The widespread adoption of lithium-ion batteries, particularly LiFePO4 (lithium iron phosphate) batteries, in electric vehicles and energy storage systems is driven by their high energy density and long cycle life. However, the persistent challenge of thermal runaway (TR) safety remains a critical bottleneck. During TR, intense internal exothermic reactions and electrolyte decomposition generate significant volumes of gases and vapors. The release sequence, concentration dynamics, and compositional characteristics of these products are intrinsically linked to the thermodynamic state of the battery. Traditional Battery Management Systems (BMS), which rely primarily on voltage and temperature monitoring, exhibit limited capability for early TR warning. Crucially, the dynamic behavior of electrolyte vapor and gaseous by-products, which serve as direct indicators of internal failure mechanisms, has not been fully exploited for predictive safety. This study investigates the spatiotemporal evolution of these critical parameters to establish a robust foundation for advanced early warning systems.
Recent research has begun to explore the potential of gas monitoring. Studies have identified key gas species like H₂, CO, CO₂, and hydrocarbons using techniques such as Fourier Transform Infrared Spectroscopy (FTIR) and mass spectrometry. Notably, hydrogen (H₂) has been flagged as an exceptionally early indicator due to its generation from reactions involving metallic lithium. Furthermore, the release of electrolyte vapor, constituting over 90% of the total ejecta mass in some cases, represents a major hazard. While progress has been made, comprehensive studies mapping the full temporal evolution of these species under varied abuse conditions, especially for LiFePO4 chemistry, and their quantitative correlation with other state-of-health parameters are still lacking. This work aims to address this gap by constructing a detailed timeline of gas and vapor release and linking it to the internal degradation state of the LiFePO4 battery.
Experimental Platform for Spatial Anomaly Parameter Monitoring
To investigate the evolution of fault parameters in LiFePO4 batteries, a specialized Spatial Anomaly Parameter Monitoring Platform was constructed. The core of the platform is a transparent acrylic glovebox purged with argon to create an inert atmosphere, preventing external reactions with the cell components after the vent is opened. The platform integrates multiple sensing modalities to capture a holistic view of the TR process.
The electrical sensing suite includes a dynamic impedance analyzer, employing a “four-wire” method to inject a single-frequency sinusoidal current and measure the voltage response, enabling real-time calculation of impedance ($Z = \frac{V}{I}$). Temperature is monitored internally using a K-type thermocouple (WRNK-191, 0–1100 °C) and externally via a high-resolution infrared thermal camera (640×480 pixels). For gas detection, an array of fixed sensors is deployed: electrochemical sensors for H₂ (0-1000 ppm), CO (0-200 ppm), and VOC (Volatile Organic Compounds, 0-200 ppm), and an infrared sensor for CO₂ (0-20,000 ppm). VOC detection primarily targets electrolyte solvent vapors and other organic by-products. Strain evolution within the cell is measured using a Fiber Bragg Grating (FBG) sensor, with temperature compensation provided by the co-located thermocouple. Visual documentation is achieved with a high-definition camera. The system is controlled by a multi-channel data logger and connected to a battery cycler for applying overcharge protocols. For low-temperature tests, a Peltier cooler is attached to the cell surface.
Temporal Sequence and Generation Rates of Gases During Initial TR
Overcharge-Induced TR at Room Temperature
A 27 Ah prismatic LiFePO4 battery was discharged to 0% SOC and placed in the argon-filled chamber. After manually opening the safety vent, the battery was overcharged at a 0.5C rate until TR initiation, defined by a surface temperature rise rate exceeding 1 °C/s.
The voltage and temperature profiles are shown in Figure 1. During normal charging, both parameters increased gradually. Upon entering the overcharge regime (after ~7200 s), voltage plateaued around 5.3 V while temperature rose linearly due to increased joule heating and onset of parasitic reactions. In the final TR stage, both voltage and temperature spiked dramatically before charge termination.
The gas concentration profiles (Figure 2) reveal a clear sequence. Electrolyte vapor (VOC) was the first to be detected at approximately 7410 s, attributable to the evaporation of low-boiling-point carbonate solvents (e.g., dimethyl carbonate, DMC). This was followed by a rapid rise in H₂ concentration at ~7510 s, generated from the reduction of solvent or moisture by lithium plating. Finally, CO and CO₂ appeared nearly simultaneously around 7550 s, resulting from the severe oxidative decomposition of carbonate solvents at the delithiated positive electrode. The generation sequence was thus: VOC > H₂ > CO ≈ CO₂.
The average gas generation rates were calculated from the linear slope of the concentration curves during the main production phase:
$$R_{gas} = \frac{\Delta C}{\Delta t}$$
where $R_{gas}$ is the generation rate (ppm/s), $\Delta C$ is the concentration change (ppm), and $\Delta t$ is the time interval (s). The rates are summarized in Table 1.
| Gas Species | Onset Time (s) | Avg. Generation Rate (ppm/s) |
|---|---|---|
| H₂ | 7510 | 0.887 |
| CO | 7550 | 0.201 |
| VOC | 7410 | 0.217 |
| CO₂ | 7550 | 0.806 |
Table 1: Gas generation characteristics during room temperature overcharge of a LiFePO4 battery.
H₂ exhibited the highest generation rate, making it a prominent and fast-evolving signal.
Overcharge-Induced TR Under Low-Temperature Conditions
To study the impact of temperature on gas generation, the same test was repeated with active cooling applied to the 27 Ah LiFePO4 battery surface via a Peltier element. The voltage profile was similar, but the temperature rise was suppressed initially.
The gas evolution sequence under cooling was altered (Figure 3). H₂ was the first gas detected at ~7462 s, even earlier than in the room temperature case. VOC appeared shortly after at ~7468 s. CO and CO₂ generation was significantly delayed, appearing at ~7484 s and ~7510 s, respectively. The sequence under cooling was: H₂ > VOC > CO > CO₂.
The calculated generation rates under low-temperature conditions are shown in Table 2. A comparison with Table 1 reveals critical insights.
| Gas Species | Onset Time (s) | Avg. Generation Rate (ppm/s) | Rate Change vs. Room Temp. |
|---|---|---|---|
| H₂ | 7462 | 0.856 | -3.5% |
| CO | 7484 | 0.165 | -17.9% |
| VOC | 7468 | 0.256 | +18.0% |
| CO₂ | 7510 | 0.268 | -66.7% |
Table 2: Gas generation characteristics during low-temperature overcharge of a LiFePO4 battery.
The data shows that low temperature had a minimal effect on H₂ generation rate (-3.5%) but severely inhibited CO₂ production (-66.7%). The earlier H₂ onset suggests that the reactions leading to its formation are less thermally activated or are initiated at localized hot spots. The suppression of CO/CO₂ is consistent with the retardation of electrolyte oxidation processes at the cathode. The increase in VOC rate could be due to altered solvent evaporation dynamics or changed decomposition pathways under cooling.
Discussion: Optimal Gas Signal for Early Warning
Based on the quantitative results, H₂ emerges as the superior signal for early TR warning in LiFePO4 battery systems. This conclusion is supported by several factors evident in both experimental conditions for the LiFePO4 battery. First, H₂ possesses excellent temporal precedence, appearing either as the first or second gas species. Second, it exhibits the highest generation rate, leading to a rapid concentration build-up that is easier to detect. Third, its generation is remarkably robust to ambient temperature variations, as shown by the minor rate change under cooling. In contrast, CO₂ is abundant in air, complicating baseline establishment. VOC sensors, often based on photoionization, are costly and their signal in early stages is weak due to limited evaporation at lower internal temperatures. Therefore, monitoring H₂ concentration provides the most reliable, sensitive, and cost-effective means for the early detection of thermal runaway in LiFePO4 battery packs.
Electrolyte Vapor Evolution and Multi-Parameter Mapping During Late-Stage TR
To investigate the later stages of TR, a more destructive test was conducted on a 13 Ah LiFePO4 battery. The cell was de-cased and overcharged at 1C within the argon chamber until full TR, with comprehensive monitoring of parameters including voltage, impedance, internal temperature, strain, gas concentrations, and visual ejecta.
Evolution of Intrinsic Battery Degradation Parameters
The synchronized data reveals the progression of internal failure in the LiFePO4 battery:
1. Voltage and Temperature: During normal charge (0-3600 s), voltage and temperature increased gradually. Upon overcharge, the voltage rose steeply to ~4.95 V before suddenly collapsing due to internal short circuit (ISC), a hallmark of TR. Internal temperature exhibited a similar trend, reaching 116.8 °C at charge termination and peaking at 145.7 °C due to continuing exothermic reactions.
2. Impedance and Strain: The dynamic impedance ($Z$) initially decreased slightly due to rising temperature, then increased sharply after ~4176 s ($\frac{dZ}{dt} > 0$) as gas generation and electrode delamination increased internal resistance. This was followed by a catastrophic drop as the ISC established ($\frac{dZ}{dt} << 0$). The FBG-measured strain remained stable initially, then showed two major peaks corresponding to intense gas generation events, indicating severe mechanical deformation before relaxation due to venting.
The co-evolution suggests a coupled degradation mechanism: Overcharge leads to lithium plating, SEI breakdown, and heat accumulation. This triggers gas-producing reactions (releasing H₂, etc.), increasing internal pressure and strain. The heat and pressure eventually cause separator collapse (ISC), leading to voltage drop, impedance collapse, and temperature spike.
Spatiotemporal Characteristics of Electrolyte Vapor Ejection
Visual observation provided critical insights into the electrolyte vapor release. The “white smoke” – primarily a mist of condensed electrolyte vapor and decomposition products – began emanating from the cell header at t = 4666 s. The ejection intensified dramatically at t = 4716 s. Charging was stopped at t = 4726 s, but vapor continued to be expelled until approximately t = 4758 s. Therefore, the main episode of electrolyte vapor release lasted for about 92 seconds, with the most intense period spanning roughly 42 seconds.
This vapor release timeline corresponds directly to the critical phase of the LiFePO4 battery’s failure. Its onset coincides with the rapid rise in internal temperature (>100°C) and the sharp increase in VOC sensor readings. The venting process is the primary mechanism for pressure relief but also represents the moment when highly flammable aerosol is released into the battery enclosure, presenting a major ignition risk.
Integrated Early Warning Strategy and Conclusion
This comprehensive study elucidates the spatiotemporal evolution of gases and vapors during thermal runaway in LiFePO4 batteries. The key findings are:
- Gas Generation Sequence is Condition-Dependent: For a LiFePO4 battery at room temperature: VOC > H₂ > CO ≈ CO₂. Under low-temperature cooling: H₂ > VOC > CO > CO₂.
- H₂ is the Optimal Early Warning Signal: It combines early appearance, high generation rate (0.887 ppm/s at RT), and robustness to operating temperature (rate change -3.5% under cooling).
- Temperature Modulates Reaction Pathways: Low temperature severely suppresses CO₂ generation (-66.7% rate change) but has a lesser effect on H₂, indicating different activation energies for the underlying reactions in the LiFePO4 battery.
- Late-Stage Vapor Release is a Discrete Event: The violent ejection of electrolyte vapor occurs over approximately one minute, aligning with the peak of internal pressure and temperature, providing a clear visual and VOC-sensor-detectable signature of catastrophic failure.
- Multi-Parameter Correlation: The evolution of gas/vapor parameters is quantitatively linked to the degradation of electrical (voltage, impedance) and mechanical (strain) states, enabling cross-validated failure diagnostics.
Based on these results, a multi-tiered early warning strategy for LiFePO4 battery energy storage systems is proposed:
- Tier 1 (Early Warning): Continuous monitoring of H₂ concentration. A sustained rise or a rate ($\frac{d[H_2]}{dt}$) exceeding a threshold triggers a pre-alarm, allowing for preventive measures like controlled discharge or isolation.
- Tier 2 (Critical Warning): Detection of a rapid increase in VOC levels or the onset of electrolyte vapor (via optical or advanced VOC sensors), coinciding with abnormal temperature rise rate ($\frac{dT}{dt} > k$). This signals imminent venting and should trigger immediate system shutdown and fire suppression readiness.
- Tier 3 (Failure Confirmation): A sudden voltage drop concurrent with a spike in temperature and internal pressure/strain confirms that thermal runaway is in progress.
The implementation of such a strategy, particularly the non-contact, spatially distributable detection of H₂ and electrolyte vapor, significantly enhances the safety of LiFePO4 battery-based储能 systems by providing earlier and more reliable warnings than traditional voltage/temperature monitoring alone. This work provides a fundamental dataset and analysis framework for developing advanced hazard mitigation systems for LiFePO4 batteries.