In the context of the global transition toward sustainable mobility, the deployment of electric vehicles has been growing at an unprecedented pace. The lithium-ion power battery, commonly referred to as the EV battery, is the central energy storage device that determines driving range, cost, and operational safety. Although remarkable improvements have been made in energy density and cycle life, the safety of EV battery systems under extreme abuse conditions remains a critical bottleneck. Thermal runaway incidents induced by overcharging or by external heating have repeatedly caused severe vehicle fires, highlighting the necessity of understanding the underlying failure mechanisms of EV battery cells across a broad range of environmental temperatures. In my research, I deliberately focused on two typical abuse categories: electrical abuse in the form of excessive charge, and thermal abuse generated by external heat while the battery is discharging. By systematically varying the environmental temperature from –25 °C to 75 °C, I examined how low-temperature and high-temperature environments alter the voltage evolution, temperature rise, gas venting, internal resistance, crystal structure, and thermal-runaway severity of different commercial lithium-ion batteries.
The safety of an EV battery is not an intrinsic quantity but rather a state-dependent response that depends strongly on temperature, charge state, current rate, and heat dissipation conditions. In my first series of tests, I used a 23 Ah square lithium iron phosphate battery at 100% state of charge and subjected it to forced overcharging at 1 C until the safety valve opened. The environmental temperatures were selected as –25 °C, 0 °C, 25 °C, 50 °C, and 75 °C. In the second series, I selected a cylindrical 26650 ternary lithium-ion battery and studied its electrical and thermal behavior when overcharged at 3 C under the same five temperature levels. In the third series, the same 26650 ternary cell was discharged at 1 C while being externally heated at 120 W, causing the cell to experience thermal runaway. The combination of electrical abuse and thermal abuse tests allowed me to compare failure routes, identify the critical warning points, and evaluate the degree of material degradation via post-mortem microscopic characterization.
1. Experimental Setup and Methodology
All cells used in my experiments were commercially manufactured and newly purchased. Before any abuse test, I conducted a pretreatment on every cell to activate the electrode – electrolyte interfaces and to eliminate the sample-to-sample variance. The pretreatment involved three charge/discharge cycles at 0.5 C inside a temperature-controlled chamber maintained at 25 °C. After the activation cycles, only cells whose discharge capacity fell inside the preselected interval and whose internal resistance measured within a narrow tolerance were chosen for the abuse study. The internal resistance was measured with a battery internal-resistance tester based on the alternating current method. During the overcharge and thermal runaway tests, K-type thermocouples were attached to the battery surface at symmetric positions, while the voltage and current were recorded by a data acquisition unit at a sampling frequency of 1 Hz. The battery temperatures and voltages were logged with synchronized timestamps so that I could reconstruct the complete sequence of events during each abuse scenario.
I performed all the electrical abuse tests inside an explosion-proof temperature chamber. The environmental temperature was first allowed to stabilize at the predetermined setpoint, then the fully charged cell was left in the chamber for another 30 minutes to ensure thermal equilibrium inside the jelly roll. The charge and discharge operations were controlled by dedicated battery cycler systems. For the forced-overcharge tests, the current was applied continuously until the cell encountered a severe voltage jump and subsequently opened its safety valve. Because the sequence of chemical reactions differs drastically with temperature, the voltage curves could be divided into different characteristic stages. I tracked not only the transient temperature maxima but also the time differences between sudden voltage rises and the hottest moments, which are valuable for early-warning design of EV battery management systems.
For the thermal-abuse tests in the discharge regime, the ternary cell was first fully charged to 4.2 V using the standard constant-current constant-voltage protocol. A polyimide-insulated heating film with dimensions of 50 mm by 50 mm was bonded to the surface of the cylindrical cell, and the center of the heater coincided with the center of the battery. The cell was then placed in the temperature-controlled explosion-proof chamber. At the moment when the chamber reached the target ambient temperature, the cell was allowed to rest for 30 minutes. After that, the cell was discharged at a 1 C rate for 5 minutes, after which I switched on the heater power and supplied a constant heating power of 120 W until thermal runaway was observed. The rapid increase of temperature, the emission of white smoke, the abrupt voltage drop, the opening of the safety valve, and the eventual jet flame were all recorded by an internal video camera. After the test, the ejected powders and the disassembled electrode materials were collected inside an argon-filled glovebox with oxygen and water contents below 2.0 ppm and 0.1 ppm, respectively. Scanning electron microscopy, energy-dispersive X-ray spectroscopy, and X-ray diffraction were used to examine the morphological, elemental, and crystallographic changes after the failure.
2. The Role of Environmental Temperature in the Overcharge Failure of a Lithium Iron Phosphate EV Battery
The first part of my study focused on the 23 Ah square lithium iron phosphate cell. The principal parameters of this battery are summarized in Table 1.
| Item | Specification |
| Positive electrode material | LiFePO₄ |
| Negative electrode material | Graphite |
| Dimensions (height × width × length) | 131 mm × 24.2 mm × 68 mm |
| Nominal capacity | 23 Ah |
| Nominal voltage | 3.2 V |
| Upper cut-off voltage | 3.65 V |
| Lower discharge cut-off voltage | 2.0 V |
All cells were charged to a fully charged state using the constant-current constant-voltage method in which the current was 0.5 C and the end current was one twentieth of the charging current. After this conditioning step, the cell internal resistance was checked again. I accepted cells whose internal resistance at 100% SOC was between 4 mΩ and 6 mΩ. Then the cell was transferred into the constant-temperature explosion-proof chamber. The overcharging procedure was performed at a 1 C rate, i.e. 23 A, while the data acquisition system recorded the voltage and surface temperature at 1 Hz. The experiments continued until the safety valve burst. The observed gas venting was always accompanied by a thick white cloud consisting of electrolyte decomposition products, lithium-containing compounds, and electrode fragments mixed with flammable gases. I identified a sequence of physical phenomena that can be written as follows: the cell initially expands because of the gases generated by the reaction between lithium deposited on the anode and the electrolyte; after the internal pressure exceeds the rated pressure of the safety valve, the valve opens; eventually the cell cools back down and reaches a safe state if no ignition occurs.
In the low-temperature cases at –25 °C and 0 °C, the gas venting lasted longer and much more gas was released than at high temperatures. This observation indicates that, at low ambient temperatures, the parasitic reactions between metallic lithium and the liquid electrolyte proceed for a longer duration before the internal short circuit becomes significant. In addition, the lower temperature delays the shutdown of the separator and slows down the kinetics of internal short-circuit propagation, meaning that the cell does not instantly enter thermal runaway. Such a phenomenon is beneficial in terms of personal safety, since the users are given more time to escape if an overcharge incident occurs in a low-temperature environment. However, the extended venting process also implies a larger accumulation of flammable gases in the surrounding atmosphere, which poses a secondary fire hazard. From the point of view of my experiments, therefore, low temperature reduces the risk of direct explosion but does not eliminate the fire risk associated with combustible gas buildup.
Regarding the voltage evolution of the lithium iron phosphate cell, I observed a marked dependence on ambient temperature. At –25 °C and 0 °C, the voltage-versus-time curve could be divided into five stages. At 25 °C, only four stages were observed. At 50 °C and 75 °C, the voltage change became even simpler and could be classified into three stages. The origin of the different stages is rooted in the competition among several processes: lithium extraction from the positive electrode, migration of lithium ions through the electrolyte, lithium insertion into the graphite anode, decomposition of the solid-electrolyte interphase, lithium-plating growth, electrolyte oxidation, and ultimately separator collapse. In the first stage, the battery voltage climbs sharply because the cathode is almost fully delithiated and the anode is saturated with lithium. The over-removal of lithium from the LiFePO₄ lattice causes a sharp rise of the electrode potential. In the low-temperature environment, the charge-transfer resistance is dramatically increased because of poor ionic conductivity, so the initial polarization is very strong and the voltage rises from 3.4 V to 5.6 V in about 26 s. At higher temperatures, the initial voltage rise is less steep, because the electrolyte remains sufficiently conductive and the side reactions begin at lower voltages.
The voltage plateaus observed in the high-SOC overcharge region are connected to the decomposition of the SEI layer and the consumption of lithium by electrolyte reduction. The main reactions responsible for the first plateau are the decomposition of metastable alkyl carbonates inside the SEI, such as the reaction that produces Li₂CO₃ and ethylene gas:
$$(\mathrm{CH_2OCO_2Li})_2 \rightarrow \mathrm{Li_2CO_3} + \mathrm{C_2H_4} + \mathrm{CO_2} + 0.5\,\mathrm{O_2}$$
Several other reactions between metallic lithium and the solvent molecules can also take place, generating saturated alkane gases. If ethylene carbonate is considered, the reaction can be written as
$$\mathrm{C_3H_4O_3(EC)} + 2\,\mathrm{Li} \rightarrow \mathrm{Li_2CO_3} + \mathrm{C_2H_4}$$
Similarly, propylene carbonate and dimethyl carbonate react with lithium to generate propane and ethane, respectively, releasing additional gas that produces cell swelling. These parasitic reactions are unavoidable when an EV battery is overcharged because the negative electrode is forced well beyond its lithium-storage limit. The onset and duration of these reactions are strongly controlled by the ambient temperature. At low temperatures, the solvent molecules have lower mobility and the reaction kinetics are slower; therefore, the voltage plateau is longer. At high temperatures, the kinetics of the SEI decomposition and the electrolyte-oxidation reactions accelerate, which reduces the number of distinct voltage stages and hastens the voltage jump that is often treated as the onset of a high-risk condition.
During the overcharging tests, I monitored the transient voltage and temperature profiles. Figure values for the key parameters are listed in Table 2. I defined the voltage surge point as the moment at which the voltage suddenly increases to a high level, usually because of localized separator dry-out or because of the opening of the positive temperature coefficient protection device. The thermal runaway risk point was taken as the moment at which the surface temperature reached its maximum value.
| Ambient temperature (°C) | Voltage surge time (s) | Time interval from surge to maximum-temperature point (s) | Maximum temperature (°C) | Maximum temperature difference from initial value (°C) |
| –25 | ≈ 940 | > 100 | ≈ 116 | ≈ 141 |
| 0 | ≈ 620 | > 100 | ≈ 105 | ≈ 105 |
| 25 | ≈ 490 | > 100 | ≈ 118 | ≈ 93 |
| 50 | ≈ 530 | > 100 | ≈ 180 | ≈ 130 |
| 75 | ≈ 295 | > 100 | ≈ 223 | ≈ 148 |
The low-temperature case at –25 °C produced a maximum temperature difference of about 141 °C, while the 75 °C case produced a maximum temperature difference of about 148 °C. This finding reveals that both extreme low and extreme high ambient temperatures enhance heat generation. In the cold environment, the electrolyte resistance and the charge-transfer resistance become very large, so the Joule heating produced by the overcharging current becomes the dominant heat source. At high ambient temperature, the electrochemical side reactions are strongly accelerated and the onset of internal short circuits is faster. In the mid-temperature range, the total heat dissipation and heat generation are better balanced, but the time from the voltage surge to the highest temperature point is still longer than 100 s. Therefore, from a safety-strategy viewpoint, the voltage surge signal can be used as a pre-warning point for a lithium iron phosphate EV battery that is being overcharged, because it allows more than one hundred seconds after the surge before the cell reaches its most dangerous state.
The internal resistance of the lithium iron phosphate cell was measured at several overcharging stages, corresponding to 105%, 110%, 115%, and 120% SOC. The resistance changes relative to the initial 100% SOC resistance are presented in Table 3. In general, before the positive temperature coefficient action is triggered, the internal resistance increases with the overcharge SOC because of continuous lithium plating and electrode-structure changes. The largest increase appears at 115% SOC, where the separator pores may close and the positive temperature coefficient device is triggered. If the overcharging continues after that point, the internal resistance surprisingly decreases, which is caused by meltdown of some separator regions and the creation of a lower-resistance local short-circuit path.
| Overcharge SOC level | Internal resistance change at –25 °C (mΩ) | Internal resistance change at 0 °C (mΩ) | Internal resistance change at 25 °C (mΩ) | Internal resistance change at 50 °C (mΩ) | Internal resistance change at 75 °C (mΩ) |
| 105% SOC | 0.892 | 0.041 | 0.587 | 1.085 | 1.324 |
| 110% SOC | 1.626 | 0.569 | 3.165 | 25.570 | 5.559 |
| 115% SOC | 2.590 | 2.924 | 247.360 | 32.639 | 80.728 |
| 120% SOC | 0.332 | 0.526 | 32.310 | 20.684 | 13.085 |
The enormous resistance increase at 25 °C and 115% SOC, as high as 247 mΩ, indicates that the positive temperature coefficient protection mechanism inside the lithium iron phosphate cell is strongly activated at room temperature. In the high-temperature environment of 75 °C, the activation also takes place but leads to a smaller resistance excursion, because the thermal protection mechanism itself is affected by the external heat. When the ambient temperature is as low as –25 °C, the resistance increase remains modest and the thermal protection is not strongly triggered, which explains why the cell can continue to vent gas for a long time without rapidly progressing to thermal runaway. These measurements confirm that a low-temperature environment can suppress the abrupt thermal acceleration of an EV battery under overcharge, whereas high ambient temperature reduces the safety margin and makes the cell more prone to thermal runaway.
3. High-Rate Overcharge Behavior of a 26650 Ternary Lithium-Ion EV Battery
The second category that I studied was a cylindrical 26650 ternary lithium-ion battery, whose main specifications are collected in Table 4. The chemistry of the positive electrode is Li(NiCoMn)O₂, and the negative electrode is graphite. This type of EV battery has a higher energy density and better low-temperature discharge capability than a lithium iron phosphate battery, but it tends to release more energy during thermal runaway. Therefore, understanding its overcharge failure under broad temperature conditions is of great importance for the battery management systems of modern electric vehicles.
| Parameter | Value |
| Cell format | 26650 cylindrical |
| Positive electrode | Li(NiCoMn)O₂ |
| Negative electrode | Graphite |
| Diameter × height | 26 mm × 65 mm |
| Nominal capacity | 5 Ah |
| Nominal voltage | 3.6 V |
| Upper charging voltage | 4.2 V |
| Discharge cut-off voltage | 2.75 V |
Before each overcharge test, the ternary cells were fully charged at 0.5 C and then rested at the target ambient temperature. The overcharge current was 3 C, which corresponds to 15 A. Because of the high current and the high energy density, the overcharge process could quickly drive the cell into dangerous states. I terminated the tests when the cell went into internal short circuit and triggered its safety mechanism. The voltage and temperature profiles obtained at five ambient temperatures are summarized in terms of the number of voltage stages in Table 5. A voltage stage is identified as a visually recognizable plateau, an inflection, or a sudden change of slope in the voltage-time curve. The number of stages decreases with increasing ambient temperature. At –25 °C, the overcharge curve can be divided into six distinct stages; at 0 °C and 25 °C, five stages were detected; and at 50 °C and 75 °C, only two broad stages appeared.
| Ambient temperature (°C) | Number of voltage stages | Main characteristics of the first stage | Main characteristics of the last stage |
| –25 | 6 | Very steep voltage increase from 4.2 V to 8.4 V in 13 s | Rapid voltage jump to 109 V followed by thermal runaway |
| 0 | 5 | Initial voltage is already affected by electrolyte oxidation | Voltage jump and internal short circuit |
| 25 | 5 | Steady but less steep voltage rise | Internal short circuit and safety valve opening |
| 50 | 2 | Gradual voltage rise over a long period | Sudden voltage jump and failure |
| 75 | 2 | Gradual voltage rise | Sudden voltage jump and thermal runaway |
At –25 °C, the voltage initially rises from 4.2 V to about 8.4 V within 13 s, indicating a severe electrode polarization that is a typical limitation of EV battery operation at freezing temperature. The high overvoltage forces the electrolyte to oxidize at the positive electrode and leads to a subsequent voltage drop that forms an inverted-U section. The two voltage plateaus observed at –25 °C are linked to the decomposition of the SEI layer and to the accumulation of dead lithium on the anode. The first plateau occurs because the SEI layer is being oxidatively decomposed, consuming active lithium. The second plateau is associated with continuous lithium plating. In the final stage, the temperature rises sharply and the voltage suddenly jumps to a very high value, indicating the complete loss of separator function and internal short circuits. At 50 °C and 75 °C, by contrast, the SEI decomposition, electrolyte oxidation, lithium plating, and electrode degradation all overlap in one single voltage-rise stage, so the clear plateaus disappear. This finding is important because it indicates that the number of recognizable pre-warning signals in the voltage curve is not universal but depends strongly on the environmental temperature.
At low ambient temperatures, the maximum temperature difference of the ternary EV battery during overcharge was found to be much larger than that at room or high temperature. The maximum temperature difference at –25 °C was the highest among the tested temperatures. This is because the ohmic resistance and the charge-transfer resistance increase dramatically as the temperature falls, leading to stronger Joule heating for the same overcharge current. In addition, the poor lithium diffusivity in the graphite anode forces more lithium to be deposited on the surface rather than inserted into the structure. The lithium deposit then reacts with the electrolyte and produces additional heat and gas. The temperature-dependent response can be approximately expressed by the energy conservation equation:
$$Q_{\mathrm{total}} = Q_{\mathrm{joule}} + Q_{\mathrm{reaction}} + Q_{\mathrm{mixing}}$$
where \(Q_{\mathrm{joule}}\) is the irreversible Joule heat, \(Q_{\mathrm{reaction}}\) is the heat released by electrochemical side reactions, and \(Q_{\mathrm{mixing}}\) is the entropic heat contribution. The irreversible heat can be written as
$$Q_{\mathrm{joule}} = \int I (V – U_{\mathrm{OCV}})\,\mathrm{d}t$$
where \(I\) is the applied current, \(V\) is the cell voltage, and \(U_{\mathrm{OCV}}\) is the open-circuit voltage. In a low-temperature overcharge event, the gap between \(V\) and \(U_{\mathrm{OCV}}\) is enlarged, so the Joule heating term increases. Moreover, the term \(Q_{\mathrm{reaction}}\) contains the heat of SEI decomposition, lithium-plating corrosion, electrolyte oxidation, and separator shrinkage, all of which are influenced by the temperature field as well.
Figure 3 shows the effect of ambient temperature on the voltage-jump time and the time difference between the voltage-jump point and the highest-temperature point. The voltage-jump time first decreases as the temperature increases from –25 °C to 25 °C, but then slightly increases at 50 °C and finally decreases sharply at 75 °C. The shortest voltage-jump time is observed at 25 °C, which means that room-temperature overcharging leads to the fastest failure in terms of voltage breakdown. This non-monotonic behavior is a nontrivial result for EV battery management system calibration. At low temperatures, the increased hardness of the separator and the lower rate of lithium dendrite growth delay the internal short circuit. At high temperatures, the viscosity of the electrolyte decreases and the lithium plating layer is more adhesive on the graphite surface, also delaying the abrupt voltage jump to some extent. However, when the temperature approaches 75 °C, the accelerated chemistry overcomes these effects and the failure occurs very quickly. Therefore, overcharging a ternary EV battery at room temperature may not necessarily provide a larger safety margin; in contrast, the time available for corrective intervention is the shortest at 25 °C.
In addition to the transient voltage and temperature trends, I measured the internal resistance of the ternary cells after they were overcharged to different SOC levels. Table 6 lists the internal resistance changes for five ambient temperatures. The values at 120% SOC reveal an enormous increase at 50 °C and 75 °C, corresponding to the triggering of the positive temperature coefficient device and the shut-down of the separator. At 130% and 140% SOC, the resistance remains at a very high level, indicating that the cell is irreversibly damaged. At –25 °C, 0 °C, and 25 °C, the onset of the dramatic resistance rise is delayed until 130% SOC. This result confirms that high ambient temperatures considerably reduce the overcharge tolerance of a ternary EV battery.
| Ambient temperature (°C) | Internal resistance change at 110% SOC (mΩ) | Internal resistance change at 120% SOC (mΩ) | Internal resistance change at 130% SOC (mΩ) | Internal resistance change at 140% SOC (mΩ) |
| –25 | 0.007 | 0.011 | 443.08 | 468.08 |
| 0 | 0.001 | 0.001 | 363.576 | 382.179 |
| 25 | 0.001 | 0.001 | 352.076 | 393.777 |
| 50 | 0.003 | 694.676 | 481.478 | 442.478 |
| 75 | 0.003 | 911.777 | 573.276 | 521.577 |
The impedance spectra of the ternary cells were measured using electrochemical impedance spectroscopy after the overcharge events. The equivalent circuit used for quantitative fitting consisted of an electrolyte resistance \(R_s\), a charge-transfer resistance \(R_{ct}\), and constant-phase elements representing the double-layer capacitance and the solid-electrolyte interphase capacitance. The fitting results indicate that \(R_s\) rises as the overcharge SOC increases because of current-collector corrosion and electrolyte decomposition. The charge-transfer resistance \(R_{ct}\), on the other hand, shows only a moderate increase with SOC, which implies that the electrode structure remains electrically percolating until the very late stages of overcharge. When the overcharge was performed at 25 °C, the smallest \(R_s\) value was found among all tested temperatures. This means that the corrosion of the aluminum current collector and the decomposition of the electrolyte are least severe at room temperature. Deviating toward either low or high temperature increases the ohmic impedance and therefore amplifies the degradation of the ternary EV battery during overcharge.
After the overcharge tests, the cells were disassembled in an argon-filled glovebox. The scanning electron microscopy images of the positive electrodes show that the secondary particles, which originally have a well-defined spherical shape, become increasingly fractured and separated from each other as the overcharge depth increases. The cracks arise because the delithiation-induced lattice contraction and the oxidative decomposition of the polyvinylidene fluoride binder destroy the integrity of the electrode network. The separation between particles becomes more pronounced at 130% and 140% SOC. At the same overcharge depth, the particle damage becomes more serious as the environmental temperature deviates from 25 °C. This indicates that room temperature is the optimum condition for preserving the structural integrity of the positive electrode during an overcharge fault. The presence of irregular white protrusions on the surface of the positive electrode is evidence for the decomposition of the binder and the formation of gas bubbles trapped under the electrode surface.
The scanning electron microscopy images of the negative electrode show that lithium-metal particles already exist on the graphite surface at 100% SOC, which is a common feature of the final stages of a standard charge because the anode is nearly saturated. When the cell is overcharged, these lithium particles grow and agglomerate into larger clusters. At the same time, the gaps between graphite particles increase, meaning that the electrode matrix loses contact and the charge-transfer resistance increases. At 50 °C, the faster kinetics encourages more severe lithium agglomeration, while at 75 °C the enhanced adhesion between lithium and graphite partially suppresses the detachment. At –25 °C, the extremely low diffusivity of lithium ions in the electrolyte hinders the lithium-plating process, so the anode degradation is less severe than one might expect from the large overpotential. Overall, the evolution of the negative electrode structure is a direct indicator of the health state of an EV battery after electrical abuse.
The X-ray diffraction patterns collected from the positive and negative electrodes of the ternary cells provide further insight. The positive-electrode diffraction pattern of cells overcharged at room temperature exhibits no observable peak shifts with SOC, suggesting that the layered \(R\bar{3}m\) structure of Li(NiCoMn)O₂ is relatively stable. At 50 °C and 75 °C, however, the (003) peak shifts toward higher 2θ values at 130% and 140% SOC, indicating a contraction of the lattice along the \(c\)-axis and a partial collapse of the layered structure. For the negative electrodes, the intensity of the LiC₁₂ phase decreases as the SOC rises from 110% to 130%, which is attributed to the consumption of active lithium by the electrolyte and to electrode damage. At 140% SOC, a slight recovery of the LiC₁₂ peak is seen, reflecting the increased lithium content in the remaining graphite regions after lithium plating. These crystallographic changes are less severe at 25 °C than at any other temperature. Therefore, in terms of material degradation, 25 °C is the most benign ambient temperature for the overcharge of the 26650 ternary EV battery.
4. Thermal Runaway Performance of a 26650 Ternary Lithium-Ion EV Battery during Discharge
In the final part of my work, I investigated the thermal runaway behavior of the same 26650 ternary lithium-ion cell when it was simultaneously discharged and externally heated. This scenario simulates a realistic EV battery failure in which a cell is in use, for example during vehicle driving, while an adjacent overheated component or a nearby fire supplies an external heat flux. The cell was first fully charged to 100% SOC, then placed inside the temperature-controlled chamber and discharged at a 1 C rate. After 5 minutes of discharge, the heating film, which was powered at a constant 120 W, was switched on. The cell temperature and voltage were recorded until the thermal runaway was complete. The entire process could be divided into four characteristic stages based on the voltage and temperature transitions: the preheating stage, the ventilation stage, the gas-evolution stage, and the thermal-runaway stage. The names refer to the phenomenological events observed in my experiments.
In the first stage, the cell voltage exhibits a slight increase as the temperature rises because the electrolyte viscosity decreases and the internal resistance drops. At the same time, the external heating continuously raises the cell temperature, eventually causing the decomposition of the solid-electrolyte interphase. The first sign of imminent failure is a sudden voltage drop, which I call the voltage-dump point. The voltage-dump point arises from the partial destruction of the anode–electrolyte interface and from the onset of intense reactions between the graphite anode and the electrolyte. The second stage starts after the voltage has returned to zero or to a very small value, although by that time the cell is already internally damaged. The internal pressure increases because of the production of gaseous species from electrolyte decomposition. When the internal pressure exceeds the burst pressure of the safety valve, the valve opens, and the cell begins to release white smoke. I record the valve-opening temperature as \(T_{\mathrm{SVO}}\). The third stage lasts from \(T_{\mathrm{SVO}}\) to the thermal-runaway onset temperature \(T_{\mathrm{onset}}\). In this stage, large volumes of electrolyte vapor and flammable hydrocarbon gases are ejected through the open valve. The duration of this stage is drastically shortened at higher ambient temperatures; at 75 °C the third stage almost disappears and the cell enters thermal runaway immediately after the vent opens.
The thermal runaway stage itself is characterized by an abrupt acceleration of temperature, which reaches values above 450 °C for all tested ambient temperatures in my experiments. The heat released during this stage is the sum of the heat from the internal short circuit, the reactions between the positive electrode and the electrolyte, the reactions between the negative electrode and the electrolyte, and the decomposition of the metal oxides inside the cathode. Schematically, the total thermal energy can be described by
$$Q_{TR} = Q_{SEI} + Q_{neg} + Q_{pos} + Q_{ISC}$$
where \(Q_{SEI}\) is the heat of SEI decomposition, \(Q_{neg}\) is the heat from anode–electrolyte reactions, \(Q_{pos}\) is the heat from cathode–electrolyte reactions, and \(Q_{ISC}\) is the Joule heat generated by the internal short circuit. The dominant part of \(Q_{TR}\) is usually associated with the reactions that involve the cathode because the cathode releases oxygen at high temperature. The released oxygen then oxidizes the electrolyte and the anode, further increasing the amount of heat and the amount of flammable gas. The overall thermal runaway process of a ternary EV battery can therefore be viewed as a self-sustaining oxidation cycle.
One of the most important observations from the thermal-runaway experiments is that, in every test, the thermal runaway did not occur until at least 89 seconds after the voltage-dump point. This consistent time interval makes the voltage-dump point a suitable early-warning feature for the thermal runaway of a ternary cell. In sharp contrast, the time interval between the opening of the safety valve and the thermal runaway is much shorter; at 25 °C and 50 °C it is only about 1 s, and at 75 °C it is virtually zero. Thus, if one waits for the safety valve to open before raising an alarm, the available response time is insufficient to protect a vehicle occupant. The voltage-dump signal, by contrast, appears much earlier in the chain of events and gives the EV battery management system enough time to trigger an alert and to disconnect the affected module.
Table 7 summarizes representative values of the voltage-dump time, the temperature at the voltage-dump point, the time from the voltage-dump point to thermal runaway, and the maximum temperature reached during thermal runaway. I should note that values depend slightly on the cell-to-cell variance, but the trends are consistent.
| Ambient temperature (°C) | Time of voltage-dump point (s) | Voltage-dump temperature (°C) | Time from voltage-dump to thermal runaway (s) | Maximum thermal-runaway temperature (°C) |
| –25 | ≈ 1220 | ≈ 320 | ≈ 165 | ≈ 462 |
| 0 | ≈ 1000 | ≈ 330 | ≈ 127 | ≈ 471 |
| 25 | ≈ 830 | ≈ 350 | ≈ 92 | ≈ 491 |
| 50 | ≈ 790 | ≈ 340 | ≈ 105 | ≈ 517 |
| 75 | ≈ 640 | ≈ 360 | ≈ 89 | ≈ 545 |
As the ambient temperature increases, the voltage dump happens earlier because the cell reaches the temperature of SEI decomposition sooner. The magnitude of the thermal-runaway temperature also increases monotonically from –25 °C to 75 °C, showing that a cold environment effectively mitigates the severity of thermal runaway. The maximum temperature increase can be more than 500 °C above the ambient temperature in the hot cases. The video recordings reveal another clear monotonic trend: the intensity of the jet flame and the volume of the flame envelope increase with the ambient temperature. At low temperatures, only a small amount of smoke is emitted and the cell remains intact without popping out its cap. At high temperatures, however, the cap is completely ejected, and a long columnar jet flame is formed. This behavior intensifies the fire hazard of an EV battery and also increases the risk of propagating thermal runaway to neighboring cells.
After the thermal runaway tests, I photographed the external appearance of the cells. The cells tested at –25 °C and 0 °C exhibit a relatively mild outer damage: the safety valve is not fully blown away, although black burn marks and gray oxidation regions can be seen on the metal can. In the 25 °C case, the cap is completely ejected and the cell interior is severely charred. The 50 °C and 75 °C cases show the largest degree of material loss, with significant portions of the electrode stack ejected as black powder. These observations indicate that the mechanical energy released by the thermal runaway is larger at high ambient temperature, meaning the danger to the surrounding personnel and components is higher. For the design of EV battery thermal propagation barriers, it is therefore necessary to account for the worst-case scenario in high-temperature climates or in a module whose cooling system has failed.
The post-mortem characterization focused on three types of samples: the ejected particulate matter that was deposited on the test bench, the positive electrode material, and the negative electrode material of the cells after thermal runaway. Scanning electron microscopy images show that the ejected particles from low-temperature tests retain a spherical multi-layer structure typical of graphite fragments, with some cracks on the edges. As the ambient temperature rises, the particles become more irregular, rough, and block-shaped. At 75 °C, almost no spherical morphology remains, indicating a much more intense mechanical and chemical degradation. The same trend is observed for the positive electrode material: the pristine-like secondary particles become fractured and agglomerated after high-temperature thermal runaway. The negative electrode material shows severe exfoliation and stacking of the graphene layers, particularly when the ambient temperature is above room temperature. The SEM observations imply that the thermal runaway of an EV battery at high temperature leaves a more chemically and structurally transformed residue.
Energy-dispersive X-ray spectroscopy of the thermally runaway samples revealed the presence of carbon, oxygen, aluminum, nickel, cobalt, manganese, fluorine, copper, phosphorus, and sulfur, among other elements. Table 8 lists normalized mass ratios of selected elements for the ejected matter, the positive electrode, and the negative electrode at the five ambient temperatures. Although the distribution varies from one location to another, some general trends are evident. Carbon becomes increasingly abundant as the ambient temperature increases, both in the ejected matter and in the electrodes, which indicates more severe carbonization during high-temperature thermal runaway. Oxygen is also present in large amounts, arising from the oxygen released by the cathode and from the electrolyte decomposition products.
| Sample | Ambient temperature (°C) | C (wt.%) | O (wt.%) | Ni (wt.%) | Co (wt.%) | Mn (wt.%) | Al (wt.%) | Cu (wt.%) |
| Ejected matter | –25 | 9.65 | 27.43 | 22.78 | 3.36 | 33.26 | 1.26 | 0.15 |
| Ejected matter | 0 | 13.39 | 40.18 | 14.54 | 2.17 | 2.41 | 25.88 | 0.36 |
| Ejected matter | 25 | 79.53 | 12.82 | 3.88 | 0.50 | 0.52 | 0.50 | 0.54 |
| Ejected matter | 50 | 98.58 | 0.00 | 0.41 | 0.05 | 0.29 | 0.18 | 0.07 |
| Ejected matter | 75 | 78.25 | 18.03 | 1.60 | 0.25 | 0.71 | 0.01 | 0.48 |
| Positive electrode | –25 | 14.80 | 38.40 | 20.11 | 3.30 | 9.24 | 12.45 | 0.13 |
| Positive electrode | 0 | 14.86 | 28.79 | 15.25 | 3.04 | 35.64 | 0.21 | 0.54 |
| Positive electrode | 25 | 15.51 | 48.41 | 7.61 | 1.16 | 3.99 | 22.21 | 0.19 |
| Positive electrode | 50 | 47.95 | 25.75 | 14.75 | 2.40 | 7.10 | 0.95 | 0.36 |
| Positive electrode | 75 | 67.77 | 28.47 | 0.66 | 0.09 | 0.36 | 1.58 | 0.04 |
| Negative electrode | –25 | 30.79 | 56.77 | 6.32 | 0.87 | 0.91 | 2.21 | 0.51 |
| Negative electrode | 0 | 52.54 | 45.31 | 0.25 | 0.00 | 0.19 | 0.03 | 0.48 |
| Negative electrode | 25 | 37.92 | 33.96 | 16.78 | 2.67 | 4.29 | 3.20 | 0.12 |
| Negative electrode | 50 | 77.82 | 16.84 | 0.21 | 0.00 | 0.09 | 0.04 | 3.41 |
| Negative electrode | 75 | 86.63 | 11.73 | 0.33 | 0.04 | 0.01 | 0.02 | 0.47 |
The existence of copper in the ejected matter and in the electrodes confirms that the copper current collector was dissolved in the high-voltage environment generated during thermal runaway. The relevant redox reactions can be summarized as follows. At the negative electrode, copper is oxidized to Cu⁺ and Cu²⁺ when the local potential becomes sufficiently high:
$$\mathrm{Cu} \rightarrow \mathrm{Cu^+} + e^-$$
$$\mathrm{Cu^+} \rightarrow \mathrm{Cu^{2+}} + e^-$$
Then, at the positive electrode, the Cu²⁺ ions are reduced back to Cu⁺ and metallic copper:
$$\mathrm{Cu^{2+}} + e^- \rightarrow \mathrm{Cu^+}$$
$$\mathrm{Cu^+} + e^- \rightarrow \mathrm{Cu}$$
Thus, copper shuttling leads to the deposition of metallic copper in unwanted regions and to the formation of copper-containing compounds on the electrode surfaces. The copper detection is therefore a footprint of an internal short-circuit event and of severe overpotential inside the cell.
The X-ray diffraction analysis of the post-thermal-runaway samples shows a strong carbon peak at 2θ ≈ 26.3° in all samples, which corresponds to graphitic carbon. The intensity of the carbon peak is higher in samples originating from high-temperature tests, again supporting the claim that carbonization is more severe at high ambient temperatures. Diffraction peaks corresponding to NiO, LiNiO₂, CoO, and LiF are detected in the ejected matter, confirming that the positive electrode material was oxidized and partially transported out of the cell by the gas jet. The presence of LiF is attributed to the reaction between the electrolyte salt LiPF₆ and water traces or to the decomposition of the lithium-containing SEI component. In the negative electrode samples, the diffraction peaks become less numerous compared with the positive electrode, indicating that most of the transition-metal oxides had already undergone redox transformation and that the graphite framework was the dominant remaining phase.
5. Synthesis and Implications for EV Battery Thermal Safety
By combining the results from the lithium iron phosphate overcharge tests, the 26650 ternary cathode overcharge tests, and the 26650 ternary cell discharge-with-external-heating tests, I obtained a comprehensive picture of how ambient temperature affects the failure and thermal runaway of modern EV battery cells. The first major conclusion is that low temperature reduces the probability of immediate thermal runaway but increases the total gas release and the time required for the cell to reach a benign safe state. This behavior is favorable for the escape of passengers in an EV battery fire at cold climate, but it also creates a longer-lasting source of flammable gas that can be ignited by any spark or hot surface. The second major conclusion is that room-temperature overcharging is not always the safest condition. For the ternary cell, the voltage-jump time at 25 °C is shorter than at –25 °C and 0 °C, meaning that a protection system has less time to react when the overcharge starts at room temperature. The third major conclusion is that high ambient temperatures unambiguously increase the severity of thermal runaway: the flame is stronger, the maximum temperature is higher, the ejected mass is larger, and the structural damage to the cell container is more severe.
The differences between the lithium iron phosphate and the ternary chemistry are also important. The 23 Ah lithium iron phosphate cells did not enter thermal runaway in my overcharge tests even at 75 °C, although their maximum temperature reached roughly 223 °C. The cells opened their safety valves and vented but did not produce a jet flame. The 26650 ternary cells, in contrast, could reach final-stage voltage jumps above 100 V and subsequently underwent violent thermal runaway with intense flame emission. This contrast underlines the higher intrinsic thermal stability of olivine LiFePO₄ compared with the layered ternary oxide cathode. However, the lithium iron phosphate cell still exhibits an abrupt voltage jump and a large temperature rise during overcharge, so a proper EV battery management system should not disregard the overcharge risk of this chemistry.
From the perspective of early detection, my experiments provide quantitative warning time windows. For the lithium iron phosphate overcharge, the time between the voltage surge point and the maximum-temperature point was always longer than 100 s, so the voltage surge can serve as a robust warning signal. For the ternary overcharge, this interval is much shorter and less predictable. In the discharge-plus-heating scenario, the voltage-dump point occurs at least 89 s before thermal runaway, which is sufficiently long for a real-time protection algorithm to cut the contactors and to trigger an alarm. By contrast, the safety-valve-opening point is too late for the ternary EV battery, because thermal runaway follows within about one second at moderate temperatures and instantaneously at high temperatures. Thus, I recommend that EV battery monitoring systems rely primarily on voltage-dump detection and on the rate-of-change of temperature rather than on gas-pressure switches or safety-valve indicators.
The material characterization results also provide useful clues for the design of more robust EV battery electrodes. The positive electrode binder degradation that I observed at high overcharge states can be mitigated by using more oxidation-resistant binders or by adding surface coatings that stabilize the cathode particle boundaries. The negative electrode lithium plating can be reduced by improving the reversibility of the graphite anode and by controlling the charging current as a function of the battery temperature. The dissolution of the copper current collector during the final stage of thermal runaway could be suppressed by implementing a careful over-discharge protection and by selecting copper alloys with higher corrosion resistance, although the extreme conditions of thermal runaway would be difficult to handle with material selection alone.
My findings also demonstrate that the ambient temperature of an EV battery is a central parameter in the failure chain. In a practical electric vehicle, the battery pack is exposed to temperatures that can easily vary from –25 °C in a cold-climate winter to 75 °C in a poorly cooled pack that sits under strong solar irradiation. The data from my experiments indicate that the operating window of maximum safety is relatively narrow, and that the safety margins shrink when the cell is pushed to high SOC and high rate simultaneously. The battery thermal management system in an electric vehicle should therefore not be designed only to prolong the life of the EV battery; it should also ensure that the temperature of every cell remains inside the safe region where the overcharge and thermal-runaway responses are the least dangerous. In particular, the cooling strategy must be sufficiently strong to prevent the surface temperature from approaching the critical thermal-runaway trigger temperature that I identified in the tests under all plausible chain-of-abuse scenarios.
To quantitatively describe the onset of thermal runaway, I used the following criterion during post-processing. The heat generation rate from the side reactions can be represented by the Arrhenius-like expression
$$\dot{Q}_{side} = A \exp\left(-\frac{E_a}{k_B T}\right) f(\mathrm{SOC})$$
where \(A\) is a pre-exponential factor, \(E_a\) is the apparent activation energy, \(k_B\) is the Boltzmann constant, \(T\) is the absolute cell temperature, and \(f(\mathrm{SOC})\) is a function that depends on the state of charge and on the extent of lithium loss inside the anode. Because \(T\) appears in the exponential term, a small increase in the ambient temperature significantly decreases the time required for the side-reaction heat to exceed the heat-loss rate. This reasoning explains why the high-temperature tests in my work always led to earlier voltage dumps and more violent thermal runaway, and why the low-temperature tests suppressed the progression from the venting phase to the full thermal-runaway phase.
6. Conclusions
Based on my systematic experimental investigation of the overcharge failure and thermal runaway of lithium-ion power batteries, I can draw the following main conclusions.
First, the overcharge of a 23 Ah lithium iron phosphate EV battery at 1 C shows that the number of voltage stages decreases with increasing ambient temperature. The low-temperature environment increases the gas release duration and the total gas volume, while the high-temperature environment accelerates the voltage surge, increases the maximum temperature, and enlarges the probability of an eventual thermal runaway. A voltage surge can be used as an early-warning signal because the time to the highest temperature is always longer than 100 seconds.
Second, the high-rate overcharge test of the 26650 ternary EV battery at 3 C reveals that at –25 °C the voltage curve has six stages, at 0 °C and 25 °C it has five stages, while at 50 °C and 75 °C it is reduced to two stages. Low ambient temperatures lead to larger maximum temperature differences. Room temperature is the condition where the voltage surge appears earliest and where the internal damage to the electrode is the smallest. At 25 °C, the corrosion of the current collector and the decomposition of the electrolyte are minimized. In contrast, temperatures that deviate from 25 °C increase the degradation of the electrode materials and the internal resistance of the cell.
Third, in the thermal-runaway test of the 26650 ternary cell during a 1 C discharge with 120 W external heating, the whole course can be divided into four stages. The thermal-runaway onset always occurs at least 89 seconds after the voltage-dump point. Thus, the voltage-dump point is a valid and reliable pre-warning signal for the ternary EV battery. The safety-valve opening point is not suitable as a standalone early warning because of the nearly immediate thermal runaway transition at high ambient temperature. The severity of the thermal runaway, the flame height, the maximum temperature, and the post-mortem material damage all increase with the ambient temperature. The ejected matter and the electrode materials contain C, O, Al, Ni, Co, Mn, F, Cu and other elements, reflecting chemical reactions of the current collector, the cathode, the electrolyte, and the SEI layer during the thermal runaway.
Finally, the overarching conclusion is that ambient temperature must be treated as a first-order design variable in the safety assessment of any EV battery. A single voltage or temperature threshold cannot be universally applied; instead, an adaptive EV battery management system should account for the current cell temperature, the charging current, the state of health, and the SOC to estimate the remaining time before irreversible damage. My experimental database of voltage-stage counts, venting durations, resistance surges, and critical transition times provides a useful reference for the calibration of such adaptive safety algorithms. Further work should extend the current results to modules and packs, include mechanical abuse as a third trigger, and explore the effect of cycling-induced aging on the temperature-dependent overcharge response. In this way, the overall safety of EV battery systems can be improved across a wide range of realistic operating and abuse conditions.