Safety concerns remain a significant technical bottleneck in the widespread adoption of li ion battery technology. Incidents of fire or explosion are fundamentally linked to the complex interplay of internal materials, operating conditions, and manufacturing integrity. Within a confined system like a li ion battery, the accumulation of heat and gas from exothermic chemical reactions can lead to catastrophic failure through thermal runaway. This process poses severe risks, not only from the immediate energy release but also from the emission of large volumes of toxic, flammable gases. A profound understanding of the gas generation characteristics, particularly during the initial stages before catastrophic venting or ignition, is therefore critical for developing effective early warning systems and mitigation strategies. This research focuses on characterizing the initial flue gas release during the adiabatic thermal runaway of a commercial square-format li ion battery. By employing a thermal abuse methodology within an accelerating rate calorimeter (ARC), we analyze the temporal evolution of temperature and gas composition. The primary objectives are to elucidate the gas release kinetics, identify key characteristic gases suitable for early detection, and propose quantitative thresholds for an effective safety warning system in applications involving li ion battery packs and energy storage systems.
The experimental investigation was conducted using a commercial square-format li ion battery with a nominal capacity of 92 Ah. The cathode active material was LiNi0.6Co0.2Mn0.2O2 (NCM622), the anode was graphite, and the electrolyte consisted of 1.1 mol/L LiPF6 in a mixture of ethylene carbonate (EC) and dimethyl carbonate (DMC) (1:1 by volume). Prior to thermal abuse testing, the li ion battery was charged to 100% State of Charge (SOC) using a standard constant current-constant voltage (CC-CV) protocol. The core experimental apparatus was an Accelerating Rate Calorimeter (ARC), which provides a near-adiabatic environment crucial for simulating the worst-case heat accumulation scenario within a battery pack. The li ion battery sample was instrumented with K-type thermocouples to monitor surface temperature and suspended inside the ARC chamber. The test followed the standard “Heat-Wait-Seek” procedure. The initial temperature was set to 40°C, with a stepwise heating rate. The onset of self-heating was defined as a temperature rise rate exceeding 0.02 °C/min. Upon detection, the ARC switched to adiabatic mode, tracking the sample’s self-heating without external thermal interference. Gas sampling was performed at critical stages: before any detectable self-heating, immediately after the safety vent opened, and at subsequent timed intervals (e.g., every 5 minutes) during the runaway process until flame or intense smoke was observed. The collected gas samples were subsequently analyzed using Gas Chromatography-Mass Spectrometry (GC-MS) to determine the precise volumetric composition of the emitted species.
The adiabatic thermal runaway process of the li ion battery was characterized by several key temperature milestones, as shown in the derived temperature profile. The entire process can be segmented into distinct phases based on the heating rate. The initial self-heating onset temperature (θ0) was identified at approximately 102.7°C, corresponding to a heating rate threshold of 0.02 °C/min. This marks the beginning of minor exothermic side reactions. A critical point was the safety vent opening (θvent), recorded at approximately 156.7°C, indicating significant internal pressure build-up from gas generation. The thermal runaway onset temperature (θ1), defined by a rapid heating rate exceeding 1.00 °C/min, occurred at about 194.6°C. This point signifies the transition into uncontrolled, highly exothermic reactions. Subsequently, the cell voltage plummeted to zero at ~204.4°C. The maximum surface temperature (θmax) recorded was 321°C, while the ARC furnace thermocouple reported 296.4°C. The temperature rise from self-heating onset (θ0) to venting (θvent) was approximately 54°C. The temperature-time curve during the self-heating phase (prior to θ1) exhibited a strong linear correlation, which can be described by the fitted equation:
$$ y = 0.2152x + 104.84 $$
where \( y \) is the temperature in °C and \( x \) is the time in minutes relative to θ0. This yields an average self-heating rate of 0.2152 °C/min during this phase. It is essential to account for thermal gradients in practical applications; a large-format li ion battery or module can have an internal-to-external temperature differential of 15-30°C. Applying a conservative 30°C differential suggests that when the external casing of a battery module reaches 72.7°C, the internal core temperature may already be at the self-heating onset (θ0 ~102.7°C). This has direct implications for setting temperature-based alarm thresholds in battery management systems.
The gas analysis revealed a complex mixture of species emitted during the thermal runaway of the li ion battery. The primary gaseous components were carbon dioxide (CO2), hydrogen (H2), carbon monoxide (CO), and various light hydrocarbons (CxHy) including methane (CH4), ethylene (C2H4), ethane (C2H6), propylene (C3H6), butane (n-C4H10), and 1,3-butadiene (C4H6). The volumetric composition evolved dynamically over time. A summary of the gas composition at selected intervals after venting is presented in the table below.
| Time Post-Vent | H2 (φ%) | CO2 (φ%) | CO (φ%) | CH4 (φ%) | C2H4 (φ%) |
|---|---|---|---|---|---|
| 0 min | 0.03 | 0.41 | 0.08 | 0.05 | 0.03 |
| 5 min | 0.04 | 0.44 | 0.08 | 0.05 | 0.04 |
| 10 min | 0.04 | 0.46 | 0.09 | 0.05 | 0.04 |
| 15 min | 0.05 | 0.49 | 0.09 | 0.05 | 0.04 |
| 20 min | 0.06 | 0.53 | 0.09 | 0.05 | 0.04 |
Notably, CO2 was present in measurable quantities even before the vent opened (0.07%), indicating early-stage decomposition reactions. However, for early warning purposes, the combustible gases H2 and CO are of paramount interest due to their high flammability and detectability. The temporal evolution of these two key gases, plotted from the point of self-heating onset (θ0), showed distinct and significant trends. Mathematical fitting of the data provided empirical relationships for their release kinetics.
For hydrogen (H2), the volume fraction (\( \phi_{H_2} \)) as a function of time (\( t_{H_2} \)) in minutes from θ0 was found to follow a linear relationship:
$$ \phi_{H_2} = 0.0014t_{H_2} – 0.85438 $$
(where \( \phi_{H_2} \) is expressed in %).
For carbon monoxide (CO), the release profile was better described by a quadratic relationship:
$$ \phi_{CO} = -3 \times 10^{-5} t_{CO}^2 + 0.0448t_{CO} – 14.507 $$
A critical finding was the difference in the time of first detection for these gases relative to key events. Calculations using the fitted equations show that CO was detectable at a volume fraction of approximately 50 ppm (0.005%) about 14 minutes before H2 reached the same concentration level. The rate of increase in volume fraction was also markedly different. The time taken for the concentration to rise from the detection threshold to 0.02% (200 ppm) was about 6.1 minutes for CO and 13.3 minutes for H2. This corresponds to average volume fraction increase rates of approximately 32.7 ppm/min for CO and 15 ppm/min for H2 during these respective intervals. The entire process from safety vent opening to full thermal runaway lasted approximately 27 minutes, underscoring that an effective early warning system must operate on a timescale of minutes to allow for preventive intervention.
Based on the experimental gas release kinetics and considering practical safety standards for combustible gas detectors, a dual-threshold early warning strategy for a li ion battery system is proposed. The primary goal is to provide a timely alert well before the gas concentration reaches the Lower Explosive Limit (LEL), allowing for countermeasures such as enhanced cooling, load shedding, or system isolation. Reference standards such as T/CEC 373 and T/JFPA 0008 suggest alarm threshold settings as low as 0.02%-0.05% of the LEL for CO and 0.05%-0.50% of the LEL for H2. Given the LEL of H2 is 4.0% and for CO is 12.5%, these percentages translate to very low absolute volume fractions.
To ensure a robust and timely warning, we propose the following thresholds based on our li ion battery test data and the principle of early detection:
- First-Level Safety Warning Threshold: A volume fraction of \( 2.5 \times 10^{-3} \% \) (25 ppm) for both CO and H2. Detecting either gas at or above this level should trigger a primary alarm, indicating the initiation of abnormal exothermic activity within a li ion battery cell or module.
- Second-Level Critical (Absolute Limit) Threshold: A volume fraction of \( 2.0 \times 10^{-2} \% \) (200 ppm) for both gases. Reaching this concentration indicates accelerated gas generation and imminence of venting or thermal runaway, necessitating immediate and decisive protective actions.
The time interval between these two thresholds is estimated from the kinetic data to be approximately 5 minutes for CO and 10 minutes for H2. This provides a crucial window for escalating response. For optimal effectiveness, highly sensitive gas sensors should be strategically placed near the potential venting paths of li ion battery modules, such as in the exhaust plenum or directly above cell vents in a pack. The detection of CO, due to its earlier release signature, offers a particularly valuable lead indicator. Integrating this gas-based monitoring with traditional voltage and temperature surveillance in the Battery Management System (BMS) would create a multi-parameter, fault-tolerant safety architecture significantly enhancing the reliability of li ion battery energy storage systems.
In conclusion, this study provides a detailed analysis of the initial flue gas released during the adiabatic thermal runaway of a commercial NCM-based li ion battery. The process is characterized by distinct thermal stages and the progressive emission of a gas mixture dominated by CO2, H2, CO, and hydrocarbons. Kinetic analysis revealed that CO is an earlier indicator than H2, providing a longer lead time for预警. The derived empirical formulas quantitatively describe the gas release profiles. Based on these findings and aligned with industry standards, a practical two-tier gas detection warning system is proposed, utilizing safety and critical thresholds for both CO and H2. Implementing such a gas-based monitoring strategy represents a proactive approach to li ion battery safety, moving beyond reliance on temperature and voltage measurements alone. Ultimately, while engineering solutions like early warning systems are vital, the fundamental path to safer li ion battery technology lies in the development of intrinsically stable materials, such as those pursued in solid-state li ion battery research, and advanced internal sensing capabilities to directly monitor the cell’s electrochemical state.