Deadly Gases and Structural Collapse: What Really Happens When EV Batteries Catch Fire
By Jacobin
When an electric vehicle bursts into flames—whether after a crash, during charging, or even while parked—the real danger may not be the fire itself. Beneath the dramatic plumes of black smoke lies something far more insidious: an invisible cocktail of toxic gases that can incapacitate or kill within minutes. A new study from the Institute of NBC Defence in Beijing is forcing automakers, regulators, and first responders to rethink how they assess—and respond to—battery thermal runaway events in real-world settings.
For years, the industry’s focus has been on preventing thermal runaway: engineering safer cells, deploying better battery management systems (BMS), and designing robust pack enclosures to contain damage. But this latest work, led by Zhou Tian, Sun Jie, and a team of electrochemical and safety experts, shifts attention past ignition—to what happens once a cell ruptures, flames erupt, and volatile chemistry takes over.
The findings are sobering. Under flame-triggered abuse conditions, common soft-packed NCM811 lithium-ion pouch cells—exactly the type used in many high-end EVs today—release at least 14 distinct volatile compounds during thermal runaway. Among them: carbon monoxide (CO), hydrogen fluoride (HF), acrolein, and acrylonitritrile—substances known for their acute toxicity, even at low concentrations. CO levels surged past 1,000 parts per million (ppm) within seconds of flame extinction and remained dangerously elevated for over 30 minutes in a confined space. To put that in context: OSHA’s permissible exposure limit for CO is 50 ppm over 8 hours. At 1,200 ppm, loss of consciousness can occur in under three minutes.
But it’s not just the gases. The study meticulously documents how the internal architecture of the battery disintegrates under stress—how the polymer separator melts, how molten electrode fragments breach containment, and how metallic residues from the cathode cross into the anode layer, triggering secondary reactions that amplify heat and gas generation. In high–state-of-charge (SOC) cells—especially those above 50%—this structural degradation isn’t gradual. It’s catastrophic, near-instantaneous, and self-propagating.
“The moment the separator fails, you’re no longer dealing with a single failing cell,” explains co-author Sun Jie, a professor specializing in advanced energy and safety. “You’re watching a domino effect unfold inside a sealed metal box—where molten nickel, cobalt, and manganese oxides mix with decomposed electrolyte, graphite, and aluminum foil at temperatures exceeding 600°C. That’s not just a fire hazard. That’s a chemical reactor gone rogue.”
A Flame Test That Mirrors Reality
Unlike many lab studies that use slow, controlled heating (e.g., oven tests or nail penetration), this team opted for a more representative—and brutal—trigger: direct flame impingement at the cell’s weakest point—the junction near the electrode tabs. Using a custom-built 50 cm³ combustion chamber, they exposed 16 Ah NCM811/graphite pouch cells to open flame at four different charge levels: 0%, 30%, 50%, and 100% SOC.
The results were stark.
At 0% SOC, the cell merely vented warm vapor—mostly unreacted electrolyte solvent (ethyl methyl carbonate)—with little flame, minor deformation, and <10% mass loss. No CO. No HF above detection limits. No structural collapse beyond surface wrinkling.
At 30% SOC, things escalated: a delayed but vigorous smoke eruption, followed by a brief (~50-second) flame jet. The outer aluminum laminate peeled like burnt paper. Internal pressure built, then released in a whoosh of vaporized organics—now mixed with benzene, toluene, and trace CO. The separator began to deform, shrinking and fusing at pore edges.
At 50% and 100% SOC? Full-scale thermal detonation.
Ignition occurred in under 30 seconds. Jets of flame—sometimes over a meter long—erupted from both ends of the cell. Solid particulates, glowing red-hot, were ejected alongside gas plumes. One 100% SOC cell launched fragments of molten electrode material upward into the chamber ceiling. Total mass loss exceeded 65% in under 90 seconds—not just from gas, but from physical ejection of active materials. Crucially, the time between ignition and violent venting shrank from 26 seconds (at 30% SOC) to just 2 seconds at full charge. “That window is too short for any human or automated system to intervene,” notes Zhou Tian, the study’s lead author and a lecturer in electrochemistry and energy safety.
The Toxic Timeline: When the Fire Goes Out, the Danger Begins
Perhaps the most operationally significant finding concerns the temporal mismatch between visible danger and invisible threat.
All sensors recorded peak flame temperatures—often exceeding 660°C—within the first 60 seconds. But crucially, CO concentrations remained relatively low during active combustion. Why? Because the fire itself was oxidizing much of the carbon monoxide into carbon dioxide.
The real spike came after the flames died.
Between 200 and 400 seconds post-ignition—when the cell appeared “safe,” just smoldering and cooling—CO levels surged, hitting 1,385 ppm in one test and holding above 800 ppm for nearly 20 minutes. Hydrogen fluoride (HF), though less persistent, spiked early in low-SOC tests but vanished in high-SOC runs—likely consumed in high-temperature fluorination reactions that produced fluorinated hydrocarbons (e.g., fluorobenzene), many of which are carcinogenic and environmentally persistent.
Meanwhile, GC-MS analysis revealed an ever-expanding roster of volatiles as SOC increased:
- 0% SOC: 7 compounds (mostly solvents and fluorobenzenes)
- 30% SOC: 10 compounds (adds benzene, CO₂, 1,3-cyclopentadiene)
- 50% SOC: 12 compounds (introduces acrolein—a potent lachrymator and lung irritant)
- 100% SOC: 14 compounds, including acrylonitritrile, a known human carcinogen with a short-term exposure limit (STEL) of just 2 ppm.
Acrylonitrile wasn’t just present—it was rising over time in high-SOC tests, suggesting ongoing decomposition in residual hot spots long after visible activity ceased.
Inside the Meltdown: How Structure Fails—Cell by Cell
Using scanning electron microscopy (SEM), X-ray diffraction (XRD), and X-ray photoelectron spectroscopy (XPS), the team reconstructed the physical degradation in chilling detail.
The cathode (NCM811) starts as a neatly ordered agglomeration of spherical secondary particles—each a cluster of primary nanocrystals. At 30% SOC post-runaway, SEM shows partial sintering: particles fuse at contact points, pores close, surface roughens. By 100% SOC? Total collapse. The layered oxide structure vanishes. What remains is a vitrified, glassy crust riddled with metallic nickel nodules (confirmed by XRD peaks at 44.5°, 51.8°) and nickel oxide—proof that the highly oxidized Ni⁴⁺ in charged cathodes violently reduced during thermal runaway, releasing oxygen that fed combustion.
Even more alarming: at high SOC, XRD detected aluminum metal peaks—meaning the cathode current collector (normally stable up to ~500°C) had melted and alloyed with active material. In one sample, graphite peaks appeared in the cathode debris—evidence that anode material had crossed the failed separator and mixed in.
The anode, meanwhile, undergoes its own metamorphosis. Fresh graphite exhibits textbook hexagonal layering under SEM. After 30% SOC runaway, the surface is coated in melted polymer (separator residue) and cracked SEI film. At 100% SOC? The layers are obliterated. The surface becomes a chaotic amalgam of carbon, copper droplets (from the current collector), and embedded transition-metal particles. XPS shows the near-total loss of organic SEI components (like ROCO₂Li), replaced by inorganic carbonates and—critically—metal carbides (e.g., Ni₃C), detectable via distinct C1s binding energy shifts.
The separator, the cell’s last line of defense, is the first to fail. Polyethylene-based films begin shrinking at ~130°C, pores closing by 150°C. At 30% SOC, SEM reveals a shrunken, pore-free film with embedded ceramic (Al₂O₃) particles still intact. At 50% SOC, those ceramic fillers detach and migrate. By 100% SOC? The separator is gone—not just melted, but chemically decomposed. XRD finds no polyethylene signature whatsoever; only traces of Al₂O₃, graphite, and metal oxides remain—likely adhered to electrode surfaces.
“This isn’t degradation,” Sun Jie emphasizes. “This is annihilation. And once the separator is compromised, the system transitions from electrochemical failure to thermochemical chain reaction.”
Implications for the Real World
So what does this mean for drivers, firefighters, and designers?
For emergency responders: Standard turnout gear and SCBA (self-contained breathing apparatus) may be insufficient. HF can penetrate some glove materials; acrylonitrile is volatile enough to permeate certain mask seals over time. The study strongly implies that “cooling and walk-away” strategies—common after EV fire suppression—could leave crews exposed to lingering toxic plumes during overhaul. Continuous gas monitoring after visible flameout is non-negotiable.
For automakers: Current battery pack designs prioritize flame containment and venting—but rarely address toxicant filtration. Could activated carbon/alkaline scrubber layers be integrated into vent paths? Could inert gas (e.g., argon) injection systems suppress post-flame CO generation by limiting oxygen re-entrainment? The data suggest such innovations are overdue.
For regulators: Safety standards like GB 38031-2020 (China) and UN GTR No. 20 focus on preventing propagation and ensuring structural integrity during abuse. Yet none mandate measurement—or mitigation—of toxic emissions. With this study providing a validated protocol (flame trigger + real-time GC-MS + multi-gas sensing), a new class of standards may be imminent.
For consumers: There’s reassurance, too. The 30% SOC “tipping point” aligns with fleet-management best practices: many EV taxi and delivery services already cap storage SOC at 30–50% when vehicles are parked for extended periods. This isn’t just about longevity—it’s about reducing runaway severity by orders of magnitude.
Looking Ahead: From Reactive to Predictive
The team is now adapting these insights into early-warning systems. By correlating specific gas ratios (e.g., CO/H₂ or acrolein/benzene) with degradation stage, they aim to develop sensors that don’t just detect “something’s wrong,” but pinpoint what is failing—and how urgently.
“We used to think of battery safety as binary: stable or runaway,” Zhou reflects. “But this work shows it’s a spectrum—of chemistry, structure, and toxicity. And on that spectrum, there are warning signs before the jet of flame. Our job now is to learn to read them.”
As EV adoption accelerates globally, such nuanced understanding won’t just improve engineering—it could save lives.
Author: Zhou Tian, Sun Jie, Li Jigang, Wei Shouping, Chen Jing, Zhang Fan
Affiliation: Institute of NBC Defence, Beijing 102205, China
Journal: Energy Storage Science and Technology, Vol. 13, No. 11, November 2024
DOI: 10.19799/j.cnki.2095-4239.2024.0519