2.2.3. Erosion/Dilation

Eroding removes the edge/outer shell of a particle [42]. Mainly 'noise dots' (single voxels), but also very small particles that result from the recording process—for example, intersecting streak artefacts—and that do not belong to particles, disappear completely. Subsequent dilatation then adds the missing edges to the particles, but not to the now missing noise dots. In this way, the particles largely regain their previous volume, but the noise dots are removed. However, previously sharp particle edges tend to be smoothened.

#### 2.2.4. Watershed Transformation

To carry out a watershed transformation, a function is initially used to assign a distance from the particle edge to each voxel of a particle. If one now interprets the equidistance lines as water level lines after fictitious flooding with water, the volume is divided into several pools. If one fills them up with virtual water, one first obtains several smaller lakes that are merged to a larger lake at a watershed. Along this watershed, the particle is divided into two [42].

#### 2.2.5. Location Retrieval

In order to follow the dissolution process of individual particles, it was necessary to retrieve the location of the particles after each discharging or charging step. Therefore, a distinctive particle was chosen for each battery in the corresponding data set. After each cycle, this particle was located to ensure that the same areas were examined in the batteries. These particles were selected from areas neither directly next to the nail, nor to the separator, as it has been suggested that in these regions that a typical particle disintegration may occur.

In the image corresponding to the RAM discharged at 100 mA current for 10 cycles, it was no longer possible to find the initially selected particle. However, by comparing a number of other areas in this data set (around the initial particles) the original location of the particle was successfully reconstructed.

#### **3. Experimental Results and Discussion**

#### *3.1. Properties of the Cells at Different Discharge Rates and Numbers of Cycles*

Figure 2a shows measured capacities of the individual cells per discharge/charge cycle. The final discharge voltage was 0.9 V. According to the manufacturer, the RAM should deliver up to 800 mAh, while the primary cell should deliver up to 1220 mAh at a discharge current of 30 mA.

**Figure 2.** (**a**) Battery capacities remaining after a given number of discharge cycles, (**b**) discharge curves of pristine rechargeable alkaline-manganese (RAM) batteries at different discharge currents, (**c**) total volume of all segmented zinc particles and (**d**) discharge curves of a RAM battery (100 mA) during different cycles.

It can be seen in Figure 2b that a higher discharge current results in a faster drop of voltage, which is associated with a lower discharge capacity after the cycle, and an increased internal resistance. The voltage also decreases with an increasing amount of cycles (Figure 2d). Of course discharging with lower currents also takes longer, as shown in the discharge curves in Figure 2b for the three different discharge currents applied in our work, namely 100 mA, 200 mA, and 400 mA.

Additionally a difference in the dissolving behavior of the zinc particles can be seen in Figure 2c. While the zinc particles in the RAM cells dissolve continuously during cycling, the zinc particles in the primary cell stopped dissolving after the third cycle.

#### *3.2. 3D Structural Analysis*

Figure 3 shows tomographic cross sections through samples in different charging stages. The cross sections of the fully charged and partially discharged batteries were always taken at the same locations (Section 2.2.5). After a first discharge, many of the smaller zinc particles were dissolved in the electrolyte (compare Figure 3, first and second column). Dissolution continuously progresses until most of the zinc particles have been dissolved into small particles that form a homogeneous gel in the electrolyte (i.e., particles are smaller than the spatial resolution of the tomography setup).

**Figure 3.** Tomographic cross sections showing three different cycle states for each of the four batteries studied. (**a**) RAM battery discharged at a current of 100 mA, (**b**) 200 mA, (**c**) 400 mA, and (**d**) alkaline manganese primary cell discharged at a current of 200 mA.

However, in the images describing the RAM cells (Figure 3a–c), larger particles, especially at the outer ring area, still remained even after 10 cycles (see also Figure 4). Their size decreased, but their inner core remained almost unchanged. This reveals a layer-by-layer dissolving behavior (see also Figure 4). On the (outer) cathode side, the cylinder comprising the MnO2 and electrolyte slowly swelled and moves inwards, while at the same time cracks formed. Especially in the primary cell, the deformation resulted in the development of pointy structures in the separator region between the electrodes. The gap between both electrodes that was filled by the separator became smaller.

**Figure 4.** Overview over four selected particles in (**a**) RAM battery discharged with 100 mA, (**b**) 200 mA, (**c**) 400 mA, and (**d**) primary alkaline-manganese cell discharged with 200 mA.

#### *3.3. Dissolution of Individual Particles During Cycling*

Cross sections through individual large particles occurring in each of the four tomographic measurement series showed the shapes of the particles after several cycling steps (Figure 4). Due to their size and good recognizability, even after several cycles, these particles were also used as "markers" for the location adjustments made when recording the data for Figure 3, i.e., they were used to find the same locations in the cell after different cycling steps. Obviously, the particles maintained their shapes over several cycling steps. However, the particle size slowly decreased and an area around the particles occurred that appeared fuzzy and in intermediate (grey) contrast, possibly containing small zinc particles that were no longer resolved by our measurement technique.

#### *3.4. Quantitative 3D Analysis*

For a quantitative analysis of the zinc particles in the batteries, the air surrounding the battery, the steel casing, and the manganese dioxide cathode were "cut off", as shown in Figure 1c, so that the respective image data contained only the anode material and the nail in the center. To this data set, the threshold filter was applied, after which the particles were white, and particle-free areas were left black. The same threshold was applied to all batteries and for all cycles. To reduce the noise in both the white and black areas, a median-3D filter was applied to the data sets (Section 2.2.3). Using the distinctive particles shown in Figure 4 as a starting point, the locations of 600 layers in each battery tomogram were adjusted to the same position, and the resulting data sets were quantitatively analyzed. This procedure ensures that exactly the same volume range for all cycling states of the battery were analyzed and any possible drift was eliminated After this, a watershed transformation was applied to improve particle separation (Section 2.2.4) using the program 'Avizo 7.0'. For the radius-dependent analysis, each particle coordinate was converted from Cartesian to cylindrical coordinates, with the nail as a center. It was taken into account that the nails stuck askew in the batteries. The graphs in Figure 5 present the results of this particle size analysis. The particle diameter represents the corresponding spherical diameter. Figure 5a,c show the results for the RAM cell discharged at either 100 mA or 400 mA, while Figure 5b,d show the corresponding results for the RAM cell and the primary cell after discharging at 200 mA. Typical particle diameters range from 100 μm to 200 μm for both battery cell types. In all four cases, a rapid drop in the overall particles volume was found after the first discharge/charge cycle. This indicates that large zinc amounts were dissolved in the electrolyte, and that many particles reached a size where they could no longer be resolved by tomography and therefore did not contribute to the overall volume in the graphs shown. After each cycle, the volume found in the analyzed (larger) particles decreases in accordance to the findings in Figure 3. The size distributions of the particles remained almost unchanged.

**Figure 5.** Particle size distributions in RAMs discharged at (**a**) 100 mA, (**b**) 200 mA, and (**c**) 400 mA and (**d**) primary cell discharged at 200 mA.

One might expect that all particles shrink continuously. However, Figure 3 demonstrates that many particles dissolved quickly and "disappeared", the smaller particles did so even faster than the larger ones. On the other hand, many larger particles became smaller and/or broke up, and they contributed to the amount of smaller particles while at the same time, several larger particles seemed to be largely unaffected by the cycling process. Eventually, the remaining particles seemed to have a size distribution that was very similar to the initial one, although many individual particles changed their size.

Figures 4 and 5 reveal that particles dissolved more rapidly at lower than at higher discharge currents. Especially, the first discharge had a large effect. From the second discharge, the dissolution process was much slower and decreased continuously with increasing cycle numbers. During the discharge, the edges of large particles, dissolved and a shell with a lower absorption coefficient was formed around the particle. During charging, these clouds became smaller again. Very small particles were completely dissolved. This process was more noticeable near the separator.

Furthermore, during charging some very small new particles are formed, indicating the development of seeds for the growth of zinc crystals (see Supplementary Video 1).

## *3.5. Local Effects*

The changes in particle sizes were not uniformly distributed over the entire cell. With progressing cycling, increased particle migration was observed, the distance and direction of which depended on the distance of the individual particle to the current collector nail. The particles in the RAM cell appeared to drift away from the collector nail towards the separator. This was observed only in the fifth cycle or later.

In the primary cell, in contrast, it seemed that starting from the first cycle, particles close to the nail migrated towards the center, while particles located close to the separator moved outwards, thus forming a ring with a lower particle density. The particle analyses of the pristine RAM cell and of the RAM cell after the 10th cycle, both discharged at 200 mA, as displayed in Figure 6. Here, the data set was divided into two annular disks (or in 3D hollow cylinders) of equal areas. The inner ring had a radius of 0.70 mm to 2.20 mm and the outer ring was 2.20 mm to 3.03 mm. The graphs show that with progressive cycling, the particles at a larger distance to the collector nail had less total volume and a smaller average particle diameter compared to other particles in the cell. Particle migration seemed to cause or at least contribute to a separation between the larger and smaller particles.

**Figure 6.** Size distributions of the zinc particles in the discharged RAM cell. (**a**) Pristine and (**b**) after the 10th cycle (discharge current 200 mA).

#### *3.6. Comparison Between the Primary Cell and the RAM*

As can be seen in Figure 5, a much larger volume of zinc particles was present in the primary cell (compare Figure 5d to Figure 5a–c). In addition, the maximum total volume per diameter class in the primary cell was observed for particles with 0.15 mm ± 0.02 mm diameter, which was slightly smaller compared to the 0.17 mm ± 0.02 mm for the RAM. Unlike the large particles in the RAM, the large particles of the primary cell dissolved faster than the smaller particles, thus leading to a shift of the average particle diameter upon cycling (Figure 5d).

Direct comparison of slices taken from the RAM (Figure 7 bottom) and from the primary cell (Figure 7 top) revealed the different design of the two battery types. The graph in Figure 2c, representing the total volume of all particles, shows that the particle volumes declined sharply in both types of batteries after the first cycle. In the alkaline primary battery, the particle volume remained roughly constant in the ensuing cycles, whereas in the RAM it continued to decrease. This can be seen as well in Figure 5.

**Figure 7.** Comparative cut of the primary cell (**a**) and the RAM battery (**b**). Both cells are shown in their pristine state.

During the first discharge, the primary cell had the largest capacity (Figure 2a), which could be explained by the amount of zinc used. After their first cycle, all batteries lost a large part of their capacities. After 10 cycles, the capacities of the RAMs were still about 50% of their corresponding original capacity, while the primary cell reached only about 25% of the initial level. Furthermore, with an increasing number of cycles, the volume of the manganese dioxide cathode also increased, and the anode as well as the separator was increasingly compressed. This did not happen uniformly, but it created splinter-shaped structures that penetrated the separator. In the primary cell, this effect was much more evident. After the 12th cycle, the primary cell was no longer rechargeable and it began to leak.

After the failure of the primary alkaline battery, the tomogram exhibited some irregularities. Figure 8 demonstrates that many manganese dioxide spikes formed that pierced the separator (red circle). In addition, the manganese dioxide in the area around the separator was slightly brighter (blue arrow). This can be explained by a damaged separator so that zinc-enriched electrolyte could freely pass through to the cathode side. Therefore, the battery discharged itself and the manganese dioxide layer continued to expand, the internal pressure increased, and the battery started to leak. Presumably, in addition to the process of discharge, gas was formed, which may also have contributed to the increasing pressure [43].

**Figure 8.** Tomographic cross section of the primary cell after 12 cycles.
