Failure gas analysis of lithium-nickel-cobalt-aluminium oxide cells from different manufacturers

Lithium ion batteries (LIBs) are now commonplace industrially and domestically, innovations in their size and capability in terms of charge and discharge rates also mean LIB applications are growing. LIBs also present a unique challenge when the undesirable happens and they fail. One of the motifs of catastrophic LIB failure is the production of large volumes of flammable and toxic gas. Characterising LIB failure and the products of such events is an area of significant interest. In this work an array of nickel-cobalt-aluminium oxide (NCA) LIBs from four different manufacturers were failed predominantly by external heating but also by nail penetration. 18 permutations based on cell type and amounts of charge (69 tests in total) have been reported. Failure was carried out in inert atmospheres of nitrogen or argon inside a sealed vessel. After LIB failure, gas samples were taken, the volume calculated and the relative amounts of CO₂, CO, H₂, CH₄, C₂H₆, C₂H₄, C₃H₈ and C₃H₆ determined using mass spectrometry. The volume of gas produced during LIB failure by each cell type at 100% state of charge (SoC) has been analysed and is reported in the range of 1.34-2.32 L Ah^-1 for cells between 2 and 5 Ah sourced across four manufacturers. The volume of gas produced by LIB failure at differing amount of charge (AoC) has been determined for 2, 3, 4 and 5 Ah cells sourced from a single manufacturer. Variations in the volume of gas produced are shown to not only be dependent on AoC but also the type of cell has a material effect on this aspect of LIB failure. This work supports the existing consensus that as AoC increases, so does the volume of gas released as a result of LIB failure. In terms of gas composition a general trend of increase in flammable components and decrease in CO₂ once SoC is >50% has been observed in this dataset. This work also demonstrates that whilst LIB failure can produce some interesting phenomena, understanding and ultimately predicting the outcomes of LIB failure is difficult. The variations reported, even within a single cell manufacturer, suggests that for safety critical applications relying on generic or typical values is less useful than testing the precise cell being considered.