In Situ Quantification of the Swelling of Graphite Composite Electrodes by Scanning Electrochemical Microscopy

The physical swelling of uncharged graphite composite electrodes due to electrolyte-binder interactions is investigated by scanning electrochemical microscopy (SECM) using 2,5-di-tert-butyl-1,4,-dimethoxybenzene as a redox mediator. A series of approach curves at the same location is conducted in order to quantify in situ and locally the physical swelling. The film thickness change δfilm amounted to 9.1 μm on average for a 80 μm thick uncharged graphite composite electrode in LP40 electrolyte between 1.1 and 5.9 h. Curves of δfilm vs. t usually reach a saturation within 12 h. The swelling ratio χ varies from 0.3% to 17.6% for uncharged graphite composite electrodes from the same batch in the same electrolyte. In contrast, the 8 μm thick polyvinylidene fluoride (PVDF) model sample swelled by χ = 99%. Approach curves demonstrate that swelling of the PVDF is the main cause for the physical swelling of uncharged graphite composite electrodes. Both PVDF model sample and uncharged graphite composite electrodes show locally different swelling ratios by SECM imaging. Based on these results a swelling model is proposed, where the uncharged graphite composite electrode swells physically on average by at least χ = 11% and the local topography is changing during swelling. © 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: 10.1149/2.1061514jes] All rights reserved.

Li-ion batteries (LIBs) are currently the most rapidly developing commercial rechargeable batteries. They are mainly used for portable electronics but increasingly find application for electrified vehicles and stationary storage applications because of the high practical energy density, good cyclability and low self-discharge. 1,2 Graphite is used in most commercial LIBs as the negative electrode material. 3 The "graphite electrode" is a composite consisting of graphite particles, conductive agents and binder.
The binder preserves mechanical stability of the composite electrode during battery operation including the stabilization of contacts between graphite particles and conductive agents as well as good adhesion of the composite material to the current collector. 4,5 Consequently, the binder is important for the performance of the LIB electrodes. 6 The binder poly(vinylidene fluoride) (PVDF) is a very stable and prominent example. 4 It is well known that PVDF takes up organic electrolyte due to electrolyte-binder interactions, i.e. both negative and positive composite electrodes containing PVDF do swell. 5,[7][8][9][10] We refer to this process as physical swelling in order to emphasize the difference to volume expansion during Li-ion intercalation to which we will refer as "electrochemical swelling". 11 This paper deals only with the physical swelling. While the electrochemical swelling of composite electrodes including the superb stability of PVDF has been studied, 12 the physical swelling without electrochemical cycling has received less attention. However, the physical swelling changes the properties of the composite electrodes as for instance the decrease of the long-range electronic conductivity. 6 The uptake of organic electrolyte by PVDF can be characterized by ex situ determination of the mass difference before and after contact to the electrolyte. 5,7,8,10 However, in this case the real film thickness change δ film remains unknown. Zhang and Tang determined δ film ex situ and globally by using a micrometer dial caliper. 11 Because the swelling depends on binder-electrolyte interactions and the electrolyte had to be removed for measuring δ film , the values are likely different from the in situ situation. In addition, a time-dependent analysis of δ film is challenging because the composite electrode is removed from the electrolyte environment for the measurement of δ film . As there is some variety in the δ film values a second composite electrode, although prepared identically, might show a different time course of swelling. Beaulieu et al. 13 determined δ film locally by in situ atomic force microscopy (AFM) during electrochemical swelling (lithiation) of a Si 0.7 Sn 0.3 electrode. However, their approach assumes a perfectly * Electrochemical Society Active Member. z E-mail: Gunther.Wittstock@Uni-Oldenburg.de smooth electrode of a few μm thickness. Since practical graphite composite electrodes exhibit a micrometer roughness and a thickness of ca. 100 μm, their approach is not appropriate for determining δ film for physical swelling. Here we propose the use of approach curves in the feedback mode of the scanning electrochemical microscopy (SECM) to determine δ film of graphite composite electrodes in situ and locally. For SECM feedback experiments an additional compound is added to the electrolyte, which is called redox mediator, because it transports electrons between the microelectrode (ME) and the sample of interest. It enters heterogeneous electron transfer reactions at both electrodes. In this study 2,5-di-tert-butyl-1,4-dimethoxybenzene (DBDMB) is utilized as a redox mediator. DBDMB was introduced by Dahn et al. 14 as overcharge protection agent for LIBs and turned out to be an excellent choice as SECM mediator in organic solvents as well. 15,16 Two other redox compounds, 1,2,4,5-tetramethoxybenzene 17 (TMB) and tri(p-bromophenyl)amine 18 (TPA), were tested as SECM redox mediators based on their proposed use as overcharge protection agents.
In the last years significant progress was made by using the SECM and related techniques to study LIB electrodes and related processes. The research fields can be divided into (i) the investigation of chemical species released or produced at the LIB electrodes [19][20][21] and (ii) the analysis of the electron transport at negative electrodes, which are covered by the solid electrolyte interphase (SEI). 15,16,22,23 This work introduces an additional application of SECM in the research field of LIB electrodes and related processes by quantifying in situ the local swell ratio of graphite composite electrodes.

Experimental
Preparation of graphite composite electrodes.-The preparation and characterization of the graphite composite electrodes was described in detail elsewhere. 15 Electrodes were prepared with a composition of 81 mass-% graphite, 6 mass-% carbon black and 13 mass-% poly(vinylidene fluoride) (PVDF). All three solid components were thoroughly mixed in dry state under dynamic vacuum (<100 mbar, R02Vac intensive mixer, Eirich GmbH & Co KG, 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 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 20 μm thick, electrochemically roughened copper foil (Carl Schlenk AG, Roth, Germany). 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 convection 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). The layer porosity amounted to 50%. 15 The graphite particles had a specified average size of 32 μm. 24 The R a value of 2.5 μm was calculated according to DIN EN ISO 4287:1997 and characterizes the overall roughness. 15 All graphite composite electrodes reported here were prepared within the same batch.
Preparation of the PVDF model sample.-The PVDF model sample was prepared by the same amount of PVDF as for the graphite composite electrode, providing an 8 μm thick layer of pure PVDF ( Figure 1b).
Scanning electrochemical microscopy.-The scanning electrochemical microscope (SECM) was running under the SECMx control software developed in house. 25 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). 26 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 as probe (Figure 2b), either the graphite composite electrode or the PVDF film as sample and a platinum wire auxiliary electrode and a silver wire as reference electrode. 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 working solution. The electrolyte was 1 M LiPF 6 in Scanning electrochemical microscopy measurements.-Before approach curves and images were recorded, the ME was polarized to a potential, where the redox mediator molecule is continuously oxidized (4.1 V vs. Li/Li + for DBDMB, 4.1 V for TPA and 3.9 V for TMB, Figures 2b and 2c). The timer was set to zero, when the SECM working solution was filled into the cell. The approach curves 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 of the approach curves to the analytical approximations by Cornut and Lefrou, 27 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). During fitting, the uncertainties of 0.5 μm and 1.5 for r T and RG with respect to the CLSM results were considered (Supporting Information (SI) 2 for further information). Approach curves were fitted by adjusting κ, i T,∞ and d 0 (within reasonable range) using a least square approach. κ is the normalized first-order rate constant, i T,∞ the diffusion-limited steady-state current in the bulk solution (at quasi-infinite 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 = 2.15 × 10 -6 cm 2 s −1 for DBDMB in LP40 and D = 2.57 × 10 -6 cm 2 s −1 for DBDMB in LP47 electrolyte solution. The diffusion coefficient was determined from the steady-state diffusion limited current of the ME i T,∞ (SI-1). 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
Choice of redox mediator.-We aim for following in situ the physical swelling of graphite composite electrodes by repeatedly recording approach curves over identical locations. The choice of the redox mediator is essential for such long-term SECM experiments, because the current response of the ME depends on many factors: (i) the ability of the oxidized redox mediator species to take up electrons from the sample of interest; (ii) the stability of both oxidized and reduced species; (iii) compatibility of the redox mediator with the active material of the ME. The effect of the mediator choice for SECM feedback images is illustrated in Figure 3 by SECM feedback images recorded from an identical region before and after a mediator exchange. Initially the solution contained 5 mM DBDMB and the feedback image shows locally different ME currents i T (Figures 3a-3c). They are caused by the combination of locally different topography (roughness R a = 2.5 μm 15 ) and locally different electron transfer rates ( Figure 4, curves 1a vs. 1b, see below) due to variations in local distribution of graphite, carbon black and PVDF at the surface of the composite. Within 2.5 h the overall shape of the feedback image of the identical region is retained using DBDMB as redox mediator (Figures 3a-3c). After exchange of the working solution to 5 mM TMB + LP 40, the obtained feedback image (Figure 3d) remains similar to the previous image in Figure 3c. Thus, neglecting the small local change in topography (as will be discussed below), the local electron transfer rates are supposed to be identical. Considering the structural analogy of DBDMB and TMB (Figure 3), this is also expected since both molecules provide a similar steric hindrance and similar interactions of the functional groups with the graphite composite electrode. There is no significant change between the two sequential images using TMB (Figures 3d  and 3e). After exchange of TMB to TPA as mediator, the obtained reactivity image was less clearly resolved ( Figure 3f). The origin of the blurred appearance of the reactivity image might be the difference in steric hindrance and interactions of functional groups of TPA on one side and the components of the composite on the other side as compared to DBDMB and TMB. Consequently, TPA is less suitable for the investigation of graphite composite electrodes. Since the relative stability of TMP as a redox shuttle is unknown 17 and DBDMB has already been identified as one of the most stable redox shuttles so far, 14,28-30 DBDMB was chosen for further experiments. Indeed, continuous SECM experimentation by using DBDMB as a redox mediator without microelectrode regeneration was recently reported for 145 h 15 and has meanwhile been extended to 670 h under favorable circumstances. Approach curves on graphite composite and PVDF model film.-SECM approach curves are a recording of the ME current i T as a function of decreasing working distance d. The ME current i T is caused by the continuous oxidation of DBDMB ( Figure 2c) and depends strongly on the type of sample and d. When the ME approached the uncharged graphite composite electrode (Figure 4, curves 1a and 1b), the oxidized DBDMB + was reduced at the graphite composite electrode to DBDMB, 15 which caused an additional flux of DBDMB to the ME and thus increased i T with decreasing d. Provided that the interfacial electron transfer kinetics does not change, approach curves can be used for determining the ME-sample separation. Usually, they are plotted as normalized current I = i T /i T,∞ vs. the normalized distance L = d/r T . Curves 1a and 1b of Figure 4 indicate the variety of approach curves that can be obtained from different regions of uncharged graphite composite electrodes (i.e. before formation of a solid electrolyte interphase, see SI-2). Local differences of electron transfer rates are expected since the local composition of the graphite composite electrode varies. The apparent heterogeneous rate constants k eff corresponding to curves 1a and 1b in Figure 4 amount to 3.0 × 10 −3 and 8.6 × 10 −3 cm s −1 . The electron transfer at graphite composite electrodes is significantly slower than the calculated curve for a diffusion-controlled reduction of DBDMB + at the sample (Figure 4, curve 2). The difference of the experimental curves to the diffusioncontrolled case is mainly caused by the binder of the graphite composite electrodes, which decreases the electron transfer rate (see below).
The approach curve to the pure PVDF model sample (Figure 4, curve 3) exhibits a k eff of 5.8 × 10 −5 cm s −1 , which is smaller by a factor of 10 −2 compared to the average k eff of graphite composite electrodes. However, the approach curve on PVDF is distinct from the curve to an insulating surface (Figure 4, curve 4), where there is no back reaction at all. Consequently, DBDMB + diffused through the 8 μm thick PVDF layer and was reduced at the Cu current collector albeit with slow rate. Hence, the 8 μm thick PVDF layer is permeable to DBDMB/DBDMB + . The electron transfer rate at the PVDF model sample was within the range of the electron transfer rate constants of the SEI-covered metallic Li electrode. 16 Thus, the electron transfer rate is reduced by the PVDF binder within the graphite composite electrode.
Concept of the in situ quantification of swelling using the SECM.- Figure 5 depicts schematically the experimental steps in order to quan- After t 2 the relative thickness z 2 is again determined by a fit of an experimental approach curve. The difference δ film = z 2 − z 1 is the film thickness change.
(f) After a repetitive approach curve fit, the film thickness change is given by tify the swelling using the SECM. The absolute height of the sample electrode is defined by z 0 (Figure 5a). When electrolyte is added to the cell (Figure 5b), the timer is started and swelling commences. In Figure 5c the first approach curve is recorded. By the fit after the complete experiment (Figure 5d), the height z 1 is obtained at which the film surface is located in the laboratory coordinate system of the SECM apparatus. After a certain time the second approach curve experiment is conducted (Figure 5e) and z 2 is obtained from a curve fit. The difference between δ film = z 2 − z 1 is the change of film thickness, which represents the physical swelling as the uncharged graphite composite electrode remains at open circuit potential during the entire measurement. Further approach curves yield δ film values for different times t (Figure 5f).
The local and in situ nature of this procedure represents the main advantage of this methodology to study the physical swelling. In addition, both physical and electrochemical swelling can be investigated separately. An intrinsic disadvantage is the fact that swelling already starts with electrolyte addition (Figure 5b). Since the approach of the ME to the sample after addition of electrolyte requires time and other experimental manipulations are necessary, the time t 1 of the first thickness determination cannot be smaller than ca. 0.5 h. Thus, swelling data are not assessable for t < 0.5 h. Figure 6a shows a typical development of δ film with time after addition of LP40 electrolyte to graphite composite electrodes. The curve represents a saturation behavior. For other locations (6-9 and 11-13) on different samples of the same batch a similar saturation behavior was found (Figure 6c). Yoo et al. 5 measured the weight difference due to solvent uptake of graphite composite electrodes using PVDF in EC:ethyl methyl carbonate (EMC) solutions. They concluded that solvent diffusion is saturated after 1 day at room temperature and before one day at 50 • C. Based on Figures 6a and 6c the saturation occurs even below 12 h at room temperature for LP40 (EC:DEC). Figure 6a emphasizes that k eff was rather stable over time and the changes are significantly smaller than the estimated error bars of 30% for the determination of an individual k eff value. The uncertainty of δ film determination was estimated to be ± 1 μm (Figures 6a and 6b). A test experiment with a glass sample, where neither mediator regeneration nor swelling occur, showed that the uncertainty of height determinations is smaller than ± 0.4 μm (SI-4).

Swelling behavior of graphite composite electrodes.-
In line with the typical data set in Figure 6a, most investigations on graphite composite electrodes suggest a significant increase of δ film (Figure 6c, measurement locations 5-9, 11-13), i.e. the electrodes swell even without Li-ion intercalation. The largest swelling amounted to 15.9 μm or 19.8% of the total film thickness ( Figure  6c, measurement location 12). The swelling ratio χ is defined by χ = δ film /d film , where d film is the absolute initial film thickness. However, there were also positions with no significant change of δ film (Figure 6b, measurement locations 1-3) because the change of δ film was within the uncertainty margin of ± 1 μm.
In order to demonstrate the potential of the method to differentiate between physical and electrochemical swelling, a short lithiation pulse was applied for 120 s at a constant potential of 0 V vs. Li/Li + after 6 h (Figure 6b). A charging state of 2.9% is estimated from the measured current and the mass of the graphite material after the pulse (SI-3). The thickness change amounts to δ film = 3.8 μm. Thus, volume expansion due to lithiation can be clearly observed, although the same horizontal position did not show any significant physical swelling before. A complete investigation of volume expansion due to Li-ion intercalation requires further instrumental developments. A lithium counter electrode must be included into the open SECM cell and a method must be devised to account for the change of interfacial kinetics accompanying the formation of the SEI. 15 Since k eff did not change significantly after the lithiation pulse ( Figure  6b), a stable and pronounced SEI was not established, because in this case k eff would be orders of magnitude smaller. 15 Furthermore, small k eff values due to SEI formation after an extended lithiation pulse do not alter the results of the swelling ratios, because δ film can be determined for both large and small k eff values by the approach curve fits. Figure 7a shows the distribution of swelling ratios χ for different measurement locations in LP40 between 1.1 ± 0.5 h and 5.9 ± 0.5 h. χ varies locally between 0.3% and 17.6%. Thus, the variations of local swelling within the same LP40 electrolyte and the same batch of graphite composite electrode are very large. When the data from the last measurements at 7.2 ± 1.8 h is included in the analysis, the variation of the swell ratios is even larger (Figure 7b, 0.3% to 19.8%). Table I summarizes all averaged χ values determined by in situ approach curves. χ amounts to 11.4% when the time interval is considered between 1.1 ± 0.5 h and 5.9 ± 0.5 h (Table I). Thus, on average graphite composite electrodes swelled significantly upon electrolyte contact (Figure 6c). When the analyzed time interval is extended to 7.2 ± 1.8 h, the swelling becomes more significant (χ = 12.5%). Thus, the average physical swelling ratio of χ ≥ 11% is larger as the reported electrochemical swell rate of 10% 4 for Li-ion intercalation in graphite composite electrodes.

Statistical analysis of the swell behavior.-The histogram in
Chang et al. 31 determined ex situ and globally χ of graphite composite electrodes after 24 h contact to 1 M LiPF 6 EC:dimethyl carbonate (DMC) (1:1) electrolyte. Depending on the preparation process, χ varied between 9 and 23%. This result is consistent with our data of graphite composite electrodes (Table I) considering the difference in soaking time, preparation procedure, methodology, electrode components and composition. Zhang and Tang 11 used a micrometer to determine χ ex situ and globally for artificial graphite electrodes after electrolyte contact. χ amounted to maximal 3.0% within 6 h, whereas χ did not increase significantly with time. Thus, the curve of χ vs. t did not show a saturation behavior in contrast to the commonly observed saturation curve in Figure 6a. χ decreased almost continuously after 6 h to −3.2%, i.e. the artificial graphite electrode shrunk. This global ex situ observation of the thickness changes of artificial graphite electrodes contradicts our local in situ observations at graphite composite electrodes, where χ rather increases after 6 h (Table I)  separator and a positive electrode (Figure 2a), i.e. the graphite composite electrodes might easily expand in the direction of the electrolyte solution.
The impact of PVDF on swelling.-Since the impact of PVDF on swelling is well documented in literature, 5,9,10 the swelling of the PVDF model sample was investigated using the feedback mode of SECM. Figure 8 shows SECM images of an identical region between 1.7 h and 3.1 h after addition of LP40 electrolyte to the cell. At (x, y) = (0, 0) the ME-sample separation amounts to d = 2 μm at 1.1 h (Figure 8a). Since k eff at the PVDF model sample is relatively small ( Figure 4) and i T increases with increasing x, there is a tilt between the PVDF model sample and the x,y-plane of the positioning system. The gray-scale is identical for all three images. The dark region at the left side in Figure 8a indicates a small d and expands with t in Figures  8b and 8c. Because the ME height is fixed during the imaging, the expansion of the dark region demonstrated unequivocally swelling of the PVDF model sample. Because of the flexibility of the PVDF film toward a touch between ME and sample and its finite permeability to the mediator diffusion, there was still a remaining flux of DBDMB to the ME leading to minimum i T values of 0.6 nA in Figures 8a-8c.  Figure S4), the ME-PVDF sample separation amounts to d = 6.1 μm at x = 240 μm for t = 1.7 h in both line scans (Figures 8d and 8e). At (x/μm, y/μm) = (240, 225) d decreases by 0.7 μm at t = 2.4 h, whereas d decreases by 2.1 μm at the same time in the case of (240, 0). Consequently, the local swelling ratio between 1.7 h and 2.4 h at x = 240 μm is three-fold larger for y = 0 μm compared to y = 225 μm. An increase of δ film = 2.1 μm for the PVDF model sample within 0.7 h is relatively large compared to graphite composite electrodes considering the total PVDF layer thickness of only 8 μm. A similar tendency is observed in approach curves to the PVDF model sample, where a swelling by 7.9 μm (χ = 99%) is found between t start = 1.1 h and t end = 6.6 h (Table I). In other words, the PVDF model sample swelled by almost its own initial thickness of 8 μm. Thus, χ is ca. nine-fold larger for the PVDF model sample compared to graphite composite electrodes of the same PVDF mass loading within a comparable time window and in the same electrolyte (Table I). Since δ film = 7.9 μm of the PVDF model sample was only slightly smaller than δ film = 9.1 μm for the graphite composite electrode for a similar time window (Table I), the swelling of the graphite composite electrodes can be ascribed mainly to the swelling of PVDF binder within the composite. The heterogeneous swelling of the PVDF model sample can be explained by a locally different rate of electrolyte uptake.  Imaging of graphite composite electrodes.- Figure 9a shows an SECM image above an uncharged graphite composite electrode with 5 mM DBDMB in LP40 electrolyte at OCP. As expected, i T was heterogeneous over the entire image because of local variations in topography (see above) and local differences in the electron transfer rate (Figure 4).  (Figures 9b, S7). In the same image frame there are also locations VI (10,30) and VII (125, 0), where i T continuously increases relative to the average i T of the encircled regions (Figure 9b). A change of i T is either caused by (i) topographic changes or (ii) changes of the electron transfer rate. Since there was no evidence for a significant change of the electron transfer rate during approach curve experiments to uncharged graphite composite electrodes (Figures 6a-6c), the change of i T was mainly caused by topographic changes. Because i T increased and decreased locally at the same time relative to the encircled positions, the swelling ratio is locally heterogeneous. Thus, the PVDF model sample (Figures 8d and 8e) and the graphite composite electrode ( Figure 9) show a heterogeneous distribution of swelling ratios. The heterogeneous swelling ratio of graphite composite electrodes is explained by the locally heterogeneous swelling ratio of the PVDF binder.
As stated elsewhere, 16 i T at the ME decreases continuously in LP40 electrolyte solution because of impurities which cover the Pt surface of the ME probe. Since each image is build up of consecutively measured locations and the relative time difference between the measured locations within one image frame remains constant, the decrease of i T due to impurities does not affect the analysis of relative current changes within one image frame of the sequence.  (Table I) over the entire surface due to electrolyte-binder interactions (Figure 8) as demonstrated by height data extracted from approach curves ( Figure 6). In addition, the swelling ratio of graphite composite electrodes is locally heterogeneous (Figure 9) because the swelling ratio of the pure binder is locally heterogeneous (Figures  8d and 8e). Thus, the topography of the graphite composite electrode changes locally after swelling (compare z 1 and z 2 in Figures 10a  and 10b).

Conclusions
2,5-di-tert-butyl-1,4,-dimethoxybenzene (DBDMB) has been used successfully as redox mediator to study the electron transport at uncharged graphite composite electrodes whereas tri(pbromophenyl)amine (TPA) is not effective. The electron transfer rate at uncharged graphite composite electrodes is heterogeneously distributed as demonstrated by approach curves at different locations. At uncharged graphite composite electrodes k eff varies between 3.0 × 10 −3 and 8.6 × 10 −3 cm s −1 and is therefore significantly larger than at charged graphite composite and metallic lithium electrodes that form a passivating solid electrolyte interphase. 16 By approach curves to a PVDF model sample it is shown that DBDMB diffuses through a 8 μm thick PVDF layer and reacts at the Cu current collector, although the electron transfer rate constant was rather low (k eff = 5.3 × 10 −5 cm s −1 ).
Height determinations from consecutive approach curves at the same location are suitable for local in situ quantification of the physical swelling of uncharged graphite composite electrodes. The average δ film amounts to 9.1 μm for a 80 μm thick graphite composite electrode in LP40 electrolyte between 1.1 ± 0.5 h and 5.9 ± 0.5 h after adding the electrolyte. The uncertainty of the film thickness change δ film determination is ± 1 μm. Thus, uncharged graphite composite electrodes swelled on average by χ = 11.4%. This value varies from 0.3% to 17.6% at different locations for graphite composite electrodes from the same batch in the same electrolyte. Consequently, a few locations do not swell significantly. In contrast, the 8 μm thick PVDF model sample swells by 7.9 μm (χ = 99%), i.e. by its own initial thickness. Because δ film of the PVDF model sample is almost as large as δ film of the much thicker graphite composite electrodes, the physical swelling of the composite electrode is mainly due to the swelling of the PVDF binder.
Both PVDF model sample and uncharged graphite composite electrodes show locally different swell ratios by SECM imaging. Based on these results a swelling model is proposed, where the uncharged graphite composite electrode swells physically on average by at least χ = 11% and the local topography is changing during swelling due to locally heterogeneous swelling ratios.
The embedding of electrochemical swelling into our swelling model is a topic of ongoing research. In addition, the swelling of composite electrodes with different compositions will be addressed as well as the impact of the electrolyte composition on physical swelling. Moreover, the swelling of different binder systems as for instance carboxymethylcellulose (CMC)/styrene-butadiene rubber (SBR), which is gaining increased significance for graphite composite electrodes, can be quantified. However, the swelling of CMC/SBR in carbonate electrolytes is at least 20 times smaller compared to PVDF. 33,34 Thus, the swelling ratios of CMC/SBR in organic electrolytes might be below the detection limit of the presented methodology.