A lithium battery fire does not need air. That one fact explains why smothering it fails

Most fires need three things: fuel, heat and oxygen from the air. Remove any one and the fire stops, which is why a blanket, a lid or a carbon dioxide extinguisher works.

A lithium-ion cell in thermal runaway breaks that rule. Once it is properly under way, it is generating its own oxygen inside the cell, released by the cathode as it decomposes. Cutting off the air supply does not stop it. Smothering it does not stop it. The fire is being fed from inside a sealed metal can that you cannot reach.

That single fact explains nearly everything about how these failures behave — why water is the correct response despite the word lithium, why a damaged electric vehicle pack can reignite two days after it was extinguished, why lithium iron phosphate is genuinely safer rather than just marketed that way, and why an airline cares what state of charge your spare battery is at.

Runaway is a loop, not an event

Thermal runaway is not a single reaction. It is a feedback loop: heat raises the rate of chemical reactions inside the cell, those reactions release heat, and the released heat raises the rate further. Below a certain temperature the loop loses; above it, the loop wins and the outcome is fixed no matter what happens next.

The sequence is a staircase, and each step matters.

Around 80 to 120 °C, the solid-electrolyte interphase — the protective film on the anode that this site has described as the quiet governor of battery life — begins to break down. That exposes lithiated graphite directly to electrolyte, and they react exothermically. The cell is now warming itself.

Around 130 to 160 °C, a conventional polyolefin separator melts. The separator is the thin porous membrane that keeps the electrodes from touching while letting ions through, and when it melts or shrinks away, the electrodes touch. That is an internal short circuit, delivering the cell's entire stored charge into its own interior as heat, in seconds.

Above roughly 180 to 250 °C, depending on the chemistry and how charged the cell is, the cathode decomposes — and this is the step that changes the character of the event entirely.

The cathode is the oxidiser

A charged layered-oxide cathode — lithium cobalt oxide, or the nickel-manganese-cobalt oxides used in most high-energy cells — is a metal oxide that has had much of its lithium removed. That leaves it in a state that is thermodynamically eager to shed oxygen, and when it gets hot enough it does exactly that, releasing oxygen gas inside a sealed cell already full of a flammable organic electrolyte.

At that point the cell contains fuel and oxidiser in intimate contact, at temperature, with no way for anyone outside to separate them. The fire is self-sufficient.

This is where lithium iron phosphate differs, and the reason is specific rather than promotional. In LFP the oxygen is held in a phosphate framework by strong phosphorus–oxygen bonds rather than sitting in a layered oxide lattice. It does not release oxygen the same way, its decomposition onset is markedly higher, and when it does fail it does so with far less energy and usually without the same violent oxygen-fed fire.

That is the mechanistic version of the claim made in passing in our article on battery ageing, that LFP is more robust. It gives up energy per kilogram, and what it buys with that is a cathode that is much less willing to become an oxidiser.

The safety of a lithium-ion cell is not a single property. It is largely a statement about one number — the temperature at which the cathode starts giving up oxygen — and about how much energy is stored waiting for that moment.

What starts it

Four routes, and they are worth distinguishing because the defences differ.

Mechanical. Crushing or puncturing a cell shorts the electrodes directly. This is the vehicle-crash case and the dropped-phone case, and it is why a visibly deformed cell is not usable.

Electrical. Overcharging drives lithium metal onto the anode instead of into it — the plating described in the ageing article — and metallic lithium is both reactive and prone to growing structures that puncture the separator later. Overcharging also generates heat directly. Over-discharging is quieter and nastier: it can dissolve copper from the current collector, which later redeposits as conductive bridges and causes an internal short weeks afterwards.

Thermal. External heat alone is enough. A cell left in a parked car in an Indian summer, or next to a heat source, can be walked up the staircase without any electrical abuse at all.

Internal defects. A metallic particle left inside the cell during manufacturing can, over time and with charge cycling, work its way through the separator and create a short. This is the failure mode behind several large product recalls, and it is why cleanliness in cell manufacturing is a safety requirement rather than a quality nicety — and why cheap, unbranded cells are a genuinely different risk proposition from cells made in a controlled facility.

There is also one trap worth repeating from the ageing article: charging a cold cell plates lithium rather than intercalating it. The consequence is not immediate — it is a cell that has quietly acquired the structures that cause an internal short later.

Why one cell becomes a pack fire

A single 18650 cell in runaway is a violent but survivable event. A pack of hundreds is a different matter, and the difference is propagation.

A cell in runaway dumps its energy as heat into whatever is around it, which is its neighbours. If they reach onset temperature, they go too, and the front moves through the pack. Much of the engineering in a good battery pack is not about the cells at all — it is about ensuring that one cell's failure stays one cell's failure: spacing and thermal barriers between cells, materials that absorb heat, deliberate venting paths that direct hot gas out of the pack rather than into the next cell, and fusing at cell level.

This is the honest reason the same cells can be safe in one product and dangerous in another. A well-designed pack with a working battery management system, thermal margin and propagation barriers contains a bad cell. A pack assembled to hit a price — no thermal management, minimal BMS, cells packed tight — turns one defect into a fire. India's pattern of electric two-wheeler and power-bank incidents has been dominated by exactly this: integration and quality control rather than any flaw in lithium-ion chemistry as such.

Fighting it, and the misconception in the name

The word "lithium" does a lot of damage here. Burning lithium metal is a classic reactive-metal fire where water is genuinely the wrong answer. A lithium-ion cell contains no bulk lithium metal — the lithium is present as ions inside the electrode materials — and the guidance for lithium-ion packs is the opposite: use water, and a great deal of it.

The reasoning follows directly from runaway being a thermal loop. You cannot remove the oxidiser, because it comes from the cathode. You cannot remove the fuel, because it is sealed inside. The only variable left is temperature, and water is the best heat sink available in quantity. Fire services attacking an electric vehicle fire are not trying to extinguish flames so much as to cool the pack below the temperature at which the remaining cells will propagate — which takes far more water than a conventional vehicle fire, and time.

Two consequences follow. Smothering agents and carbon dioxide may knock down visible flame and will not stop runaway, because they do nothing about internal heat or internal oxygen. And reignition hours or days later is normal, because cells that were heated but did not yet go can cross their threshold long after the visible fire is out. This is why damaged EV packs are quarantined at a distance, sometimes submerged, and monitored for days rather than declared safe.

What a person can actually do

A swollen cell is gassing and is finished. Bulging means internal decomposition has generated gas. Stop using it, do not charge it, do not puncture or compress it, and dispose of it through e-waste channels. A swollen power bank or phone battery is the single clearest warning sign most people will ever get.

Do not charge unattended on soft furnishings. Beds and sofas both insulate and burn. Charging on a hard surface with air around it costs nothing and removes the worst case.

Use the charger and the pack the device was designed for. Uncertified chargers and replacement packs are where overcharge protection and cell quality are most often missing, and they are heavily represented in incident reports.

Respect the temperature warnings. A device that refuses to fast charge when hot or cold is protecting itself from exactly the mechanisms above, and that refusal is worth more than the convenience it costs.

Where materials work is actually happening

The nanomaterials contribution here is real and specific, and it is mostly about raising the temperature of each step on the staircase.

Ceramic-coated separators are the clearest example. Coating a polyolefin separator with a thin layer of alumina or similar oxide particles dramatically reduces thermal shrinkage and raises the temperature at which the separator fails, which buys margin at the most dangerous step in the sequence. This is a direct metal-oxide application in a safety-critical role, and it is now standard in many cell designs.

Cathode particle coatings — thin conformal oxide shells of the kind deposited by atomic layer methods — improve thermal stability and reduce reactivity with electrolyte at the surface, where decomposition starts.

Electrolyte additives and flame-retardant formulations raise the onset of the reactions that start the loop, and solid electrolytes remove the flammable liquid entirely, which is the strongest structural answer available and brings its own unresolved problems of interfaces, dendrites and manufacturability.

The honest framing for all of it: none of these makes a high-energy cell inert. They shift onset temperatures and reduce the energy released. A claim of a "safe" battery should always be read as a question about how much safer, measured how, and at what cost in energy density — because energy density and safety are coupled, and a great deal of stored energy in a small volume is never going to be intrinsically harmless.

Why it matters for students and researchers

This is the clearest example in energy engineering of a coupled trade-off that cannot be designed away. Packing more energy into less volume necessarily means more energy available to release if something goes wrong, so safety engineering here is about controlling the release rather than eliminating the possibility. Students who internalise that stop looking for the safe high-energy chemistry and start working on onset temperatures, propagation barriers and detection.

It also illustrates the limits of abuse testing. Nail penetration, crush, overcharge and thermal tests are standardised and necessary, and they test a cell's response to a defined insult — not its likelihood of developing an internal short from a manufacturing particle three years later. The most consequential real-world failures have often been ones that abuse testing does not reproduce, which is an uncomfortable and important thing to understand about qualification.

The open problems are pressing. Early detection of an incipient internal short, before runaway, would change pack safety more than any material change. Propagation-resistant pack architectures that do not cost unacceptable weight and volume remain an active design problem. Non-flammable electrolytes with usable conductivity are a long-standing goal. And for India specifically, the binding constraint on battery safety right now is cell quality, pack integration and standards enforcement rather than frontier chemistry.

Frequently asked questions

Can I use water on a lithium-ion battery fire?

Yes, and in quantity — this is the standard guidance for lithium-ion packs. The confusion comes from lithium metal fires, where water is genuinely wrong, but a lithium-ion cell contains no bulk lithium metal. Since you cannot remove the fuel or the oxygen from inside a sealed cell, cooling is the only mechanism available, and water is the most effective coolant available in bulk.

My power bank or phone battery has swollen. Is it dangerous?

Treat it as finished. Swelling means gas has been generated by internal decomposition, which indicates the cell has already degraded in a way that raises its risk of runaway. Stop charging and using it, avoid puncturing or squeezing it, keep it away from heat and flammable surroundings, and take it to an e-waste collection point rather than a household bin.

Why do electric vehicle fires reignite hours later?

Because cooling the outside of a pack does not necessarily cool every cell inside it. Cells that were heated but had not yet crossed their runaway threshold can do so long after the visible fire has gone out, and propagation can restart from there. This is why damaged packs are kept isolated and monitored for an extended period rather than assumed safe once flames stop.

Is LFP genuinely safer, or is that marketing?

Genuinely safer, for a structural reason. In lithium iron phosphate the oxygen is bound in a phosphate framework by strong phosphorus–oxygen bonds, so the cathode does not readily release oxygen when hot, its decomposition begins at a higher temperature, and failures are markedly less energetic. The trade-off is lower energy per kilogram, which is why LFP dominates stationary storage and cost-sensitive vehicles while high-nickel oxides persist where range matters most.

Why must batteries be shipped at around 30% charge?

Because a cell's stored energy and its willingness to run away both depend on state of charge. A partially charged cell has less energy to release and a higher onset temperature for cathode decomposition, so both the likelihood and the severity of an incident fall. That is why air transport rules limit the state of charge of lithium-ion cells shipped as cargo, and why loose spare batteries are required in the cabin, where a problem can be seen and dealt with, rather than in the hold.