Study on the fire risk associated with a failure of large‐scale commercial LiFePO4/graphite and LiNixCoyMn1‐x‐yO2/graphite batteries

The fire hazards of fully charged large‐scale commercial LiFePO4/graphite and LiNixCoyMn1‐x‐yO2/graphite batteries are experimentally studied using a bench‐scale calorimetry apparatus. The battery burning process can be roughly summarized into three stages with significant criteria. The fire behaviors associated with LiNixCoyMn1‐x‐yO2/graphite battery give more splash spark, explosion, and gas/smoke ejection, while LiFePO4/graphite battery presents more jet flame. The sound signal may be a good choice for reflecting the battery state during thermal failure. The battery catches fire when average surface temperature (ST) reaches about 150°C. The maximum average STs for LiFePO4/graphite and LiNixCoyMn1‐x‐yO2/graphite batteries are approximately 535.3 and 658.7°C, respectively. The maximum heat release rate (HRR) of two batteries is comparable, while the total heat release for LiFePO4/graphite battery is higher than LiNixCoyMn1‐x‐yO2/graphite battery. The normalized heat release by initial mass of battery is found to be 2.304 and 3.133 kJ/g for LiFePO4/graphite and LiNixCoyMn1‐x‐yO2/graphite batteries, respectively. Besides, LiNixCoyMn1‐x‐yO2/graphite battery releases more CO and exhibits larger mass loss compared with LiFePO4/graphite battery. Finally, fire risk assessment for two batteries is also performed and discussed. In conclusion, LiNixCoyMn1‐x‐yO2/graphite battery is more hazardous than LiFePO4/graphite battery in current condition.

Previous researches on thermal behavior and fire hazards of LIB have been performed at small format battery. [9][10][11][12][13][14][15][16][17] Factors including the battery materials, 18,19 states of charge (SOC), 20,21 battery states, 22 heating modes, 23 ambient pressures, 13 battery modules, 24 and incident heat fluxes, 8 impacting the fire behavior of LIBs were focused. Ribiere et al 25 identified and quantified the toxic emissions (NO, HCl, HF, etc.) and energy release of small commercial pouch cells by using the Fire Propagation Apparatus. They found that the fully charged battery gave a slightly low total heat release while a relatively high toxic gas emissions. Ditch et al 26 conducted the free-burn fire tests to evaluate the flammability characteristics of cartoned small-format LIBs in warehouse scenarios. Larsson et al 27 exposed the commercial lithium iron phosphate cells and laptop battery packs to a controlled propane fire. They found that battery with higher SOC presented higher heat release rate (HRR) peak and a lower total emission of HF. Chen et al 19 employed a fire calorimeter to investigate the combustion behavior of two commercial 18 650 LIBs at different SOCs. They reported that the HRR and total heat release increased with SOC, and LiCoO 2 18 650 LIB gave higher explosion risk than LiFePO 4 18 650 LIB. A Copper Slug Battery Calorimetry (CSBC) was used by Liu et al 28 to measure the total energy generated inside the battery of 18 650 LIB at various SOCs. Combined with a cone calorimeter, it could quantify the flaming combustion of materials ejected from battery. Liu et al 18  In this work, the large-scale commercial fully charged LiFePO 4 /graphite and LiNi x Co y Mn 1-x-y O 2 /graphite batteries were used to explore the burning behaviors and fire risk. To characterize the fire risk of LIBs, the main parameters including sound signal, temperature, heat release, gas emissions, and mass loss were measured. The fire hazards corresponding to two composition schemes batteries were also compared and discussed further. The findings may be extend to the battery modules composed of different types of small format batteries, and the larger size batteries to some extent.

| EXPERIMENTAL SETUP
Two kinds of large-scale commercial batteries tested in this work are manufactured by China Aviation Lithium Battery Co., Ltd., which are applied to energy storage devices. The  Table 1 presents the detailed information of LIBs. The LIB is charged to the expected SOC and then is allowed to a rest for 24 h prior to all tests. Figure 2 presents a schematic diagram of the thermal hazard test apparatus. The apparatus was developed to fill the needs of battery thermal runaway and burning test based on the full scale room test (ISO 9705) and the cone calorimeter test (ISO 5660). The battery sample was placed and installed horizontally on a sample holder made by stainless mesh. An electrical balance was applied to record the mass loss of sample. The K-type thermocouples with diameter of 1 mm were attached on the surface of battery to measure the surface temperature (ST) profiles. The gaseous products of combustion were collected and analyzed by a Servomex 4100 gas analyzer (Servomex, East Sussex, UK), which would be devoted to measuring the concentrations of CO 2 , CO and O 2 , and estimating the HRR based on the oxygen depletion method. An electric heater with a power of 2 kW was taken as an external heating source to heat the battery and trigger the thermal runaway of battery. A digital camera with a high resolution of 25 fps was used to record the whole experimental process for obtaining the visually visible photographs. Figure 3 shows the typical sequence photos of the experiment. The significant thermal behaviors both 135 LIB and 148 LIB can be clearly seen. The whole process can be roughly divided into three stages: stage (a)-LIB is heated to ignite, stage (b)-LIB is involved in splash spark, hissing sound, jet flame, general flame, explosion and gas/smoke ejection, and stage (c)-LIB is gradually out of flame. Multiple outbursts from the battery are visually observed. However, the difference of the composition between the 135 LIB and 148 LIB may cause completely different burning behaviors. Apparently, 135 LIB presents more vigorous bright flame, while 148 LIB exhibits more violent explosion and gas emission. In addition, the time to ignition (TTI) of 135 LIB and 148 LIB is approximately of 525.4 and 368.2 seconds, respectively. The duration of combustion process is about 443.5 and 156.1 seconds for 135 LIB and 148 LIB, respectively. All the above indicates that 148 LIB is easier to ignite and burns faster than 135 LIB, which reflects that the thermal stability of 148 LIB is less than that of 135 LIB to some extent. Figure 4 depicts the original sound recorded in the whole experimental process. The data can also reflect the fierce level of combustion and the different thermal stages. This may provide

| Surface temperature
The ST is a key parameter reflecting the thermal failure of LIB. Considering the nonuniform temperature distribution of large-scale battery, six thermocouples are used to monitor the ST. The temperature profiles of LIB are shown in Figure 5.

| Heat release and gas emissions
The heat release rate results for 135 LIB and 148 LIB are illustrated in Figure 7. In current work, the HRR estimated is based on the oxygen consumption, corrected by carbon dioxide and carbon monoxide production. Another critical factor related to battery combustion is the gases depletion and emission. During the whole experimental process, the O 2 is consumed and the CO and CO 2 are emitted. The gas concentrations and emission yields are continuously measured by a Servomex 4100 gas analyzer. The gas analysis from the battery combustion allows assessing the heat release, combustion efficiency, and toxic threat. Figure 8 shows the evaluation of O 2 consumption. The maximum consumption value of oxygen is about 2.65% and 2.25% for 135 LIB and 148 LIB, respectively. Figure 9 illustrates the concentration of gaseous emissions. It can be noted that the content of CO 2 is significant higher than CO, regardless of the battery formats. The ratios of peak and total contents between CO 2 and CO are 58.44 and 48.88 for 135 LIB, while they are 17.83 and 19.56 for 148 LIB. This result highlights that the 135 LIB has higher combustion efficiency and lower toxic risk.

| Mass loss
The mass loss of battery is measured by using a Mettler Toledo XP10002S with a range of 0-9 kg and a readability of 0.01 g. The battery mass at different times is normalized by the original mass. 20.81% mass loss of 135 LIB is obtained and the data are about 27.54% for 148 LIB. The maximum mass loss rate is also estimated. They are 0.25% s −1 and 4.71% s −1 for 135 LIB and 148 LIB, respectively. Note that 148 LIB reveals a larger and faster mass loss, which indicates that the 148 LIB may exhibit violent chemical reactions and incomplete combustion combining the analysis of heat release and gas emission (Figure 10).

| DISCUSSION
In the process of thermal analysis of LIB, it was always be simplified as an isotropy and homogenous structure. For a prismatic LIB in current work, the bottom surface of battery was heated by an electric heater, and the top and lateral surfaces were exposed to the ambient atmosphere. The thermal balance of battery can be dominated by heat accumulation and heat loss, as expressed in the following conversation equation: 5 where C p and T are the density, the specific heat and the temperature, respectively, t is the time, Q input , Q chem and Q loss are the input heat of the external heater, the heat generation by the chemical reaction, and the heat loss from battery, respectively.
As the battery temperature increasing, several superheated side reactions including the decomposition of solid  where ΔH, M and R are the reaction heat, the mass of reactants and the gas constant, respectively, n, A and E a are the reaction order, the pre-exponential factor, and the activation energy, respectively. The decomposition temperature of SEI is about 80-120°C. As the temperature increases to 110°C, some combustible gases will be released due to electrolyte decomposition. With a continuous increase of temperature, the separator will melt at 135-160°C, and the cathode will decompose above 200°C. Finally, the battery becomes unstable and is out of control.
The 135 LIB with LFP as cathode has good thermal stability than the 148 LIB employed NMC as cathode. Compared with the NMC, the LFP shows excellent thermal tolerance against LiPF 6 -based electrolyte, which can suppress the decomposition of electrolyte and has a slight structure change to some extent. 33 It may be reflected in the earlier and easier thermal failure for 148 LIB than that for 135 LIB. However, NMC underwent a phase transition from layered structure to crystalline structure with a space group Fm3m in the range of 220-350°C. 34 The overall decomposition reaction of NMC can be expressed as: 15 Thus, it should be noted that the oxygen was released. Similarly, the oxygen generated from the LFP cathode can also described as: Following, the released oxygen would react with EC, as shown in the reaction: Based on these analyses, it can be conjectured that the thermal hazard of LIB was induced by the large amount of heat accumulation and gases generation caused by the chemical reactions involving the cathode, electrolyte, and dropped electrode materials. The 135 LIB released combustible gases, accompanied by inner pressure decreasing and immediately followed by a considerable jet fire. In addition to fire and explosion, the 148 LIB experienced multiple gas/smoke ejections without flame. However, the emitted flammable and toxic gases can be a serious problem for LIB fire. Swelling was also observed for two formats battery. Finally, the distinguishing thermal behaviors can be explained by the different thermal failure mechanisms of two composition batteries attributed to the varied type and intensity chemical reactions. That is deserved to have an in-depth study in future work. Table 2 shows the comparison of parameters associated with fire hazards of the LIBs. It can be seen that the composition scheme of battery is a key factor for fire risk of battery. Compare to 135 LIB, the 148 LIB is the most hazardous and illustrates a potential threat for the safety based on an evaluation method proposed by Petrella. 35 Besides, the toxic risk is also obvious for 148 LIB according to the production of CO and CO 2 . In summary, 148 LIB has a higher overall reactivity and fire hazard than 135 LIB in current study.

| CONCLUSIONS
In this work, the fire hazard associated with the failure of large-scale battery was examined experimentally under fully charged. Two commercial composition schemes of LIB commonly applied in electric vehicles were selected to facilitate the study. One is LiFePO 4 /graphite battery (135 LIB). The other is LiNi x Co y Mn 1-x-y O 2 /graphite battery (148 LIB). The thermal behavior, sound signal, ST, heat release, gas emission, and mass loss were recorded and measured. The effect of composition scheme on fire hazard of battery was also discussed. (2) (4) 4FePO 4 → 2Fe 2 P 2 O 7 + O 2 . The bold values represent the key parameter The whole experimental process can be roughly divided into three representative phases, regardless of 135 LIB and 148 LIB. However, 148 LIB presents more splash spark, explosion, and gas/smoke ejection in stage (b) while the jet flame is the main thermal behavior for 135 LIB. The sound signal can also reflect the battery failure state during the experiments. Besides, it is illustrated that the thermal failure and ignition of 148 LIB are relatively easier than that of 135 LIB. Compared with 135 LIB, 148 LIB combusts incompletely, taking the consumption of O 2 , the generation of CO 2 and CO, and the mass loss of battery into account. But, the normalized heat release by initial mass of battery was calculated. The value obtained as 2.304 kJ/g for 135 LIB is less than 3.133 kJ/g for 148 LIB, which is significantly lower than that of commonly solid combustibles. Meanwhile, 148 LIB releases approximately 3.0 times more toxic CO than 135 LIB. The total mass loss is about 380.2 and 250.6 g for 135 LIB and 148 LIB, respectively. Finally, we conducted the systematic assessment of fire risk for batteries and the fire risk of two batteries was evaluated by Petrella method. In summary, 148 LIB is more hazardous than 135 LIB based on normalized heat release and toxic gas emissions in current work. Certainly, the active components inside battery mainly determine the fire behavior of integral battery. Thus, the fire performance of each active component will be further checked and the comprehensive comparative study will be fulfilled in the future work.