Thermal Runaway — What Actually Happens Inside a Cell Before It Catches Fire
Five stages. Distinct temperature windows. One cascade that cannot be reversed once it starts. Here is the actual mechanism — not the simplified version.
Table of Contents
ℹ
This is the Expert level of the EVPulse Thermal Runaway series. It assumes working knowledge of lithium-ion electrochemistry, BMS architecture, and pack design. If you need the fundamentals first, start with the Intermediate article.
The Framing Problem
Every thermal runaway explainer says the same thing: the cell gets too hot, reactions accelerate, it catches fire. That framing is not wrong. It is just useless for engineering decisions.
Useless because it collapses five chemically distinct stages — each with its own temperature window, gas signature, heat generation rate, and critically, its own intervention window — into one undifferentiated event called "thermal runaway."
If you are designing a BMS protection strategy, specifying a thermal management system, writing AIS-156 compliance documentation, or investigating a field failure, you need to know which stage you are dealing with. Stages 1 and 2 are recoverable if caught. Stage 3 is borderline. Stages 4 and 5 are not.
By the time you see smoke from a lithium-ion cell, you are already in Stage 4. The intervention window closed somewhere around Stage 2. Everything after that is containment, not prevention.
This article goes through each stage in sequence — the temperature at which it begins, what is chemically happening, what it produces, and what a well-instrumented BMS can detect at that point.
Stage 1 — SEI Decomposition
Onset Temperature
90–120°C
The solid-electrolyte interphase (SEI) is the passivation film that forms on the graphite anode surface during the first few charge cycles. It is essential for normal operation — it allows Li⁺ ions to pass through while preventing direct electrolyte contact with the graphite.
The SEI is metastable. At 90–120°C, the exothermic decomposition begins:
SEI Decomposition
LixC6+SEI→Li2CO3+LiF+heatLi_xC_6 + SEI \rightarrow Li_2CO_3 + LiF + \text{heat}LixC6+SEI→Li2CO3+LiF+heat
The heat generated is modest at this stage — approximately 40–120 J/g of anode material. Not enough to self-sustain on its own. But enough to raise the cell temperature further, which accelerates the decomposition rate, which generates more heat.
This is the first runaway loop. It is slow. And it is the last point at which passive intervention — active cooling ramped to maximum, charge/discharge terminated — has a realistic chance of arresting the cascade.
⚠
Most commercial BMS implementations do not have sufficient temperature sensor density to detect Stage 1 reliably. A cell core experiencing SEI decomposition may be at 110°C while the nearest surface NTC sensor reads 72°C. By the time the surface sensor hits the protection threshold, Stage 2 has already begun.
What the BMS sees at Stage 1:
Cell surface temperature elevated above pack average, rising at a measurable dT/dt
Possible slight voltage deviation from neighbouring cells if decomposing SEI affects impedance
No gas detection signal yet — gas generation is minimal at this stage
Stage 2 — Electrolyte Reaction with Anode
Onset Temperature
120–180°C
As the temperature rises past the SEI decomposition window, the lithiated graphite anode begins reacting directly with the liquid electrolyte. The primary reactions involve ethylene carbonate (EC) and dimethyl carbonate (DMC):
Electrolyte-Anode Reaction
2Li+C3H4O3(EC)→Li2CO3+C2H4↑2Li + C_3H_4O_3 \text{(EC)} \rightarrow Li_2CO_3 + C_2H_4\uparrow2Li+C3H4O3(EC)→Li2CO3+C2H4↑
This produces ethylene gas and heat. Heat generation at this stage is significantly higher — approximately 100–400 J/g — and the reaction rate is strongly temperature-dependent.
💡
The highly lithiated anode (charged state) is far more reactive than a discharged one. This is why thermal runaway risk is highest at high SOC — not just because of voltage, but because the lithium-rich graphite is a more energetic reactant. Always store cells at 30–50% SOC for long-term parking.
Why high SOC is disproportionately dangerous
At full charge, graphite exists as LiC₆ — one lithium atom per six carbon atoms, the most lithiated state. The chemical potential of lithium in LiC₆ is very close to that of metallic lithium (near 0V vs Li/Li⁺). This means the thermodynamic driving force for reaction with electrolyte is at its maximum. A cell at 100% SOC has roughly 3–4× the heat generation potential during anode-electrolyte reaction compared to a cell at 20% SOC.
SEI Decomp. Onset
90–120°C
Anode Reaction Onset
120–180°C
Heat Generation (Stage 2)
100–400 J/g
Gas Produced
C₂H₄ (ethylene)
What the BMS sees at Stage 2:
Rapidly rising surface temperature, dT/dt accelerating
Cell voltage may begin dropping as internal resistance rises
Gas venting may begin — pressure relief vents on prismatic cells open between 120–160°C
If gas sensors are present: CO₂ and hydrocarbon detection
Stage 3 — Cathode Decomposition and Oxygen Release
Onset Temperature
170–300°C (chemistry-dependent)
This is the stage that determines whether a cell merely vents aggressively or transitions to combustion. The cathode material begins decomposing, and the critical consequence is oxygen release.
For NMC cathodes, decomposition follows:
NMC Cathode Decomposition
LixNi0.8Mn0.1Co0.1O2→rock salt phase+O2↑+heatLi_xNi_{0.8}Mn_{0.1}Co_{0.1}O_2 \rightarrow \text{rock salt phase} + O_2\uparrow + \text{heat}LixNi0.8Mn0.1Co0.1O2→rock salt phase+O2↑+heat
The oxygen released reacts with the flammable electrolyte vapours that have been accumulating since Stage 2. This is the transition from an overheated cell to a cell that is now producing its own oxidiser internally.
LFP | NMC 811 |
|---|---|
TR onset: ~270–310°C | TR onset: ~170–210°C |
Heat of reaction: ~100–200 J/g | Heat of reaction: ~500–900 J/g |
Oxygen release: Minimal | Oxygen release: Significant |
Propagation rate: Slow | Propagation rate: Fast |
Can achieve TR? Yes | Can achieve TR? Yes |
The LFP advantage here is structural. The phosphate groups in the olivine lattice create a strong covalent P-O bond that holds oxygen far more tightly than the layered oxide structure of NMC.
🔑
The difference between LFP and NMC safety is not that LFP cannot have thermal runaway. It is that LFP's cathode decomposition starts 100°C higher and releases a fraction of the oxygen. Both can burn. Only one gives you enough time to evacuate the vehicle.
What the BMS sees at Stage 3:
Temperature above protection thresholds — contactors should already be open
Cell voltage collapsing toward zero
Significant gas venting — CO₂, CO, HF detectable
At this point, BMS protective actions are irrelevant. The cell is beyond electrical intervention.
Stage 4 — Electrolyte Combustion
Temperature Range
300–600°C
The electrolyte vapours, now mixed with cathode-released oxygen and exposed to temperatures well above the autoignition point (~300–400°C for common electrolyte solvents), combust. This is the first visible flame.
At this point the thermal runaway is fully self-sustaining. The combustion generates more heat than any cooling system can remove.
🔴
HF generation during Stage 4 is the primary acute hazard to first responders and anyone near a burning EV pack. A single 100 Ah prismatic cell can release 15–30 mg of HF during combustion. A 96S1P pack releasing all cells simultaneously produces quantities that are immediately dangerous to life in an enclosed space. Standard N95 masks provide no protection against HF gas.
The combustion products include:
CO₂ and CO from organic solvent oxidation
HF (hydrogen fluoride) from LiPF₆ electrolyte salt decomposition — toxic at concentrations above 3 ppm
H₂ from lithium-water reactions if moisture is present
Smoke particulate containing lithium compounds
Stage 5 — Full Thermal Runaway and Propagation
Peak Temperature
600–900°C
The cell reaches peak temperature. The separator melts completely (PE/PP separators melt at 130–170°C — they were gone long before this stage). The cell case ruptures or explodes from internal gas pressure.
Propagation timescales in real packs
In a tightly packed NMC prismatic module with no propagation barriers: cell-to-cell propagation typically takes 30–90 seconds per cell once the first cell reaches Stage 4. In a 96S pack, complete pack involvement can occur in 10–15 minutes from the first cell event. LFP prismatic packs with the same geometry show 3–5× slower propagation.
Propagation mechanisms:
Conductive: Heat transfers through cell-to-cell busbars and module structural components
Convective: Hot gases from the venting cell flow through the module enclosure
Radiative: At 600–900°C, radiative heat transfer is significant at close cell-to-cell distances
LFP vs NMC — The Safety Gap
✅ Pros
Exceptional thermal stability — no runaway below 270°C
3,000+ cycle life at 80% DOD
No cobalt — lower cost and ethical sourcing
Flat discharge curve — consistent voltage
❌ Cons
Lower energy density — 150 vs 250 Wh/kg for NMC
Flat OCV makes SOC estimation unreliable
Poor performance below 0°C — 40% power loss
Heavier packs for same range
The correct framing: LFP has a higher trigger threshold and lower consequence severity. An LFP pack that reaches thermal runaway is still a serious fire. It is simply more likely to be contained within the pack structure before propagating.
Trigger Mechanisms
Thermal runaway has three primary trigger classes:
Thermal abuse: External heat source raises cell temperature into Stage 1 directly. Fire exposure, adjacent cell event, BTMS failure in high ambient.
Electrical abuse: Overcharge pushes cell voltage past the cathode stability limit. Overcurrent generates I²R heat at rates that can reach Stage 2 in seconds.
Mechanical abuse: Crush or penetration creates an internal short circuit — a direct conductive path between anode and cathode inside the cell.
Battery Technology Timeline
1991
Sony commercialises the first lithium-ion cell — LiCoO₂ cathode, graphite anode
1997
Goodenough and Padhi discover LiFePO₄ — olivine structure, exceptional thermal safety
2008
Tesla Roadster launches — first highway-capable BEV, 53 kWh NCA pack
2020
BYD launches Blade Battery — cell-to-pack LFP, passes nail penetration without fire
2023
Na-ion enters mass production — CATL and HiNa ship first commercial cells
2024
Solid-state prototypes demonstrate >500 Wh/kg in lab conditions
Chemistry Comparison
LFPNMC 811Na-ion
Lithium Iron Phosphate — olivine crystal structure with strong P-O covalent bonds. Flat OCV plateau (3.2–3.3V) makes SOC estimation challenging. Thermal stability up to 270°C. Dominant in Indian commercial EVs, buses, and stationary storage. Cycle life exceeds 3,000 at 80% DOD.
Nickel Manganese Cobalt (8:1:1) — layered oxide structure with high nickel content for maximum energy density. Sloped OCV enables accurate SOC estimation. Thermal runaway onset at 170–210°C. Dominant in passenger EVs where range is the priority. Cycle life 1,200–1,500 typical.
Sodium-ion — emerging post-lithium chemistry using abundant sodium. Hard carbon anode with moderate OCV slope enables easy SOC estimation. Lower energy density (~160 Wh/kg) but excellent low-temperature performance. No lithium supply chain dependency. Entering commercial production in 2023–2024.
Quick Knowledge Check
Q: Which condition creates the highest thermal runaway risk during DC fast charging?
High ambient temperature onlyLow SOC with cold batteryHigh SOC with cold battery at maximum charge rateFull charge at moderate temperature
High SOC means the lithiated graphite anode is at maximum chemical potential — most reactive with electrolyte. Cold temperature means lithium plating risk is highest, and plated lithium can create internal short circuits. Maximum charge rate maximizes I²R heating. All three conditions simultaneously create the worst possible combination.
Q: What is the most underdeployed thermal runaway precursor technology in commercial EV packs?
dT/dt monitoringVoltage divergence at restGas detection sensorsAC impedance measurement
Gas detection is the highest-value, lowest-cost precursor technology. A CO₂ or VOC sensor costs under ₹500 per pack and provides 2–8 minutes of warning before Stage 4. The reason it is not standard is inertia, not engineering logic.
Reader Poll
Poll: How important is thermal safety vs range when choosing an EV?
Thermal safety is my #1 priority—Range matters more than safety specs—I want both equally—I don't know how to compare them—
Tap to vote
Precursor Detection
🔑
Gas detection is the single most underdeployed thermal runaway precursor technology in commercial EV packs. A CO₂ or VOC sensor costs under ₹500 per pack. The detection lead time of 2–8 minutes before Stage 4 is the difference between a contained venting event and a vehicle fire.
The BMS has four potential precursor signals:
dT/dt monitoring: Rate of temperature rise above predicted I²R load. Lead time: minutes to tens of minutes.
Voltage divergence at rest: Internal short causes self-discharge faster than neighbouring cells. Lead time: hours to days.
Impedance anomaly: AC impedance shows elevated charge-transfer resistance before temperature rises. Lead time: potentially days.
Gas detection: CO₂ and VOC sensors detect Stage 2 venting before visible symptoms. Lead time: minutes.
What the BMS Cannot Do
Once a cell reaches Stage 3, the BMS has no remaining electrical intervention that affects the outcome. Opening contactors removes the external current path — but the cell is now generating heat entirely from internal chemistry.
The BMS at Stage 3+ has exactly two useful functions:
Electrical isolation — open contactors to prevent the failing cell from being fed additional energy
Alert broadcast — transmit the highest-severity fault on CAN immediately
⚠
A BMS that tries to "manage" thermal runaway through active logic is dangerous. It creates false confidence, delays evacuation, and may interfere with the isolation step. Isolate and alert. Nothing else.
Pack-Level Propagation
In a production commercial EV pack, the thermal runaway event in a single cell is not the design challenge. The catastrophic scenario is propagation — the cascade where one cell triggers its neighbours, which trigger theirs, and the entire 300–500 kg pack becomes involved.
Propagation resistance is determined by:
Cell-to-cell gap and thermal interface: More gap means less conductive transfer. But more gap means lower energy density.
Vent path design: Prismatic cells need vent channels that direct hot gas away from adjacent cells.
Propagation barriers: Intumescent materials between cells that expand when heated, slowing propagation by 3–10×.
Module-to-module isolation: Structural gaps and thermal breaks between modules.
The regulatory minimum in AIS-156 Phase 2 and UN ECE R100 is that the pack must not propagate to a full-pack fire within 5 minutes of a single-cell thermal runaway event.
Key Takeaways
Thermal runaway is five distinct stages, not one event. Stages 1 and 2 are potentially recoverable. Stage 3 is the point of no return.
LFP's safety advantage over NMC is real — onset temperature 100°C higher, heat generation 5× lower, propagation rate significantly slower.
The BMS cannot prevent thermal runaway once Stage 3 begins. Its role is electrical isolation and alert broadcast.
Gas detection is the highest-value, lowest-cost precursor technology that is systematically underdeployed.
Propagation resistance is a pack mechanical design problem. The BMS cannot compensate for bad mechanical design.
High SOC is the highest-risk state for thermal runaway initiation.
💡
When designing pack architectures for Indian commercial EVs operating in 40°C+ ambient, prioritise LFP chemistry with propagation barriers, CO₂ gas sensors, and BMS firmware that isolates and alerts at Stage 2 detection. Do not attempt Stage 3+ "management" logic — it adds complexity without adding safety.
Resources and References
ℹ
All references verified as of May 2025. Paywalled papers include DOI for library or Sci-Hub access.
Standards and Regulations
AIS-156 Phase 2 (2023) — Automotive Industry Standard for Electric Power Train Vehicles, Phase 2. Bureau of Indian Standards / Ministry of Road Transport and Highways. Applicable to M and N category EVs from October 2023. https://www.bis.gov.in
UN ECE R100 Revision 3 (2021) — Uniform provisions concerning the approval of vehicles with regard to specific requirements for the electric power train. United Nations Economic Commission for Europe. https://unece.org/transport/documents/2021/03/standards/un-regulation-no-100-rev3
IEC 62660-2 — Secondary lithium-ion cells for the propulsion of electric road vehicles — Part 2: Reliability and abuse testing.
GB/T 31485-2015 — Chinese national standard for EV battery safety requirements — referenced for NMC and LFP thermal abuse data used in comparative studies.
Research Papers
Feng, X., Ouyang, M., Liu, X., Lu, L., Xia, Y., & He, X. (2018). Thermal runaway mechanism of lithium-ion battery for electric vehicles: A review. Energy Storage Materials, 10, 246–267. DOI: 10.1016/j.ensm.2017.05.013
Ren, D., Feng, X., Liu, L., Hsu, H., Lu, L., Wang, L., He, X., & Ouyang, M. (2020). Investigating the relationship between internal short circuit and thermal runaway of lithium-ion batteries under thermal abuse condition. Energy Storage Materials, 34, 563–573. DOI: 10.1016/j.ensm.2020.10.020
Spotnitz, R., & Franklin, J. (2003). Abuse behavior of high-power, lithium-ion cells. Journal of Power Sources, 113(1), 81–100. DOI: 10.1016/S0378-7753(02)00488-300488-3) — foundational paper on SEI decomposition and stage-wise thermal abuse.
Koch, S., Fill, A., & Birke, K. P. (2018). Comprehensive gas analysis on large scale automotive lithium-ion cells in thermal runaway. Journal of Power Sources, 395, 135–144. DOI: 10.1016/j.jpowsour.2018.05.080 — HF and gas composition data.
Wang, Q., Ping, P., Zhao, X., Chu, G., Sun, J., & Chen, C. (2012). Thermal runaway caused fire and explosion of lithium-ion battery. Journal of Power Sources, 208, 210–224. DOI: 10.1016/j.jpowsour.2012.02.038
Tianfeng Gao a b, Jinlong Bai a b, Dongxu Ouyang a, Zhirong Wang a, Wei Bai a, Ning Mao a b, Yu Zhu a, Effect of aging temperature on thermal stability of lithium-ion batteries: Part A – High-temperature aging
Technical Reports
NTSB (2020). Thermal Runaway in Lithium-Ion Battery Packs: Investigations and Recommendations. National Transportation Safety Board, Washington DC. https://www.ntsb.gov
EUCAR (2019). Battery Hazard Levels and Safety Testing Standards for Li-ion Batteries in Automotive Applications. European Council for Automotive R&D.
ARAI (2022). Testing Protocol for AIS-156 Phase 2 Compliance: Nail Penetration and Propagation Assessment. Automotive Research Association of India, Pune.
Further Reading — EVPulse Series
← Intermediate: Thermal Runaway — The 5 Stages Your BMS Is Racing Against
→ Master: Thermal Runaway — ARC Testing, Propagation Modelling, and Pack Certification
This is the Expert level of the EVPulse Thermal Runaway series.
Published on EVPulse — India's most technically rigorous source for battery technology and EV engineering coverage.