Thermal runaway

Thermal runaway is the uncontrolled rise in temperature of a reacting material or device when the heat produced by its own chemical reactions exceeds the heat it can lose to its surroundings. The higher temperature speeds up the reactions, which produce still more heat, until the material ignites or its reactive content is used up. In the SFPE Handbook the term is used as a synonym of runaway reaction, originally for chemical reactors. Today it is best known as the failure mechanism behind a lithium-ion battery fire, in which a battery cell heats itself and can set fire to its neighbours.

The same physics explains the self-heating of haystacks, oily rags and coal piles, so the general theory of self-heating is the starting point for understanding batteries. In an electric vehicle, thermal runaway can start at temperatures of about 70 to 250 °C, depending on the type of battery. It releases flammable and toxic gases and can make the fire reignite until the battery's energy is used up. The peak heat release rate of an electric vehicle fire is similar to that of a petrol or diesel car. Its slow start, long duration and the difficulty of extinguishing it are what set it apart.

Definition

Gray describes spontaneous combustion as the reaction of an unstable, usually oxidisable, material that releases heat. Much of this heat stays inside the material because it, or its container, conducts heat poorly. If the process ends in flaming, it is called spontaneous ignition.

Mild self-heating that settles at a slightly raised temperature is called subcritical. Self-heating that ends in ignition is called supercritical, and the sharp boundary between the two is the critical condition. Runaway reaction and thermal runaway are the names used for supercritical behaviour in a chemical reactor, which is almost always well stirred. Other chapters of the SFPE Handbook use the term in the same sense. In the thermal theory of autoignition, ignition is taken as the instant at which the temperature in the reacting mixture rises rapidly. This event is called thermal runaway and also marks the start of a rapid consumption of the reactants.

In the theory of heat conduction, thermal runaway is the case in which the temperature of a heat-generating solid grows exponentially with time and no steady state exists. For batteries, a report by the Dutch Institute for Public Safety (NIPV) on car parks quotes a definition by Colella and co-workers. In that definition, thermal runaway is a failure mechanism that leads to self-heating in a battery or battery cell and that can result in fire. The NIPV report notes that thermal runaway and the burning of the released gases are different processes, and that the link between them is not yet well understood.

Heat generation and heat loss

The mechanism is a competition between two heat flows. Most chemical reaction rates increase almost exponentially with temperature, while heat loss by conduction increases only linearly. Once the heat production grows faster than the loss, the temperature can only rise. A larger body and a higher ambient temperature both reduce the relative heat loss and so favour ignition. Inside a self-heating body the temperature profile is roughly parabolic, so runaway starts at the centre, where the temperature is highest.

Semenov's theory, the simplest description, writes this competition as a heat balance for a body with a uniform temperature. The body stores heat according to its heat capacity, gains heat from the reaction in its whole volume and loses heat through its surface. The rate of change of the stored heat equals the heat generated minus the heat lost:

$$C_v\rho V\frac{dT}{dt}=VQf(c)e^{-E/RT}-S\chi (T-T_a)$$

Symbol	Description	Unit
$C_v$	heat capacity at constant volume, per unit mass	J/(kg·K)
$\rho$	density of the material	kg/m³
$V$	volume of the body	m³
$T$	temperature of the reacting material	K
$t$	time	s
$Q$	heat of reaction per unit concentration of fuel	J/mol
$f(c)$	kinetic rate law, a function of the fuel concentration $c$	mol/(m³·s)
$E$	activation energy of the reaction	J/mol
$R$	universal gas constant, 8.314	J/(mol·K)
$S$	surface area across which heat is lost	m²
$\chi$	heat transfer coefficient	W/(m²·K)
$T_a$	ambient temperature, constant in time	K

The first term on the right is the heat generated by the self-heating reaction, and the second is the heat lost to the surroundings. Their difference drives the change in temperature. The equation cannot be solved by classical methods, but the critical condition can be found from it exactly.

The relation assumes a uniform temperature, a single reaction of simple order, and a heat of reaction and activation energy large enough to give ignition behaviour. It suits well-stirred liquids and gases. It is a poor approximation for poorly conducting materials such as cellulose, with a conductivity of about 0.05 W/(m·K), although it still gives semi-quantitative results there. It is a general model of self-heating and not a model of a battery.

The balance is easiest to understand as two curves of heat flow against temperature, the Semenov diagram. At a low ambient temperature the straight heat-loss line cuts the curved heat-generation curve twice. The lower crossing is a stable state, usually 5 to 20 °C above ambient, in which the material may degrade or discolour but does not ignite.

Semenov diagram: heat generated by the reaction (curve) and heat lost to the surroundings (straight lines) for three ambient temperatures. Based on Gray (2002), pp. 2-213–2-214.

At the critical ambient temperature the line just touches the curve. Above it there is no balance point at all, and the temperature rises until ignition, at a rate proportional to the imbalance between heat production and loss. The slope of the loss line depends on the ratio of surface area to volume. For a given ambient temperature there is therefore also a critical size, above which a body will run away.

A worked example shows why the critical point lies close to ambient conditions. Many spontaneous combustion reactions, in particular those of cellulosic materials, have activation energies of around 100 kJ/mol. Gray shows that criticality exists only if the activation energy is at least four times the product of the gas constant and the critical ambient temperature. At an ambient temperature of 30 °C (303 K) this product is 8.314 × 303 = 2.52 kJ/mol, so four times it is 10.1 kJ/mol. The activation energy of 100 kJ/mol is about ten times larger, so the reaction accelerates strongly enough for a sharp critical condition.

The ratio of the gas constant times the squared temperature to the activation energy is 8.314 × 303² / 100,000 = 7.6 K. Kanury points out that this ratio is generally small, which is why the critical temperature of a reacting mass lies close to the temperature of its surroundings. As a rule of thumb for materials with an unknown activation energy, self-heating of more than 30 °C above ambient, typically a body temperature above 60 to 70 °C, indicates imminent spontaneous ignition.

Size, storage and time to ignition

Because heat loss depends on surface area and heat generation on volume, size matters as much as temperature. Bodies placed in contact act as one larger body. Two bales that touch have a considerably lower critical ambient temperature than a single bale, so a test on one item says little about a stack unless a theory is used to scale it up. Bodies that are not in contact can still interact by heating the air around them. In tests with kegs of 14 kg in still air, the critical ambient temperature fell from 62.5 °C for a single keg to 54 °C for 18 kegs packed in a steel box.

The time to ignition is very hard to predict. It becomes infinite at the critical condition and is extremely sensitive to conditions close to it. Larger bodies generally take longer: oily rags in a wastepaper basket may ignite within a few hours, haystacks after a few weeks and coal stockpiles after a few months. Bodies that are stacked while still hot ignite much sooner. Deep-seated burning that starts by spontaneous ignition is difficult to extinguish completely and often reignites days after it seemed to be out.

The International Fire Engineering Guidelines note that self-heating depends not only on the chemical nature of a substance but also on the size and shape of the aggregation. In continuously stirred chemical reactors, the scale-up from a prototype to a commercial size has led to the critical condition being exceeded. The same lesson applies to stacks and stores: the critical condition belongs to the whole assembly, not to a single item.

Thermal runaway in lithium-ion batteries

A battery pack in an electric vehicle is built up in steps. Cells deliver about 4 V, cells are combined into modules of up to 60 V, and modules are connected into packs of 300 to 1,000 V. Packs in passenger cars can store 100 kWh or more. Packs can also be connected into larger battery systems of 10 to 400 kWh, for example for trucks and electric ships. Stationary installations of this kind are described under battery energy storage system.

Each battery pack has its own safety window. If it is exceeded, the pack can heat up further and thermal runaway can occur. The onset temperature depends on the type of battery and ranges from 70 to 250 °C. Elsewhere the NIPV report gives a range of 50 to 270 °C, and it stresses that "lithium-ion" covers many battery types with different fire behaviour. A battery management system monitors the cells. When a threshold is exceeded it can apply active cooling during charging, stop the charging or disconnect the whole pack, and a fault in the system itself also switches the pack off.

The causes can be internal or external. Internal causes are rare design or manufacturing errors in the electrode, separator or electrolyte, which can lead to failures of the management system and to overcharging; such errors usually lead to recalls. External causes include outside heat, misuse and extreme conditions such as a collision that damages the battery. In car parks, damaged charging cables or stations, vehicles that hit a charging station and vehicles with an unnoticed defective pack add to these causes.

Published failure probabilities differ widely, from one in a million or lower to about one in 100 million. Overall the probability of battery failure seems relatively low. The NIPV report adds, however, that no lithium-ion battery is intrinsically safe. The design of a pack cannot completely prevent the release of enough energy for thermal runaway.

Once a cell runs away, the process can spread. Heat from one cell warms the adjacent cells, which then also go into thermal runaway, in a chain reaction through the pack. As the battery heats up it releases flammable gases, which ignite if enough oxygen is present and the ignition temperature is reached; the hot cell housing may be enough. Gases trapped in a pouch or cartridge can escape when the enclosure or a pressure valve fails. In the right mixture with air they can cause a flare fire or a gas cloud explosion, and mechanical damage can throw out burning fragments.

Structure of an electric vehicle battery pack and spread of thermal runaway from cell to cell. Based on Van der Graaf et al. (2021/2023), pp. 21–22.

Ageing changes the behaviour. As a battery ages, its maximum state of charge falls, and since the intensity of thermal runaway depends on the state of charge, the intensity decreases. An older battery in thermal runaway also produces fewer flammable gases than a new one.

Fire behaviour of electric vehicles

A battery fire develops in two stages. While the cells warm up and start to release gas, the temperature rises slowly, at about 0.2 °C per minute. Once thermal runaway occurs, the rise becomes exponential, at more than 10 °C per minute. A conventional car fire burns for just over half an hour and peaks at about 5 MW after 10 to 15 min. Without intervention an electric vehicle burns out in about 90 min. After extinguishing, the energy remains in the battery and the fire can keep reigniting for several hours.

Schematic development of a battery fire: a slow rise while the cells warm up and release gas, then a steep rise once thermal runaway occurs. Based on Van der Graaf et al. (2021/2023), p. 26.

The peak heat release rate is not much different from that of a conventional car. A review by Sun and co-workers found a similar peak of about 6 MW for both. The time to the peak was 6 to 40 min for electric vehicles and 6 to 35 min for conventional cars. The total heat released was 4.7 to 8.5 GJ and 3.3 to 10 GJ respectively. Vehicles with packs of 15 to 25 kWh give about 4 to 6 MW, and the growth of packs to more than 100 kWh has raised this to about 8 MW.

Peak heat release rate in full-scale fire tests of battery electric, plug-in hybrid and conventional cars. Based on Sun et al. (2020), as cited in Van der Graaf et al. (2021/2023), p. 24.

Measured flame temperatures do not differ significantly either. In tests by Lam and co-workers, electric and conventional vehicles both reached about 800 ± 100 °C over periods of 5 to 30 min. In the Netherlands, the car fire curve in the standard NEN 6098 for car parks is based on a TNO design fire from 1999. In those tests, fire could spread to neighbouring cars 10 to 15 min after the first car. The book Introduction to Fire Safety Engineering notes that these curves may be less conservative for modern cars with more plastics.

The gases differ more than the heat. Burning batteries release methane, carbon monoxide, carbon dioxide and hydrogen, as well as the toxic gases hydrogen fluoride and hydrogen chloride. A burning pack releases about 20 to 200 mg of hydrogen fluoride per Wh of capacity. For a 100 kWh pack, that amounts to 2 to 20 kg, depending on the state of charge. Hydrogen fluoride is a strong irritant. Purser lists 200 ppm as the concentration predicted to impair escape in half of the population and 900 ppm as the concentration predicted to incapacitate half.

For people escaping, smoke from any vehicle fire is already very toxic because of the burning plastics. On that basis the extra hydrogen fluoride from the battery is considered to have no effect on escape safety. Smoke with more hydrogen fluoride can still spread from the car park to escape routes and stairs of buildings above it.

The released gases also create an explosion hazard. If they do not ignite at once, an explosive mixture can build up in a confined or poorly ventilated space. When it reaches the lower explosion limit and meets an ignition source, a fire gas explosion can follow. Hydrogen, one of these gases, has a very wide flammable range: at room temperature a flame can propagate upwards through mixtures with 4.1 to 74 % hydrogen by volume in air. Zalosh describes explosion protection as breaking the chain of events: preventing a fuel-air mixture by ventilation or inerting, removing ignition sources, or limiting the pressure rise by venting or suppression. Fires of electric vehicles can involve repeated small explosions, flying parts and flash fires.

Firefighting and design measures

A fire in which the battery pack is involved is very difficult or impossible to extinguish. The pack is closed, and the gases released by the self-propagating thermal runaway keep reigniting. The coolant, often water, cannot reach the cells directly because the pack is sealed against moisture and dirt. Cooling therefore needs a lot of water for a long time, and the situation is safe only when all electrical and chemical energy in the battery has been used up. Putting out the flames needs no more capacity than for a conventional car, but long-term cooling to prevent reignition needs much more water.

At present, an immersion container is the best available method for the long-term cooling of a battery pack. Blanketing the vehicle with foam is not yet a useful tactic, because laboratory studies show that packs can reignite during thermal runaway even with little oxygen. Sprinklers and water mist cannot extinguish a burning electric vehicle, because the droplets cannot enter the pack. They do reduce the probability that the fire grows and spreads to nearby vehicles and to the structure.

Ways to bring water into the pack are being studied. One is a sprinkler inside the pack, which could cool the cells in thermal runaway more effectively and possibly extinguish them. Another is a hose connection in the casing of the pack, tested by a car manufacturer. In a study of several detector types, carbon monoxide detectors gave the fastest alarm.

A fire in the Singelgarage in Alkmaar on 1 July 2020 illustrates these problems. A police robot was used to tow an electric vehicle out of the car park because its battery pack kept reigniting, and the pack was then cooled for a long time in an immersion container. For indoor car parks in the Netherlands, the NIPV report suggests, among other measures, not placing charging bays near pedestrian escape routes, because toxic and caustic products of thermal runaway can spread there. It also suggests a means of switching off all charging points at once.

Limits and misconceptions

The theory of thermal runaway is well developed for self-heating materials and chemical reactors, but it rests on simplifications. Semenov's model assumes a uniform temperature and a single simple reaction, and the classical corrections are often impossible to apply because data are lacking or uncertain. No formula is available in these sources for the runaway of a battery cell or for its spread from cell to cell. For batteries, the NIPV report lists open questions: the relation between thermal runaway and fire development, the conditions for reignition, the effect and duration of external cooling, and the thermal load on car park structures.

A common misconception is that electric vehicles burn hotter or with a much larger peak heat release rate than conventional cars. The data summarised by NIPV show similar peaks and similar flame temperatures. The differences lie in the slow start, the long duration, the risk of reignition and explosion, and the larger amount of hydrogen fluoride. Another misconception is that a fire that seems to be out is safe. Both in self-heating stockpiles and in battery packs, deep-seated reactions can restart hours or days later.

See also

• Car park fire
• Tunnel fire
• Warehouse fire

References

• Gray, B. (2002). Spontaneous combustion and self-heating. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 2-211–2-228). National Fire Protection Association; Society of Fire Protection Engineers.
• Iqbal, N., & Salley, M. H. (2004). Fire Dynamics Tools (FDTs): Quantitative fire hazard analysis methods for the U.S. Nuclear Regulatory Commission Fire Protection Inspection Program (NUREG-1805). U.S. Nuclear Regulatory Commission, Office of Nuclear Reactor Regulation.
• Kanury, A. M. (2002b). Ignition of liquid fuels. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 2-188–2-199). National Fire Protection Association; Society of Fire Protection Engineers.
• National Research Council of Canada, International Code Council, Department of Building and Housing, & Australian Building Codes Board. (2005). International fire engineering guidelines (2005 ed.). Australian Building Codes Board.
• Purser, D. A. (2002). Toxicity assessment of combustion products. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 2-83–2-171). National Fire Protection Association; Society of Fire Protection Engineers.
• Rockett, J. A., & Milke, J. A. (2002). Conduction of heat in solids. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 1-27–1-43). National Fire Protection Association; Society of Fire Protection Engineers.
• Van der Graaf, J., Hessels, T., & Rosmuller, N. (2023). Fire safety of indoor car parks accommodating electrically powered vehicles. Nederlands Instituut Publieke Veiligheid. (Original work published 2021)
• Van Rossum, E. (2027). Introduction to Fire Safety Engineering. Fire Safety Academy.
• Zalosh, R. (2002). Explosion protection. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 3-402–3-421). National Fire Protection Association; Society of Fire Protection Engineers.