Anup Barai’s research while affiliated with University of Warwick and other places

Lithium-Ion Batteries (LIBs) have been at the epicentre of the global push to electrify transport. Having already become mainstream in the automotive industry, vehicles powered by LIBs are gaining traction in other transport sectors such as aerospace and freight trucks. These sectors have more stringent requirements for performance as well as safety. Therefore, academia as well as industry are aggressively exploring battery diagnostic methods that can be implemented on-board [1]. A common feature shared among several well-developed tools for battery diagnostics is their reliance on an electrochemical technique called Incremental Capacity Analysis (ICA) [2,3]. ICA is ideal for in-field implementation because it: (a) requires only voltage and current data, which are available irrespective of application and (b) provides additional insights into electrode-level phenomena that are direct indicators of battery health. Therefore, ICA is also used by Battery Analytics companies for health and fault monitoring [4]. The long-term evolution in the understanding of ICA has captured the effects of factors such as C-rates, temperature, charge-discharge hysteresis etc. in literature [5]. However, one fundamental assumption omnipresent across all previous studies is the use of the entire battery voltage range: majority of ICA studies consist of the battery being charged/discharged over the entire 0-100% % State-of-Charge (SoC) interval (100 to 0 for discharge). While this assumption is valid for laboratory-based testing and validation, batteries in transport applications seldom undergo a complete charge/discharge. For instance, road EV manufacturers limit the available SoC windows to prolong the battery life. On the other hand, the electric aviation industry reserves 30% of the battery capacity for emergency scenarios. Therefore, there is a lack of studies as well as data considering the implementation of ICA with partial cycling data, which is a true representative of real-life use cases. The presented work addresses the identified gap through an experimental investigation of ICA results obtained from partial charging data. Commercial Li-ion cells were cycled using multiple C-rates and Depths of Discharge (DOD). ICA was conducted on data obtained from full as well as partial charging regimes. As shown in Figure 1, the obtained ICA results highlighted a direction-dependence: the ICA curves from partial charge cycles commencing from the same SOC indicate a disparity based on the direction of the previous cycling history (charge or discharge). This direction-dependence has not been considered in battery research literature and could invalidate the previously established protocols of degradation diagnostics performed using ICA. Moreover, the results highlight a major deficiency in large battery dataset creation efforts: being generated to support the ultimate implementation of Artificial Intelligence (AI) [6] in predicting battery behaviour, the datasets are incomplete due to a lack of partial cycling data and would not be relevant for real-life application. Further work is ongoing to investigate the phenomenon further and evaluate the implications in current practices of battery diagnostics and large dataset generation. Figure 1

The conventional approaches of estimating the lifecycle of lithium-ion batteries utilize battery ageing models combined with data from accelerated ageing tests. Despite the prevalent use of these accelerated techniques in battery research over the years, there still exists a lack of consistency in the testing parameters and methodology. Particularly in cycle ageing tests, the rest period between cycles is typically selected without a standard criterion and is generally minimised to facilitate higher cycle throughput. Although this parameter has been considered to have minimal influence on the test results ¹ , a recent study ² has demonstrated that cycle-to-cycle relaxation may have a significant impact on the cell cycle life, even though the investigated cells were cycled in otherwise identical conditions. This work aims to expand on the findings presented in ² , by investigating the underlying mechanism of this behaviour, using X-ray CT imaging, forensic and post-mortem analysis. In this study, commercially available cylindrical lithium-ion cells were subjected to prolonged cycling with different rest periods between cycles. For the relaxation time after discharge, three time constants were selected, namely 1min, 10min and 60min, while the relaxation period after charge was kept constant at 1min. The cells were cycled at 10°C between the charge/discharge voltage cut-off limits specified by the manufacturer, with 0.3C/1C current rate for charge/discharge respectively. A clear correlation between relaxation time and cell degradation was observed in the cycling results, as the cells cycled with longer rest periods exhibited higher capacity fade (Fig.1(a)). Non-invasive diagnostic analysis conducted with differential voltage (DVA) and incremental capacity (ICA) revealed increased losses of lithium inventory and active material in the longer relaxation test cases. As the main relaxation time was after discharge, when the cells were at a low SOC and in sub-ambient temperature (10°C), calendar ageing effects are expected to be negligible, which indicates the presence of a different underlying mechanism responsible for the observed degradation pattern. At the end of the cycle ageing test, X-ray computed tomography (XCT) scans were obtained for all the tested cells, and one cell per test case was selected to be dismantled for autopsy and visual inspection. The XCT images revealed severe deformations in the jelly roll of all tested cells, in the form of kinks located in the innermost layers, between the cell's hollow core and the positive current collector tab (Fig.1(b)). Furthermore, the cell structural integrity appears to deteriorate as the relaxation period increases, since more jelly roll layers appear to be affected in the 10min and 60min test cases (Fig.1(b)). Upon dismantling, significant electrode delamination was observed in the innermost electrode areas for all test cases (Fig.1(c)-(e)), which was inherently linked to the jelly roll deformation. For the cells cycled with 60min relaxation, a bright coloured passivation layer was detected in the mid to outer layers of the negative electrode (Fig.1(e)-(f)), implying the potential occurrence of lithium plating ³ . Furthermore, the opposing parts of the cathode electrode corresponding to the location of the passivation layer in the anode exhibited a unique wave-like deformation pattern (Fig.1(f)), which indicates that these jelly roll regions where subjected to high mechanical stresses during cycling. The autopsy results support the preliminary diagnostic conclusions, as the presence of the covering layer and the morphological changes in the electrodes can lead to loss of active lithium and active material, due to local passivation and particle cracking. Overall, this study demonstrates that longer cycle-to-cycle relaxation periods cause higher mechanical and electrochemical stresses within the tested lithium-ion cells, which can even lead to the formation of local passivation layers, affecting their electrochemical performance and structural integrity. This effect, which has not been previously reported, does not appear to be present in all cells, as a preliminary investigation conducted with three different commercial cells of similar format did not exhibit the same trend. Although the underlying mechanisms responsible for the results presented in this study needs to be investigated further, the potential influence of relaxation on the cycling behaviour of lithium-ion cells needs to be considered in the design and interpretation of accelerated ageing tests. Reichert, D. Andre, A. Rösmann, P. Janssen, H. G. Bremes, D. U. Sauer, S. Passerini, and M. Winter, Journal of Power Sources, 239 45-53 (2013). G. Darikas, A. Barai, M. Sheikh, P. Miller, M. Amor-Segan, and D. Greenwood, in "Electrochemical Society Meeting Abstracts 242", p. 2590-2590. The Electrochemical Society, Inc., 2022. A. J. Smith, Y. Fang, A. Mikheenkova, H. Ekström, P. Svens, I. Ahmed, M. J. Lacey, G. Lindbergh, I. Furó, and R. W. Lindström, Journal of Power Sources, 573 233118 (2023). Figure 1