Detection of a Cobalt-Containing Interphase at the Li <sub>6</sub> PS <sub>5</sub> Cl-NMC111 Interface by In Situ μXANES and EIS

Results

Electrochemical cycling.-Cycling of ASLBs is shown in Fig. 2, comparing cells prepared with bare NMC and coated NMC-LLSTO. The cells began with a charge step, followed by discharge, with voltage curves shown in Fig. 2a. Five cycles were performed at C/ 10, with later cycles at C/3. The specific capacities are shown in Fig. 2b.

With a cathode interfacial coating, the full NMC capacity was accessed during the first charge (172 mAh g -1 ). During the first discharge, the was a loss in capacity caused by irreversible reactions (121 mAh g -1 ). With uncoated NMC, the first charge resulted in significantly lower capacity (132 mAh g -1 ) than theoretically available in the NMC (∼160 mAh g -1 ). After the first discharge, capacity was even lower (76 mAh g -1 ).

A loss in capacity can be caused by (i) conversion of active material to an electrochemically irreversible phase, or (ii) introduction of an internal resistance large enough to push the electrochemical activity of active material into a potential range outside of the battery operational window. An example of the latter case is formation of a resistive blocking layer that isolates available active material from the rest of the cell, rendering it unavailable.

It was clear from the above cycling results that the uncoated NMC suffered from a capacity loss on first charge that was not seen with a cathode coating. This loss during the first charge could be due to electrochemical events that happen under applied current, or chemical events that occur at rest, before cycling begins. Previous reports have shown evidence of S and P containing breakdown products. 10 However, to our knowledge evidence of alteration of the NMC has been less discussed. If species originating in the NMC are converted to other materials, this would mean capacity loss mechanism (i) occurs to some extent, and the capacity losses are not only due to resistive layers adding overpotentials to the cell voltage.

In situ μXANES of the Li 6 PS 5 Cl-NMC111 interface.-Presumably, redox reactions occur at the interface where LPSC is in contact with the cathode active material. If LPSC is oxidized, the reduced compounds originate in the NMC. To specifically probe the LPSC-NMC interface, we employed a spatially resolved X-ray spectroscopic technique. The NMC particles (Fig. 3a) were roughly spherical with an average diameter of 9.3 μm. However, there was distribution of particle sizes, with examples in Fig. 3b ranging across a factor of 7, from 2.9 μm to 20.5 μm.

XRF mapping allowed composite materials to be imaged. Figure 4a shows a composite cathode with 10% NMC by mass and Fig. 4b shows a composite cathode with 70% NMC by mass. Areas with a low Co signal were occupied by either LPSC or void. The 70% composition was the optimal for battery performance, for example in the data shown in Fig. 2. Given the high-flux onto the sample and the high fluorescence yield from the particles, the 70% CAM sample was difficult to analyze and find isolated particles for measurements.

To resolve isolated NMC particles, a 10% CAM case provided ample distance between particles. The brightness of the NMC particles indicated their relative proximity to the PEEK window. This means the particle at coordinates (65,65) in Fig. 4a was closest to the PEEK window.

Figure 5 shows the 10% NMC composite cathode with the individual elemental XRF mapping of Ni, Mn, Co, and S. The colocation of Ni, Mn, and Co marked the NMC particles. The S signal was anti-correlated to these, as S was found in the LPSC and was present in the regions left vacant by NMC. Fluorescence response was in proportion to element Z values: Ni > Co > Mn ? S. From this series of scans, a particle of interest was chosen, marked in each panel by a white dashed box. This particle had other NMC particles in its "west" and "southwest" directions. However, to the north and east, only LPSC was in its vicinity.

This particle of interest was analyzed using μXANES collected across the NMC particle interface. Figure 6a shows a detail of the XRF map with 21 μXANES positions indicated by gray dots placed in the middles of the pixels. By collecting μXANES at a grid of positions which spanned the particle interface we scanned for changes between the particle bulk and the interphase region formed at the interface where LPSC and uncoated NMC were in contact. Two such scans were performed in the "north" direction and one in the "east" direction. These were assigned a position value based on radius (P1-P8).

In Fig. 6 panels (b-e) these positions are given different colors. The XRF map in Fig. 6a was taken with the Si filter removed, causing a "halo" effect due to signal saturation in the bulk of the particle. At positions beyond those marked by dots in Fig. 6a, the Co and Ni K-edge μXANES signals vanished, meaning these positions were beyond the particle interface. The Mn K-edge μXANES data had a higher signal-to-noise ratio than that of Co and Ni, and consequently small variations in the μXANES could not be observed with confidence.

μXANES scans were used to examine the change in composition in the interphase region in contact with LPSC, as compared to the NMC bulk. The interface of the LPSC and NMC was measured at every position because these particles were in a 3-dimensional composite cathode. However, the interphase signal was drowned out by the bulk particle data at the positions closer to the center of the particle. The normalized absorption coefficients at each position are plotted for the K-edges of Co (Fig. 6b) and Ni (Fig. 6d). Black traces were the center of the particle and light blue was the location furthest from the center of the particle, which was largely composed of interphase.

Co metal has a K-edge of 7708.9 eV. In Fig. 6b the absorption edges were higher, indicating Co was in a more oxidized state. In NMC111 Ni, Mn, and Co have the nominal oxidation states of 2+, 4+, and 3+. 23 In Ni-rich NMCs some of the Ni is oxidized to maintain electroneutrality. Surface analyses of uncycled NMCs have been reported with a Co oxidation state of 3+ and also a mixture of 3+/4+. [24][25][26] In the derivative plot of the Co μXANES (Fig. 6c) the edge energy is indicated by the large peak before the zero-crossing point, which occurs at 7727.2 eV. Reports of Co XANES in the literature indicate that a shift to lower photon energy by about 4 eV is expected when comparing a Co 2+ compound to an analogous Co 3+ compound. 27,28 In the positions closer to the interface of the particle (P7, P8) a smaller peak was observed at 7722.7 eV (marked by an orange arrow), indicating the formation of a Co 2+ species. Spatiallyresolved intensity of the 7722.7 eV peak is shown in Fig. 6f, demonstrating a significant increase at the particle interface. The continued existence of the original peak at a higher energy indicated that some amount of Co 3+ remained and was expected as XANES is an additive technique. For Ni there was no change in oxidation state between the interphase and bulk (Fig. 6e).

The reduction of Co suggests that NMC undergoes a redox reaction with LPSC electrolyte, as oxidation to Co 4+ would be expected upon self-discharge. No change was observed in the oxidation state of Ni (Fig. 6e). The decoupling of Co and Ni oxidation states further indicates the breakdown of the NMC material at the interface. The presence of Co 2+ suggested the formation of a cobalt sulfide such as CoS or Co 3 S 4 .

Jung et al. have previously reported XPS and SIMS analyses of both aged and cycled pellet cathodes. 10 In aged cathodes that were not cycled, they found evidence of decomposition to sulfides, raising the possibility of NiS, CoS, MnS, or Li 2 S if the components of the NMC participated in sulfide formation. They also found evidence of NiS in the XPS spectra. 29 However, involvement of the other metals was not specifically indicated. By using a microscopic technique to probe redox states at the LPSC-NMC interface, the results above demonstrated a measurable reduction of Co 3+ to Co 2+ at the NMC surface. This verified that Co participates in a redox reaction during chemical degradation. With an uncoated NMC, this confirms that active material was converted to other species, and that capacity loss was at least partially caused by active material loss.

Piecewise in situ EIS during cycling.-The evolution of electrochemical kinetics was studied by EIS. The piecewise EIS for cell charging shown in Fig. 7c (coated ASLB), Fig. 8a (uncoated ASLB), and Fig. 8b (liquid-electrolyte cell) were performed in the manner of a GITT-type experiment, with a galvanostatic period followed by a rest. The GITT cutoff voltage was 4 V, although there was relaxation during the rest after this. On the EIS plots, voltage after relation was the value reported.

Figures 7a-7b shows the piecewise cycling data of a full cell made with coated NMC-LLSTO. The cell underwent 8 charge steps at C/10 before the upper voltage limit was reached, followed by six discharge steps. Piecewise Nyquist plots for the charge are shown in Fig. 7c. For all piecewise steps the Nyquist plots had the appearance of a single semicircle with a low frequency tail characteristic of mass transport. At 1 MHz there was a high frequency resistance (HFR) of 38 Ω, which was mostly invariant during cycling.

The transition from semicircle to tail was at about 1 Hz, so the real impedance value (Z) at 1 Hz could be used as a characteristic impedance for the limiting process at each stage after the HFR was subtracted, essentially the width of the Nyquist plot semicircle. After the first hour of charge the characteristic impedance was 562 Ω. This value decreased with each step until 3.68 V when it was 214 Ω. Charge steps 7 and 8 showed an increase to a final value of 286 Ω. During discharge (Fig. 7d) the impedance increased to 938 Ω. The increase was particularly steep below 3.44 V. Evolution of the Nyquist plots did not follow a reversible path during the first charge-discharge cycle. This was expected because the cycling data showed a large capacity loss between the first charge and first discharge. The impedance increase began in the final steps of charge and continued through discharge. This indicated the processes causing capacity loss occurred during this period, and primarily at low voltages at the end of discharge, when the biggest impedance increase happened.

The first charge of identical cell with uncoated NMC is shown in Fig. 8a. HFR was 41 Ω. Characteristic impedances were much larger than the cell with NMC-LLSTO. The initial Nyquist plot showed an impedance of 1794 Ω, 3.1× higher than the cell with coating. The trend during charge was however similar, as impedance reached a minimum from 3.52 V to 3.67 V of about 1110 Ω. This was 5× higher than the minimum in the coated cell. At the end of charge the impedance increased again to 1590 Ω, 5.4× higher than the coated cell.

The uncoated cell reached a lower open circuit voltage than the coated cell, only 3.77 V compared to 3.92 V, although they both had a 4 V cutoff. This was because the higher overall impedance caused a larger iR contribution, therefore triggering the cutoff voltage earlier. While the overall trend was the same, the increasing impedance disparity with the coated cell (3.1× to 5× to 5.4×) revealed that the decomposition reactions triggered by the electrochemical reaction caused a greater impedance rise in the uncoated cell.

Figure 8b shows the same experiment with a Li-ion cell of similar loading with liquid electrolyte. The first Li-ion Nyquist plot had two semicircles. The semicircle at higher frequency is typically attributed to Li + migration through the SEI, and the one at lower frequency is the charge transfer resistance of the NMC electrochemical reaction. By 3.75 V these two distinct processes had merged into a single semicircle. The characteristic impedance decreased steadily during charge. At 3.75 V it was 267 Ω and at the end of charge it was 202 Ω. These were lower than the coated ASLB cell, but only by a relatively small amount.

This impedance analysis demonstrated the difference between chemical and electrochemical degradation at the LPSC-NMC interface. The chemical degradation resulted in an interphase characterized by μXANES, and this interphase caused the uncoated cell to have a far higher initial impedance than the corresponding coated cell. While the impedances seen during charge of the coated ASLB did not different greatly from a liquid electrolyte cell, the uncoated cell displayed much higher impedance. Electrochemical degradation was evident in the coated cell, which developed high impedance during the discharge, not returning to its pre-charged state. The coated and uncoated cells displayed the same trend during the initial charge, but the impedance of the uncoated cell increased to a greater extent. This suggested that coating also mitigates electrochemical decomposition.

Transmission line model of the cell with uncoated NMC.-The electrochemically active surface area in ASLBs is likely lower than in liquid electrolyte batteries because solid electrolyte does not flow and wet interfaces like liquid electrolyte. Instead, point contacts between the solid electrolyte and active material particles are the locations of electrochemical activity. 17,30 Evidence of an interphase region containing reduced Co 2+ was observed above. From the EIS, a high initial impedance was observed for uncoated ASLBs. This interphase layer presumably also can occlude the electrochemically active surface area. This is where the electrochemical reaction can occur, given in Eq. 1.

x x Li Ni Mn Co O Li Ni Mn Co O Li e 1 x 1 13 13 13 2 1 13 13 13 2 ⇌ + + [ ] / / / -/ / / + -

Microstructural information about a composite cathode can in principle be extracted from EIS data, but this requires application of a model. It is desired that the model should be physically meaningful and have few enough parameters that can reasonably be obtained by data fitting.

Single-particle or single-interface models can apply when current is uniform through the thickness of the composite cathode, i.e. when all surface area experiences the same extent of reaction. We have previously reported spatially-resolved non-uniformity in the electrochemical reaction during ASLB cycling in 120 μm thick cathodes. 17 In such a thick cathode with 70% CAM, the maximum gradient in 1-x was measured to be 29% of the total cyclable cathode capacity.

Figure 9 shows operando EDXRD data on a thinner, 10 mg cathode, the same size as used in the EIS work above. Charge and discharge in Fig. 9a were correlated to the localized NMC a and c lattice parameters in Fig. 9b. These are reported as a function of depth in the cathode, with the distance from the LPSC separator represented by different colors. Using literature data for NMC111, the localized values of 1-x were calculated and are shown in Fig. 9c. 31 As expected, non-uniformity was far less than in thicker cathodes. However, during the maximum non-uniformity, around 3.5 h or the midpoint of charge 1, the gradient in 1-x was 0.05 or 10% of the cyclable capacity.

The decomposed interphase around NMC particles could not be observed by EDXRD. Because the sampled gauge volume contained many particles, any signal from the interphase region was extremely small, being dominated by the bulk signal. We previously demonstrated structural decomposition of a sulfide solid electrolyte using EDXRD. 32 This large-scale decomposition occurred when carbon was present in the cathode. However, with carbon excluded it was not observed.

A TLM is represented by a ladder-like circuit such as Fig. 10a, in which there are two parallel rails: one for ionic conduction and one for electronic conduction. 33 The electrochemical reaction in Eq. [1] demonstrates that both ionic (Li + ) and electronic (e -) conduction are required for any particular reaction site on the NMC to participate in cycling. In Fig. 10a, one example reaction site is marked by a red dot. During charge, Li + and e -transport away from this deintercalation site, as shown by dashed lines. These transport processes accomplish current flow across the composite cathode and complete the cell circuit. In this way the two rails of the TLM are connected by an interfacial impedance Z int (Fig. 10b). The parallel RC circuit in Z int accounts for the electrochemical reaction kinetics (R 1 ) and the interfacial capacitance (C 1 ) and has an impedance of Z RC = R 1 /(1+jωC 1 R 1 ), where j is the imaginary number and ω is frequency. Solid-state diffusion of Li + in the NMC crystallites is represented by a constant phase element (CPE), with Z CPE = Q 2 /(jω) a2 . For semiinfinite planar diffusion a Warburg element would be used with a 2 = 0.5, but in this system the more general CPE allowed for deviations from ideal geometry. This is a modification of the Randles circuit used to describe electrochemical reactions. 34,35 Figure 10c shows the electronic end element Z q , accounting for the contact resistance between the cathode and the metal current collector. Use of a CPE here allowed the model to also account for any capacitive behavior at the current collector. However, the fit results showed a q was nearly zero, making this almost a pure resistor (see below).

In a TLM, there are an integral number of possible reaction sites, shown as n Z int elements. This accounts for the fact that the electrochemical reaction does not have to occur uniformly throughout the electrode but can be preferentially distributed through the thickness. This phenomenon is often seen in composite electrodes, due to the coupling of transport and kinetics shown in Fig. 10a. 36 An early example of a TLM was provided by De Levie in 1967. 37 In the De Levie model, current enters and exits the circuit on different rails, as in the cathode model shown in Fig. 10a. Later, Siroma et al. provided several equations for general TLMs, one of which is the De Levie model. 38 In cases where the reaction is uniform, a single-interface equivalent circuit model can adequately model the system, and this has frequently been used in the study of ASLBs. 5,[39][40][41] However, in our system the EDXRD results above demonstrated that the reaction was non-uniform, requiring a TLM approach.

Equivalent circuit EIS models are meant to characterize multiple phenomena that occur on different time scales. This was important for our system, which we previously found to be transport-limited, making kinetics difficult to characterize using EDXRD data alone. 17 However, they are usually more simplistic than physics-based models, sacrificing some accuracy for ease of interpretation. For example, current flow in liquid electrolyte is not solely a conduction phenomenon. Geng et al. have combined a TLM framework with physically-meaningful descriptions of transport and species concentration. 42 Moškon et al. have developed powerful TLM versions that can account for realistic transference numbers in the liquid electrolyte of Li-ion batteries. 43,44 However, ASLBs based on single-ion conductors like sulfides are particularly well-suited for analysis with a TLM, because ionic current is solely carried by Li + . In this way, an ASLB can be more accurately modeled with a TLM than a liquid electrolyte cell. Recently TLMs have been reported in analysis of ASLBs, to quantify both conduction and kinetic phenomena. 10,13,45 TLMs can be used to quantify details about electrode microstructure. Costard and co-workers used a TLM to determine the kinetic parameters for charge transfer between NMC active material and liquid electrolyte. 46 For several NMC stoichiometries, they calculated values of the charge transfer resistance that were independent of the electrode microstructure, normalized to the real electrochemically active surface area A act between NMC particles and the electrolyte. 47 This allowed calculation of true exchange current density i 0 values. Their approach was to determine A act independently via FIB-SEM tomography. The charge transfer resistance R CT obtained from fitting EIS data to a TLM could then be normalized with this true surface area. The four electrodes they studied had quite different microstructural properties, but their calculated i 0 parameters fell in a compact range from 2.08-4 A/m 2 . This is as opposed to the rather large range of i 0 parameters reported across the literature from various methods.

Our approach in this work was to adapt the TLM technique reported by Costard et al., but to determine the real active surface area A act . To do this, we assumed the microstructure-independent i 0 matches the average values found from literature. By determining R CT via a TLM, A act could thus be calculated. This approach addresses the difficulty of accurately determining the microstructural parameters, e.g. A act , for the ASLBs used in this study. Sulfide ASLBs are typically studied under heavy compression, and it has been shown that performance is a function of the stack pressure. 48,49 In the composite cathode, high cycling pressure is likely to increase A act , pressing sulfide particles into more intimate contact with the active material. Contact area between the sulfide particles themselves and between active material particles will also increase the cross sections of conduction pathways, improving effective conductivities. Tomographic FIB-SEM would be challenging to accomplish under compression. Nanoprobe X-ray tomography has demonstrated that contact area improves under compression. 49 However, if the area is blocked by an interphase formed through decomposition, this may be mere nm in thickness and difficult to observe via tomography. We hypothesize that an estimate of A act can however be extracted from EIS through the application of a TLM in this work.

Impedance of the composite cathode TLM in Fig. 10a Z cc was mathematically given by Eq. 2.

This is the Z-type open-Q arrangement as defined by Siroma et al. 38 This impedance was placed in series with a resistance for the solid-state separator layer, R SSE , as well as an RQ circuit that modeled the In-Li anode foil. The values for the separator and anode parameters were determined from symmetric cell experiments and were held constant during EIS data fitting for the full cells. Fitting was done for the data in Fig. 8a, using a non-linear least squares regression in Python.

Before using the TLM fits to quantify the microstructure and calculate A act , we first normalized to an electrode length scale, to allow comparison to literature results normalized in a similar way. This confirmed the fitted values were reasonable. The raw fit parameters were normalized using the cell superficial area 1.247 cm 2 and are reported in Table I. The fits for the seventh EIS step (112 mAh g -1 ) showed relatively anomalous values that disagreed with the trends in the rest of the charge. This indicated fitting of this step was prone to error, although it is included in Table I for completeness.

Charge transfer resistance r CT was equal to r 1 from the interfacial impedance. The trend in charge transfer resistance r CT described changes in the electrochemical reaction during battery charging. In liquid electrolyte cathodes, r CT is reported to initially decrease during the charge. 50 In contrast, the ASLB r CT started at 25 Ω cm 2 (at 16 mAh g -1 ) and increased to 117 Ω cm 2 (32 mAh/g) and 169 Ω cm 2 (48 mAh g -1 ). Beginning at 64 mAh g -1 it decreased to 112 mAh g -1 . In the final charge step, there was another increase, to a final value of 98.9 Ω cm 2 .

To assess the validity of the model fits, these values were compared to reported r CT values from the literature. Charbonneau, et al. measured r CT for NMC111 in liquid electrolyte. 50 They found r CT decreased until 100 mAh g -1 with minimum values of 1.92 Ω cm 2 . Moškon, et al. used EIS and a TLM to model NMC811-Li cells using varying concentrations of LiPF 6 liquid electrolyte. 43 They reported values for a 1 M electrolyte adjusted for their electrode area (2 cm 2 ) giving an r CT of 32 Ω cm 2 for pristine NMC811. Costard, et al. used EIS, TLM, and Butler-Volmer kinetics to calculate parameters for NMC111, NMC622, and blended cathodes. 46 They found r CT of 63.97 Ω cm 2 .

While there is a significant variation in r CT values reported in the literature, the values in our study were within reasonable ranges. This provided confidence that the model delivered physically meaningful results. Double layer capacitance c dl was equal to c 1 . We also compared c dl to literature values, which are generally in the mF/cm 2 to μF/cm 2 range. Moškon, et al. found a c dl of 5 mF cm -2 for pristine NMC811. 43 Charbonneau, et al. measured a value of 1.1 mF cm -2 .

Calculation of electrochemically active surface area A act .-In this section we used the raw fit parameters to calculate A act , using a method similar to that in Costard et al., detailed above. In the literature, values from equivalent circuit models are often normalized by the electrode superficial area. The values in Table I were normalized this way, to compare directly with literature values. However, a more physically meaningful normalization can be done using the electrochemically active surface area in the cathode, A act . For example, R CT can be converted to an area normalized r CT using Eq. 3.

This provides a way to estimate A act from EIS data. Because solid electrolyte does not flow to wet the active material in the manner of liquid electrolyte, it is believed that A act will be lower than in conventional liquid electrolyte cells. This can be due to voids in the composite cathode causing reduced contact between SSE and NMC. Other phenomena will also result in reduction of A act , for example growth of a resistive interphase or physical disconnection of the NMC caused by volume change. 5 Active surface area is often reported as volume specific active surface area a , act given by Eq. 4.

a A Volume of electrode 4 act act = [ ]

A theoretical calculation for a act in a porous electrode can be derived based on the active material diameter and the volume fractions of the active material and electrolyte phases. 51 Based on a 10 μm NMC particle diameter and the densities of NMC and LPSC, a theoretical a act value of 2216 cm -1 is expected for the ASLB system. In the section that follows, we use an alternative calculation based on R CT to estimate the real active surface area in the ASLB.

The Butler-Volmer equation describes charge transfer kinetics at the interface between cathode active material and SSE and is displayed in Eq. 5. Here n is the number of electrons transferred, i 0 is the exchange current density, η is the surface overpotential, and α is typically 0.5. Based on a literature review in Table II of Table I. Superficial area-normalized parameters from TLM data fitting of the uncoated ASLB EIS results. r 1 c 1 q 2 a 2 q q a q Charged capacity mAh/g [Ω cm 2 ] [mF/cm 2 ] [ Ω cm 2 s -a2 ] [ Ω cm 2 s -ap ] 16 25.2 0.363 0.028 0.016 44500 0.053 32 117 0.050 5.13 0.197 84700 0.034 48 169 0.030 16.8 0.216 118000 0.024 64 104 0.045 9.83 0.214 136000 0.024 80 86.7 0.053 9.73 0.221 151700 0.021 96 66.4 0.071 7.82 0.235 154300 0.018 112 28.1 0.161 4.31 0.240 192800 0.012 128 98.9 0.046 15.4 0.255 194700 0.009 exchange current density of NMC111, an average i 0 value of 2.3 × 10 -4 A cm -2 was used. 46 i i nF RT nF RT exp 1 exp 5 CT 0

For small currents like those applied during EIS, a linear approximation of the BV equation can be used, as in Eq. 6.

The area specific charge transfer resistance r CT is equal to overpotential divided by charge transfer current as seen in Eq. 7.

Equation 7 can therefore be used to solve for the active surface area in the system A act given known values for R CT and i . 0 This is given in Eq. 8.

Table III displays the calculated values of A act and a act based on the non-normalized values of R CT obtained from the EIS fitting.

A theoretical a act value of 2216 cm -1 was calculated for the ASLB system. The values of a act in Table III were all lower than this value, indicating the inefficient packing of solid particles in an ASLB results in lower active surface area. Interphase formation and disconnection due to volume change are also expected to impact this value. Early in the charge (16 mAh g -1 ) the a act value was calculated to be 49% of the theoretical value. The tomography data of Sakka et al. showed that with similar materials and the same compression (50 MPa) the contact area fraction between NMC and sulfide was 84%. 49 The lower value estimated from the TLM could be due to the decomposition interphase blocking the connection even in cases where particles are touching. Because the tomography voxel was 500 nm, the interphase and small disconnections might not be discernible. This demonstrates that chemical decomposition results in reduced charge transfer between the active material and electrolyte.

The sudden drop in a act to 234 cm -1 following the first segment is the most prominent observation upon continued charging. Because of uncertainty inherent in the EIS fitting and the selection of i 0 , the variation after segment 2 is possibly due to experimental variations. To compare active surface area more generally, we use the median value of 296 cm -1 . This is 7.5× lower than the theoretical value of 2216 cm -1 , or 13%. This is far lower than the 84% value calculated by tomography. Due to the voxel size, tomography may only be able to observe loss of contact area when it is caused by pore space. The extra loss could be due to physical disconnection or additional electrochemical interphase formation. Our finding demonstrates that contact areas may be lower than expected in ASLBs, especially when using uncoated active material that undergoes decomposition and interphase formation.

The use of a TLM and piecewise EIS has been shown as a possible tool to probe microstructural characteristics of composite ASLB electrodes, especially in cases where active surface may be blocked by extremely small interphase layers that are difficult to observe in situ. Future efforts to produce composite electrodes with well-controlled microstructures could be used to tie microstructure characteristics to TLM parameters, which can be useful for more extensive cell characterization.