ME. During the mechanical touch by the ME, the ME was pushed 1.5 μm inside the metallic Li foil indicated by the fit and the overflow currents. Consequently, the repassivation of the SEI on metallic Li occurs within some tens to one hundred minutes after a gentle ME touch.
Although the approach curves of Figure 14 suggest a complete repassivation at x = 0 μm/y = 0 μm 2 h after a mechanical touch by the ME, imaging experiments revealed increased currents within 60 μm around the position x = 0 μm/y = 0 μm after 4 h and 7 h (triangle symbol in Figures 5g/5h). This apparent contradiction might be explained by two facts: (i) Because the ME was pushed into the Li foil for a few μm, permanent topographic changes might occur, which will change i T because of the distance dependence ( Figure 3). The affected area of 60 μm around the position x = 0 μm/y = 0 μm is also reasonable, because the ME including the insulating glass sheath provides a radius of a similar size. (ii) A SECM feedback image of this study provides information for a single position only for a very short time compared to the duration of the entire imaging experiment. Thus, the development of i T at a single position is unknown in between two subsequent image frames. Since the images in Figures 5g/5h were recorded after the approach curves and strong long-term changes of SEI passivity at OCP took place (Figure 5c), there could be an additional long-term change after the approach curves around x = 0 μm/y = 0 μm as a consequence of the contact before.
The occurrence of increased currents at positions of mechanical touch by the ME was also observed for graphite composite electrodes. 11 The measurements at graphite composite in LP40 electrolyte showed that the increased currents during repassivation occurred for at least 14.3 h.
As stated above long-term changes of SEI passivity take place after approaching the ME to contact the Li foil (Figures 5g/5h). Figure  11b characterizes the short-term changes of SEI passivity, which occur during the forward and reverse images corresponding to Figures 5g/5h and three additional images between Figures 5g/5h. Figure 11b depicts only one event. Although this event is located within the region where long-term changes took place in Figures 5g/5h (marked by a triangle), it is not repeated. Therefore, Figure 11b does not demonstrate significant short-term changes of SEI passivity similar to Figure 11a without a mechanical touch. In summary, the event density is only increased during the mechanical contact ( Figure 13). A few hours after the contact, the event frequency is comparable to the situation without preceding mechanical contact ( Figure 11).

Conclusions
In this study the electron transfer at metallic Li was characterized using the feedback mode of SECM and 2,5-di-tert-butyl-1,4dimethoxybenzene (DBDMB) as redox mediator. Li foils showed a local variation of electron transfer rate similar to charged graphite composite electrodes. The SEI at Li and graphite are not electronically insulating for DBDMB as redox mediator. Thus, DBDMB is useful to study the electron transfer at Li. The ranges of electron transfer rate constants at charged graphite composite and Li metal overlap, with the average electron transfer rate at Li being slightly smaller than on lithiated graphite. The SEI passivity at Li was higher for 1 M LiClO 4 in PC compared to 1 M LiPF 6 in EC:DEC (1:1) solution.
We have already reported about the local variation and temporal development of SEI properties on graphite composite electrodes. 11 Several reasons might be responsible for this behavior at composite electrodes: (i) Due to calendaring, mechanical stress is build up within the composite. Releasing the mechanical stress by particle movement may damage the SEI. Electrolyte-binder interactions and electrochemical swelling due to lithiation further increase the stress between the particles of the rough graphite composite electrode. (ii) Inhomogeneous SEI formation on individual particles may be due to different surface structures, functionalization or nucleation phenomena. Release of mechanical stress is avoided in using smooth Li metal electrodes. The inhomogeneous formation of SEI indicates that inhomogeneous SEI formation seems to be an intrinsic property of this interface and does not only depend on preceding mechanical processing (i.e. calendaring, dissecting of cells etc.).
Both metallic Li and graphite composite electrodes showed locations of changing electron transfer rates over a time of several hours of continuous imaging after addition of electrolyte at OCP. Since shortterm and long-term changes of the SEI occur even at the metallic Li foils, the changes are rather caused by the SEI itself than by interactions within a composite electrode in a LIB. A 12-fold spontaneous increase of the electron transfer rate was observed in 1 M LiClO 4 PC solution. The long-term changes of SEI passivity at metallic Li were dependent on the electrolyte: Unidirectional long-term changes were observed in 1 M LiPF 6 EC:DEC (1:1) and bidirectional in 1 M LiClO 4 PC solution.
Similar to charged graphite composite electrodes, metallic Li showed significant short-term changes of SEI passivity over a time span of hours. However, the frequency of such events was much lower at Li. Events were not repeated at the same location independent of the solution composition. A strong decrease of SEI passivity and a strong increase of event density were observed when the ME touched the metallic Li foil. The SEI passivity decreased 200-fold 20 s after mechanical touch by the ME in LiClO 4 PC solution. Thus, the decrease of SEI passivity after mechanical touch was an order of magnitude larger compared to the spontaneous changes at OCP without ME touch. Repassivation occurred within 2 h after touch in 1 M LiClO 4 PC solution. In addition, a local SEI damage by the ME decreased not only the SEI passivity locally, but also a few hundred μm apart.
The use of smooth Li electrodes in this experiment also eliminates the uncertainty connected to the quantitative interpretation of SECM approach curves for rough graphite composite electrodes. Such setups might be ideally suited for in situ studies of the impact of electrolyte additives on the formation potential of SEIs, their passivating properties and stability. A suitable additive is expected to decrease the extent of long-term changes and cause a lower frequency of shortterm changes.