Virtual unrolling of spirally-wound lithium-ion cells for correlative degradation studies and predictive fault detection

This method for virtually unrolling the jelly roll of a lithium ion cell using X-ray CT highlights the predictability of macro deformations formed while cycling. The failure is shown to propagate from nucleation points present since production.


Introduction
Lithium-ion batteries have become ubiquitous in modern life due to their high energy and power densities, relatively long cycle life and portability. This widespread deployment has generated a huge market and increased the need for improved performance, longevity and reliability of cells. The cycle life of a battery is governed by a wide array of factors 1,2 including, the temperatures that cells are exposed to, 3 rate of operation 4 and the depth-of-discharge of each cycle. 5 There are many varied degradation mechanisms 6 and similarly varied techniques for tracking them including acoustic time-of-ight measurements, 7 electrochemical methods, 8 and X-ray computed tomography (CT) among others. X-ray CT has been shown to be a powerful, non-destructive technique to investigate batteries with studies highlighting its efficacy in identifying microstructural evolution 9,10 and cell failure. 11,12 Using X-ray CT, Waldmann et al. 13 identied deformations in spirally-wound cells (which exhibit what is commonly referred to as a 'jelly-roll' architecture), of a number of cell chemistries as a result of high rate cycling. Comparison to cells cycled at low rates showed substantially larger deformations; similar deformations, observed towards the centre of the cell, were observed by Pfrang et al. 14 The capacity fade observed in the cells was ascribed to delamination of the cathode in the central region of the jelly-roll.
In this work, the structural deformation of the electrodes of a commercial cell are examined using X-ray CT. Virtual unrolling of the jelly-roll structure provides a radially resolved deformation map of a battery electrode revealing, for the rst time, that initial defects in the jelly-roll are nucleation points for electrode deformation which can be correlated to ageing and ultimately cell failure. The rapid nature of the X-ray CT scans, coupled with fast analysis, provides a pathway to an in-line quality assurance process, which currently rely on optical assessments 15 of the electrodes prior to assembly, which could have a signicant impact on safety, durability and recyclability of Li-ion batteries.

Methods & results
The major computational challenge in this study was the rigorous extraction of the jelly-roll. The 'edge' of the repeating layer was chosen to be the boundary between the anode and separator phase and the cathode as this offered the greatest contrast. The contrast was enhanced using an adaptive edge enhancement algorithm consisting of a linear deployment of a small kernel median lter, an unsharp mask and a Gaussian difference lter. The Cartesian coordinates of the edge were extracted using a modied marching cubes algorithm around the mean of the image. 16 Once extracted, the 2-dimensional jelly-roll should form a continuous Archimedean spiral, which in cylindrical coordinates is: The radius (r) plotted versus the angle (f) would produce a series of straight lines, each representing a different loop of the spiral. Fig. 1(b) demonstrates this trend between 0 and 270 degrees.
A coordinate system transform is performed on the extracted data. The centre of rotation of the cell is found by trying multiple centres and minimizing the resulting eccentricity of the cylindrical casing of the cell. Following that, the radius and angular position can be determined according to: where the zero subscript indicates the coordinates at the centre of the casing. An example of an extracted spiral can be seen in Fig. 1(a and b) shows the unrolled electrode contour for a number of different scan times. Fig. 1(b) also highlights the efficacy of this method for rapid in-line quality assurance. The inset image shows that essentially identical unrolled contours can be extracted with an 80 second or 26 minute acquisition. Coupled with computationally efficient methods presented here, the full shape of a spiral wound electrode can be imaged and analysed in under two minutes representing a substantial improvement in characterisation times using more traditional methods, which would require qualitatively analysing the greyscale images of the battery to assess the shape and quality of the jelly roll. Additionally these methods are completely non-destructive and non-intrusive.
The battery was cycled until its capacity had faded to approximately 79% of the initial capacity, which is the end of its primary life making it only useable in secondary applications. 17 Non-destructive 3D tomography of the battery aer cycling showed a small mechanical failure in the inner region of the jelly-roll, near the centre of the cell. Fig. 2(a) shows the capacity fade if the MJ1 over the rst 1097 cycles and Fig. 2(b) a series of orthoslices from tomograms at times between 0 and 1089 cycles. When comparing these two images it is clear there is a strong correlation between the amount of deformation in the centre of the jelly roll and the capacity loss experienced by the cell. Fig. 3 shows a progression of orthoslices from tomography performed at this point. Fig. 3(a) is a YZ orthoslice showing the deformation of the jelly-roll, the most prominent feature is the damage to the jelly roll. The middle image Fig. 3(b) shows an XY orthoslice through the area that shows the most signicant deformation and the right side Fig. 3(c) is a region-of-interest (ROI) scan of the most affected region of the jelly-roll. The capacity loss observed is likely from the delamination of and damage to the electrodes. The active materials in both electrodes are no longer in good electrical contact with one another, with the electrochemical pathways signicantly elongated as a result of the delamination. The 'kink' or deformation in the coil is relatively small; however, it spans a region of more than 25 mm inside the electrode itself, which results in noticeable capacity fade. Lastly, the high-resolution ROI scan shows the cathode inside the kink has fractured and delaminated from the current collector, further reducing the available active mass in the battery.
The mechanism of this deformation and delamination can be explored by analysing the unrolled jelly-roll as seen in Fig. 1(b) before and aer cycling. Comparing the results one orthoslice at a time would be very arduous, so the analysis was simplied by dividing the jelly-roll into different equal-sized layers of interest. Here, ve different regions are presented, each representing 100 orthoslices, totalling 3.6 mm of electrode, each. The electrode was divided into these regions based on the location of the external current collector tab which can be seen at the 270 degree position (bottom) of the orthoslice in Fig. 3(b).
The 'top' regions, shown in Fig. 4(b) are not inuenced directly by the presence of the current collector while the bottom regions are Fig. 4(a). The middle region, shown in blue spans the edge of the current collector tab, with roughly half above and half below its end. The data obtained from each of these regions is presented by extracting the coordinates of the unrolled electrode for each of the 100 slices and then overlaying the results on top of each other.  Theoretically at a given state of charge the electrode position should remain identical, regardless of cycle number, but this is not the case. Over extended cycling, architectural anomalies within the cell (for example the current collector tab), act as anchor points, constraining 'jelly-roll' movement. By contrast, other features (for example the free volume in the centre of the cell) provide freedom for electrode movement; consequently there is a differential displacement depending both on angular and radial position. We would expect that these irreversible deformations would form faster for a battery being cycled at higher rates and slower for a battery cycled at lower rates.    The major changes in the jelly-roll can be seen most clearly through the red and blue regions in Fig. 6. The jelly-roll irreversibly deforms outwardly (blue) into any space le available during the initial winding of the battery. These spaces are created by inconsistent surfaces external to the jelly-roll. The bottom of the battery imaged here is being inuenced by two of these features. The rst is the external current collector, visible at 270 degrees. On either side of the collector, there is signicant outward deformation in the jelly-roll. The second feature is the end of the spiral electrode, which is visible between 90 and 120 degrees, near the top of the gure. Again, aer cycling the electrode migrates and deforms into this space. In the top of the cell, the external current collector is no longer present but still has a signicant effect. While not an inconsistency in the manner described above, the end of the current collecting tab is a very signicant discontinuity in the electrode, and its abrupt termination leads to a very signicant void to expand into. Wherever the electrode was unable to deform outwardly it will deform inwardly (red). This leads to the most signicant inward deformation to be located in line with the anchor points described above. If sufficiently high forces are applied to inward deformation, a delamination event will occur, as is clearly evident in the cell demonstrated here. This will occur where the highest stresses are being generated which will likely be in line (radially) with the current collector.

Conclusions
The method presented here, and accompanying results, demonstrate a signicant advance in understanding Li-ion cell failure and degradation mechanisms. X-ray CT was used to image and examine an 18650 Li-ion cell as it aged and its capacity faded beyond primary use. By virtually unrolling the jelly-roll structure, the failure mechanism was able to be correlated with physical features and defects. Specically, the location of the eventual failure and delamination in the electrode was near an obvious imperfection in the jelly-roll that was visible before the battery was cycled. Additionally, this nucleation point for eventual failure existed because the electrode was imperfectly coiled due to the presence of the external tab for current collection which also facilitated the uneven stresses which developed in the electrode, causing the failure and delamination in the innermost coil, where those stresses would be the largest. The duration of the X-ray CT performed in this work can be substantially reduced at the expense of some resolution; however, the authors have noted that alternative scanning approaches using lab systems which can take as little as ca. 80 s are sufficient to achieve the required quality to perform the analysis described here. By virtually unrolling the battery, we are able to detect features and trends that would otherwise be very difficult to determine, which has led to improved understanding of how the battery operates and degrades. The rapid scan and calculation time of this technique provides an opportunity to use a similar method as part of an industrial quality assurance scheme. Doing this will increase the condence a manufacturer can have as to the reliability of the cells produced, potentially enabling a selection process in which cells used in high-stress environments or high-value such as electric vehicles can be tested in-line. The development of such a system, which would impose maximum allowable deviations in the jelly-roll, could also provide pass/fail criteria for recycled cells. This would greatly enhance the prospects for second life applications of cells which have the capability to perform in less demanding, e.g. stationary, applications improving the whole life efficiency and economics of the batteries produced.

Conflicts of interest
There are no conicts to declare.