Thermal runaway and thermal runaway propagation in batteries: What do we talk about?

Thermal runaway and thermal runaway propagation in batteries: What do we talk about?

Journal of Energy Storage 24 (2019) 100649 Contents lists available at ScienceDirect Journal of Energy Storage journal homepage:

2MB Sizes 0 Downloads 4 Views

Journal of Energy Storage 24 (2019) 100649

Contents lists available at ScienceDirect

Journal of Energy Storage journal homepage:

Thermal runaway and thermal runaway propagation in batteries: What do we talk about?


Alexander Börgera, , Jan Mertensa, Heinz Wenzlb ⁎

a b

Volkswagen AG, LetterBox 1695, D-38440 Wolfsburg, Germany TU Clausthal, Institut für Elektrische Energietechnik und Energiesysteme, Leibnizstraße 28, D-38678 Clausthal-Zellerfeld, Germany



Keywords: Thermal runaway Thermal runaway propagation Lithium-ion battery Semënov theory Fail-safe battery management system

Investigations describing incidents with electrochemical cells, in which the cell heats up and is either destroyed or severely damaged, causing damage to the immediate environment of the cell, are frequently referred to as “thermal runaway“. However, the term is used in most papers without a clear definition of what is meant. This article analyses in more detail what exactly could be meant by this term, and intends to offer a precision of the usage of the term “thermal runaway” which will help to find solutions for thermal runaways and their propagation and safe handling in an electrochemical system like a lithium-ion battery. In addition, the key role of a fail-safe battery management system in preventing a thermal runaway is discussed.

1. Introduction It is a well-known fact that chemical and electrochemical systems can come out of control [1]. In the case of an unstoppable self-heating, one usually talks about a thermal runaway (TR). In systems (like modern traction batteries) that consist of several sub-systems (e.g. cells or modules), the thermal runaway of one sub-system may impact the other sub-systems; in the extreme, they also come out of control. That is what can be called thermal runaway propagation (TRP). Several papers have dealt with the different aspects of thermal runaway propagation with focus on lithium-ion batteries [2–13]. Thermal runaway exists in other battery technologies as well where the effects range from catastrophic to limited damage. However, despite a wide discussion of the phenomenon [11], the underlying basic concepts are not very well defined. Often, a TR definition such as “a process of uncontrolled heat release and rapid temperature rise” [3] or “higher heat generation than heat removal” is used – but it is not difficult to show that there are drawbacks in their application, at least with respect to electrochemical systems like batteries. Most batteries heat up during use because the heat generated is higher than the heat removed, and an uncontrolled heat release with a rapid temperature rise does not necessarily lead to dangerous temperature levels and subsequent damage to the cell. In the case of lithium-ion batteries, thermal runaway propagation is

also of great practical significance due to the increased usage of lithium-ion batteries as traction batteries in electric vehicles and the corresponding danger for passengers. These batteries consist of several, often dozens of cells and if one cell has a TR problem, it might affect the others. This is what is described by thermal propagation as the “sequential occurrence of thermal runaway within a battery system triggered by thermal runaway of a cell in that battery system” [14]. However, it makes sense to differentiate even further between thermal propagation (TP) as a propagation of a thermal event and thermal runaway propagation (TRP) as the triggering of thermal runaways in adjacent cells (or modules) due to the occurrence of a first (often socalled “trigger”) runaway in a first cell. Since traction batteries contain many cells and modules to achieve the required energy and power ranges, this is a problem that becomes more and more severe with the increase of electromobility as stimulated now by many governments in the world. Therefore, attempts to unify technical regulations for electric vehicles have looked at thermal runaway propagation aspects of lithiumion batteries and have tried (but so far, without success) to find adequate test conditions [14]. This article tries to approach the thermal runaway phenomenon especially in lithium-ion batteries in a more fundamental way to path the way for a more science-based discussion for cell, module and battery development and its future standardization and regulation.

Corresponding author. E-mail address: [email protected] (A. Börger). Received 14 August 2018; Received in revised form 23 December 2018; Accepted 11 January 2019 2352-152X/ © 2019 Published by Elsevier Ltd.

Journal of Energy Storage 24 (2019) 100649

A. Börger, et al.

2. Theory The temperature in an electrochemical cell is the result of heat generation and heat exchange with the environment. Heat transfer in particular differs greatly between different points in the cell. Anisotropic thermal conductivity is a well-known phenomenon, and the temperature difference to neighboring areas depends on the exact point which is considered. In addition, heat generation differs because of inhomogeneous current density, local differences in the equilibrium voltage and the extent of side reaction currents. Many studies have measured or simulated the temperature differences between cells in a module or battery and between different points within a cell [2–13]. It is not surprising therefore, that the inside and/or surface of cells can locally reach such a high temperature that the cell and sometimes its surroundings are destroyed by thermal effects: melting, fires, explosions. In the following discussion we want to show that such incidents need to be classified according to definite criteria, although these criteria are not always easy to apply. A good starting point for discussions about TR is to leave the electrochemistry away for a moment and to look at the situation in purely chemical systems. In these, thermal runaway is a well-known and well understood phenomenon that is treated in standard textbooks for chemical engineering [1]. Semënov (for fluids) and Frank-Kamenetzkii [1,15] (for mixtures of solids, esp. dusts) theories have been worked out and give criteria for the occurrence or non-occurrence of thermal runaways. To be more precise, a thermal runaway of an exothermic reaction will occur under the following circumstances according to the Semënov theory of heat explosion [1],:

Fig. 1. Different thermal events (standard condition with fluctuations (1a and 1b), temperature curve as a result of technical interventions (2) and uncontrollable event (3)) and the threshold (4) between them.

It is clear that a catastrophic thermal event is not only the result of the available materials but also the result of chemical engineering solutions, such as cooling options and temperature and pressure control. The occurrence of a TR (scenario 3) therefore is not and should not be linked to any damage that may or may not result. In order to avoid deviations from the normal and expected behavior, risk analysis, for instance based on a differential thermal analysis, and risk reduction techniques are used in chemical engineering science [17]. The starting point for a risk analysis is an understanding of the processes which can take place, including processes initiating from all products of the reactions and reactions of the materials used in the reaction vessels. For example, if the maximum theoretical temperature of a possible reaction is higher than the ignition point of the container material, than the exothermic reaction of the container material with the surrounding air needs to be considered as well. Reactions which may cause volume expansion and gas evolution are of particular importance and the reaction with air after venting needs to be analyzed. Total energy, maximum temperature and rate of energy release all need to be known. In the chemical industry, the risks of gas evolutions and high rates of exothermic reactions which may take place, for whatever reason, are analyzed and controlled largely by the following precautions or changes in the way the process is carried out:

a) when the equilibrium between heat generation Qgen and (maximum) heat removal Qrem, max from the system under investigation is left in favor of heat generation (1)

Qgen > Qrem, max

• when at each temperature T higher than the initial temperature T , 1


i.e. the equilibrium or stationary value of the initial state, the temperature gradient of the power released by the reaction is higher than the power of heat removed.


dQ gen dT

)T1 > (

dQ rem, max )T1 dT


Heat removal in this context is the maximum heat removal capacity of the system under investigation. A thermal runaway under this definition is an uncontrollable event in the sense of ISO 31000 [16] for risk management and DIN 820-12 [17] which gives guidelines for the inclusion of safety aspects in standards. There are three general scenarios for an exothermic reaction:

- mixing of materials which causes a uniform reaction and thermal energy generation rate, avoiding hot spots, - decreasing the reaction rate using inert additives or lowering of concentrations in solutions, reducing the rate of adding educts etc., - removing reaction products from the reactor, - increasing the ratio of surface and volume to increase heat losses, - cooling of the reactor.

a) The temperature rise is expected to exceed a lower temperature limit and stay below a higher temperature threshold. The reaction proceeds as planned. b) The temperature rise exceeds a temperature threshold. The reaction is for whatever reason abnormal and actions need to be taken, where possible. This could be done by maximizing the cooling or by reducing or even stopping the addition of reactands. c) The temperature exceeds a second threshold, the reaction rate increases as a result of the higher temperature and it is no longer possible to limit the temperature rise by external measures. Temperature and pressure will reach their maximum which may be close to the theoretical maximum. Major damage may occur unless the reaction vessel has been built in such a way that the reaction is contained minimizing or avoiding any damage to the surroundings.

The onset of a thermal runaway should be stoppable by extensive cooling and allowing the reacting materials to spread out over larger areas and volumes. Engineering solutions to achieve this are an integral part of the design. Thus, applying the Semënov criteria to chemical industry reactors leads to the following detailing of thermal runaway criteria. A thermal runaway will occur when a) the generation of heat is higher than the loss of heat, and b) heat generated cannot be removed by the cooling system, causing a further increase in temperature, and c) a heat increase cannot be stopped anymore by outside interference

Fig. 1 illustrates these scenarios. 2

Journal of Energy Storage 24 (2019) 100649

A. Börger, et al.

is switched off. It is theoretically possible, that the two Semënov conditions are fulfilled during normal charging or discharging processes, because, at the end of charge, there are strong side reactions (charging with constant current leads to a thermal disaster in the case of almost all battery technologies). - especially in battery systems, the mass transport characteristics are completely different from typical chemical systems. The main driving force for mass transfer is migration (transport in the gradient of the electric field), accompanied by diffusion. Normally, there is no convection at all (i.e. no stirring), leading to large inhomogenities when the (electrochemical) reaction proceeds. Therefore, the classical chemical concepts for a thermal runaway situation like the widely used definition of TR as “a process of uncontrolled heat release and rapid temperature rise” [3,14] cannot be applied without modification, e.g. without a reference to cooling power and current flow during the event. A quite pragmatic approach to come to practical TR definitions has been used in the IWG GTR EVS (= Informal Working Group for developing a Global Technical Regulation for Electric Vehicle Safety, now also referred as GTR 20) which is written down in the current GTR 20 document [14]:

Fig. 2. Different events in dQ/dt vs. T plot (so-called Semënov plot): exponential curve is heat generation by an exothermic reaction. Line c corresponds to heat removal in a thermal runaway case and the line a corresponds between points 1 and 3 to heat removal in a safe case. Line b is the borderline case (the line is a tangent to the heat generation parabola). The safe area (corresponding to the safe zone in [3]) is the dashed area.

since internal heat generation releases more heat than can be removed.

Detection of thermal runaway. Thermal runaway can be detected by the following conditions: (i) The measured voltage of the initiation cell drops; (ii) The measured temperature exceeds [the maximum operating temperature defined by the manufacturer]; (iii) dT/dt ≥ [1 °C/s] of the measured temperature. Thermal runaway can be judged when: (a) Both (i) and (iii) are detected, or (b) Both (ii) and (iii) are detected.

A visualization of the three different scenarios described above in terms of the Semënov theory is given in the so called Semënov-plot (Fig. 2). A Semënov-plot shows the heat generation dQgen/dt and heat removal dQrem/dt versus temperature. Exothermic reactions have an exponential form and heat removal is usually given as a linear function of the temperature difference between the temperature of the reactants and ambient temperature corresponding to heat removal by conduction or convection. The safe zone for an exothermic reaction following Arrhenius law is characterized by these conditions for any temperature Tsafe where heat generation is lower than or equals heat removal and which does not yet lead to any damage, except possibly a destruction of an individual cell without damage to its environment1. The occurrence or non-occurrence of a thermal runaway therefore depends not only on the heat of reaction but also on the engineering solutions to remove a sufficient amount of heat.

It is interesting to note that these criteria give no guidance to detecting the beginning of a TR at a time, when countermeasures are still possible and effective. That these conditions are incomplete can be shown by assuming one string of cells with one cell of high resistance. Assuming 100 cells in series, the current will drop to nearly zero and the voltage of the low resistance cells will be only minimally above their equilibrium voltage. The high resistance cell will have a voltage in the range of 20 V with a small current flowing. As long as the cooling system operates, the temperature gradient may be low. However, eventually the cell will vent and may cause a fire, if the gases and particles from the inside of the cell are not prevented from igniting. The high resistance cell however will eventually reach very high temperatures even in a well cooled system, but not necessarily fulfill either condition (i) or (iii) of the above criteria. To exclude such an event from a safety oriented discussion of a TR is inappropriate. Another approach of the definition for a TR is the usage of the EUCAR hazard level (HL) classification, i.e. considering a TR has only happened in case something dangerous has occurred (fire, flame, rupture, explosion) [18]. The general drawback of a “pragmatic” or “technical” usage of the term thermal runaway is simply that it is misleading since it creates a false sense of safety. In fact, it can even be more dangerous to use these criteria since they lead to detection mechanisms that look only at these criteria and thus they can lead to non-detected thermal runaways because dangerous real-world situations are not fully covered by them. Since safety is of highest concern with lithium-ion technologies and especially, their potential use as a new basis of mobility, a fully comprehensive treatment of thermal runaway and propagation is necessary. Thus, the focus of the usage of the term “TR” in the current GTR EVS document is detection, not avoidance and understanding the underlying processes. In the light of the above discussion, the GTR EVS definitions, depending on the values chosen, describe either the occurrence of an unusual situation calling for emergency measures to be taken or the

3. Discussion 3.1. Application of the general Semënov theory to batteries Unfortunately, the definitions used in chemical engineering cannot be transferred to electrochemical cells and especially batteries without further analysis. This is due to the following reasons: - the materials that can release large amounts of energy are separated, i.e. there are two electrodes and in most cases a separator. Conditions regarding the occurrence of a TR are different for each of the two separate electrodes and the two electrodes in contact with each other. - electric current plays a major role. External currents can be switched on and off with an immediate effect on heat generation in a cell, in contrast to the curves shown above. The temperature can have abrupt changes and discontinuous values of dT/dt when the current

1 The difference between a TR and a thermal propagation event are characterized by Tsafe. At a lower Tsafe, the destruction of an individual cell without damage to the immediate environment (neighboring cells) may occur, at a higher Tsafe thermal propagation may take place.


Journal of Energy Storage 24 (2019) 100649

A. Börger, et al.

convection, conduction and heat radiation2 (with as heat conduction transfer coefficient, A as effective cell (or battery) surface, Tsurf as surface temperature, Tamb ambient temperature , h as heat convection transfer coefficient, t as time, as emissivity of the cell surface and as Stefan-Boltzmann constant):

onset of a process which cannot be stopped any more. This can also apply to cases where the damage to the cell is small, the environment is not affected, the process can be managed by current control and cooling engineering. The same abnormal exothermic reaction either falls under the definition of thermal runaway, possibly causing catastrophic events, or does not fall under the definition because of high cooling power and other good engineering solutions on a systems level. To analyze the process and target options for handling thermal events, however, a more elaborate approach would be desirable using stability criteria as in the case of pure chemical systems. A first step has been made by the introduction of the safe zone concept by Barnett et al. [3] who found out in their simulations that there is always a zone of physical parameters that do not allow a thermal runaway to take place at all (cf. explosion theory where beyond the upper or the lower explosion limit, explosions cannot take place either). Of course, a TR can still only occur when more heat is generated than the system is able to cope with. In lithium-ion battery systems, the causes of heat generation are much more diverse than in the case of purely chemical systems and require the consideration of (at least) the following aspects:

Qrem, max = A +


d lnQrem, max dT

+ hA

d (Tsurf



4 Tamb )




Qgen = Q1 + Q2 + Q3 + Q4 + Q5

with Q1 as external heat flux and Q5 as sum of all other heat fluxes,

Q2 =

I ²Ri



with I as current, Ri as one of several resistances (e.g. from electrolyte, current collectors, etc.),

Q3 = U Ai 0i (exp( i

i zi F i





i) z i F i




with U as cell voltage, A as effective electrode surface, i 0i as exchange current density of the i-th reaction, i as the corresponding transition factor for the i-th reaction, z i as number of exchanged electrons in the ith reaction, F as Faraday constant, i as overpotential of the i-th reaction, R as universal gas constant, and

Q4 =


VBatt k 0i (

R Hi )exp(



) cij j


with VBatt as battery volume, k 0i as rate constant of the i-th reaction, cij as concentration of the j-th component in the i-th reaction, mij as reaction order of the j-th component in the i-th reaction, R Hi as reaction enthalpy of the i-th reaction, andE Ai as activation energy of the i-th reaction. 3.2. Classification of TRs A closer inspection of the different contributions to heat generation makes clear that different cases should be considered, e.g. - the contribution of external currents which can be terminated by means of a master switch, - the contribution of one high internal resistance cell which is a function of the number of cells in series and in parallel, and - the contribution of an external heat source Q1 which will differ greatly between individual cells and cells in a module. Depending on the aspect to be investigated, three different groups of categories for defining a TR may be useful based on a) phenomena which are observed, b) root causes and c) countermeasures that can be taken. The possibility to test different TRs allows a separate classification (see section 3.4 below). Please note that none of the definitions proposed here takes into account whether damage will be caused outside the cell or not. Please note also that the classification presented here is not exhaustive because, depending on the actual battery design, root causes of internal short circuits (see above) and other aspects can also be used as (even more detailed) classification categories.

Thus, an analogy to the criteria in the Semënov theory for the border of the safe zone would be – in analogy to [19] – (after combining Eqs. (1) and (2) for Qgen = Qrem, max in the following general expression (3)):




dt 4 d (Tsurf

while the maximum heat production rate is given by the following expression:

1 Heat flux from the surrounding Q1: Heat radiated, convected or conducted from nearby components (i.e. an engine in a hybrid electric vehicle) can cause damage and/or initiate processes leading to a further temperature increase. This includes fires for whatever reasons, such as ignition of gases vented from a cell in air. In cases of intentional external battery heating (as used in traction batteries for commercial vehicles esp. in cold countries), this additional heating has to be considered, too. In the case of battery cooling, this is a possible countermeasure against thermal runaways. 2 Heat flux from current flow Q2 as part of normal operation or faults in the battery management system (BMS) and its parametrization. This term includes heat from e.g. ohmic (i.e. electrical resistance) heating due to the internal resistances of the electrolyte, the electrodes, the wires etc. which is omnipresent in electrochemical systems, and also heat from high currents caused by external short circuits and heat from high currents by internal short circuits caused by dendrites, mass particles (i.e. manufacturing defects) or other causes such as contact between the electrodes (a: current collector to other current collector or b: current collector to active mass or c: active mass to active mass) etc. 3 Heat flux from electrochemical reactions Q3 , i.e. the main reaction, anodically and cathodically coupled electrochemical reactions on one electrode, and side reactions during overcharging and high currents at very low voltages which lead to significant temperature rises. 4 Heat flux from purely chemical reactions Q4 , e.g. oxidation of anode materials from oxygen evolution on the cathode or air from the surrounding. In general, we have two separate terms, an enthalpic and an entropic one. (Due to temperature rises, also endothermic reactions can take place at higher system temperatures if they are entropy-driven (i.e. due to the formation of gaseous molecules) and can lead to cooling instead of heating.) Nonetheless, to keep consistency with standard textbooks [1], we include the entropic part in 5. 5 Other sources of heat Q5 , e.g. by volume expansion of non-ideal gases or by the Joule-Thompson effect etc.

d lnQ gen

d (Tsurf


2 Obviously, heat radiation becomes more and more important, the higher the actual cell temperature is.

where Qrem, max is given by the sum of the terms corresponding to heat 4

Journal of Energy Storage 24 (2019) 100649

A. Börger, et al.

circuit as originating from a contact between anode and cathode, anode and cathodic current collector, cathode and anodic current collector etc.

3.2.1. Phenomena This classification describes different processes leading to a TR: (i) a battery thermal runaway in the narrow sense is given in a case where the (external) current has been shut down to zero by means of a switch or fuse but the cooling power of the residual system is not able to remove enough heat from the system to stop the temperature from rising more and more. This class of TR corresponds to good, fail-safe battery management systems. (ii) a battery thermal runaway in a wider sense (or: in a practical sense) is given in a case where the external current is still flowing and the cooling power of the complete system is not able to remove enough heat from the system to stop the temperature from rising more and more (iii) a third situation, between (i) and (ii), is given in the case that at a certain time or temperature, the external current will be reduced to zero or near zero. This is e.g. the case when a shut-down separator in the cell limits the maximum internal current and temperature rise. A thermal runaway in this sense (termed: current-reduced thermal runaway) will occur when the point at which the current shut-down occurs comes too late for the system to avoid that the temperature rise becomes unstoppable. (iv) an even more complex situation might be given when the flowing current is outside the specification. This could be due to e.g. faulty or wrong charging characteristics or unintentional abusive conditions. Since wires, current collectors etc. only support certain current values without being destroyed (e.g. by melting), one has to consider a situation with the maximum current that is technically possible. Characteristic for these kinds of thermal runaways are at least one or even several dT/dt discontinuities, one for each component becoming non-conductive. (v) a purely heat source driven thermal runaway, i.e. a heat source from outside (e.g. a fire) produces a thermal runaway without electric currents playing a role. Since the theory for chemical reactions can be used for this situation, it will not be in the focus of this article.

3.2.3. Countermeasures It is well known from the literature that thermal runaways only occur when a specific temperature threshold is reached [3]. Thus, there are thermal runaways that can be stopped during the temperature rise if additional cooling mechanisms are used that may start only in the case that a temperature rise is detected. These may or may not be accompanied by a current shut-down. This leads to the following classification: (i) thermal runaways starting at temperature and current conditions within the permitted range without any countermeasures being taken or possible, (ii) thermal runaways stoppable by immediate current shut-down, (iii) thermal runaways only stoppable by immediate current shut-down and additional cooling measures. Especially the case of a well-designed and operating battery management system (BMS) will prevent many incidents of occurring, e.g. a BMS will increase cooling as much as possible, preventing the usage of a battery at temperatures with the highest possible margin below the critical temperature for a thermal runaway etc. In this context it is also necessary to distinguish between fail-safe BMS, which will ensure safe operation even if it becomes faulty and those BMS which will not be able to prevent TR if one (or several) of its functions is faulty or becomes unavailable. 3.2.4. Root causes and countermeasures: a matrix of combinations Combining root causes and countermeasures leads to the following matrix (see Table 2): This matrix is useful to focus engineering and development work on minimizing the risk of a TR taking place.

These different types of thermal runaways are summarized in Table 1.

3.3. Stability criteria The general formula (5) can be simplified in order to come to welldefined stability criteria that lead also to a cascade of special safety measures that can be used to show safe system behavior. a) The term Q2 can be limited by defining the maximum resistance and current that is technically possible in the system under consideration. Detailed values can probably only be established by designing appropriate test procedures. However, estimates based on theoretical considerations can be made. Rm, (T,I) is determined by the maximum resistance at the maximum possible current. This resistance may be dependent on temperature and current. In most cases, the maximum current Imax will be given either by the self-heating and thus melting of the current collectors, or by the mass-transport limitation (i.e. the diffusion resistance RD for the application of a corresponding current I against the diffusion overvoltage D ). In the first case, the maximum current can be calculated by setting the heat generation from the maximum current equal to the maximum heat loss by conductive processes:

3.2.2. Root causes of thermal runaways For design purposes and development a classification of the possible reasons for the occurrence of thermal runaways may be useful: - abuse conditions (i.e. penetration by external objects like nails or a faulty BMS), - an inherent faulty system design, e.g. by use of materials which have a too narrow window of operation or a wide variation of specified properties which makes the occasional cell dangerous, - the introduction of undetected defects during the manufacturing process (e.g. mass particles), - ageing conditions, producing e.g. lithium plating, dendrites or pinholes in the separator. Please note: according to the concrete internal set-up of the battery, the root causes can be detailed further, e.g. by defining an internal short

Table 1 Phenomenological classification of different types of thermal runaways and their description. Type of thermal runaway



TR in the narrow sense TR in a wider sense TR with reduced current TR with current out of specification purely heat source driven TR

|I| = 0 during TR |I| > 0 during TR |I| 0 during TR |I| 0 during TR |I| 0



after nail punching in a system with a faulty BMS that does not detect temperature rises in systems with shutdown separator caused by external or internal short circuits or faulty chargers by fire

Journal of Energy Storage 24 (2019) 100649

A. Börger, et al.

Table 2 Matrix of thermal runaways. non-stoppable Abuse Faults Defects Ageing

non-stoppable non-stoppable non-stoppable non-stoppable

Imax =



by by by by

abuse (Ia) faults (Ib) defects (Ic) ageing (Id)


stoppable by current interruption

stoppable by current interruption and additional cooling

current-stoppable current-stoppable current-stoppable current-stoppable

cooling-stoppable cooling-stoppable cooling-stoppable cooling-stoppable


zFADcEl . yt (10)


with cEl . yt as concentration of the electrolyte, D as diffusion coefficient and N as thickness of the diffusion layer. This can be combined with the (diffusion) overpotential to give the maximum resistance in the case of diffusion limitation:

Rmax =

d D dImax


Tamb )

Qrem = Qchem + Rmax Imax 2


Tamb ) dt

Qrem = Qchem + Rmax IISC 2



where IISC corresponds to the current that will flow in the case of an internal short circuit (ideally, it is even 0 so that this summand disappears or it affects only a few cells). An interesting feature of Eq. (12) and (13) is that they allow a sequence of safety measures, e.g.:

For all tests appropriate temperatures and SOC at the beginning of testing need to be determined. Using these test procedures, the decisive information about a TR is simply its ability to cause external damage. No part of the surface may exceed a dangerous temperature limit which will lead to damage to any outside component, the geometrical shape must be unchanged and none of the gases emitted may cause either toxic or corrosive effects and may not ignite. Whether a cell or battery can still be used after such an event is irrelevant.

a) if the heat removal is not enough, cooling possibilities have to be enhanced; b) if the heat capacity is not enough, it has to be increased; c) if the ohmic heat is too high, it has to be reduced by system measures to 0 (i.e. by using shut-down separators) or it has to be reduced by reducing the maximum current (i.e. by lowering the melting points of current collectors) or the maximum resistance for the components outside the cell. d) Furthermore, it may become important to analyze side reactions on both electrodes in greater detail than so far considered necessary to limit the heat released by them, e.g. by finding

3.4.1. TR driven by external current The information to be established is: At which combination of current, temperature and voltage and after what time of exposure to these conditions is it necessary to switch off the external current so that no catastrophic behaviour takes place. The properties under high rate discharging and charging conditions need to be established for this. For discharging, externally caused short circuit currents at different SOC and temperatures need to be investigated, and for charging, the maximum charging power (voltage and current of power electronics) need to be taken into account. It is necessary to differentiate between individual cells and modules/battery systems as cells in modules may suffer much more severe conditions than individual cells depending on the nominal voltage of

da) additives that reduce the exchange current density etc., db) additives that remove heat because at higher temperature, they undergo endothermic reactions that are entropy-driven at higher temperature according to the Gibbs equation RG





abuse (IIIa) faults (IIIb) defects (IIIc) ageing (IIId)

(i) TR driven by external current where a reliable BMS with appropriate settings prevents a TR by switching off the external current sufficiently early. It may be appropriate to differentiate between standard BMS and BMS with a fail-safe characteristic. (ii) TR driven by short circuits which cannot be switched off by switches or fuses. (iii) TR driven by external heat sources. (iv) It is further suggested to treat mechanical damage and the resulting dangers separately.

In the case that the current is switched off by the system controller or by the system design, this reduces to:


by by by by

It would be desirable if test conditions and groups of countermeasures would be drawn up to prevent or reduce the risk of a thermal runaway taking place. In principle, the twelve different combinations of thermal runaway and countermeasures according to the above matrix in Table 2 require each their own test scenarios. The different root causes do not lead necessarily to special test conditions because they depend on the devices under test (DUTs). Four different test categories are suggested:




3.4. Resulting test possibilities

Rmax for the case of melting collectors is more complex and cannot be given without a detailed analysis of all factors. An estimation is the metallic resistivity of the material at the melting temperature. b) effects of heat radiation, entropic (i.e. effects from changes in the entropy of the reactants, the S term), Joule-Thompson effects (with expanding gases), mixing effects and others are neglected. This leads to the following expression for the limits of the safe zone [3]: CP

abuse (IIa) faults (IIb) defects (IIc) ageing (IId)

with R G as free reaction enthalpy, R H as reaction enthalpy and R S as reaction entropy, dc) additives that undergo side-reactions that change the thermal properties of the cell inside (e.g. by polymerization), dd) additives that influence the thermal and/or electrical properties otherwise. The consequence of all this is that it allows and accounts for a cascading of the thermal propagation mitigation measures to be taken and it accounts for a sequence of possible tests to confirm (or disprove) that the required level of safety has been reached. A possible sequence of providing evidence for reaching the safe zone is given in Fig. 4.


where corresponds to the thermal conductivity of the material, κ is the specific resistance of the current collector, Tf is the melting point of the collector and Tamb to the temperature of the environment. Imax for the case of mass-transport limitations is given by

Imax =

by by by by

(14) 6

Journal of Energy Storage 24 (2019) 100649

A. Börger, et al.

will be available during the whole test?

Table 3 Classification of some exemplary TR and TRP test data taken from the literature. The references marked by an asterisk correspond more to rapid cycling than to actual ageing. test category


(a) abuse (b) faults (c) defects (d) ageing

[20] [21] [22] [23] [24] [25] [26]* [27]*

3.4.4. TR driven by other reasons, esp. mechanical impact The above three groups of tests do not refer to mechanical damage of cells or modules which are a separate issue. Crushing cells or modules in such a way that a catastrophic thermal event will take place is probably easy, as it is easy to define tests to make a petrol tank explode. However, only those tests need to be defined which may occur under real conditions. 3.5. Classification of existing data according to the above classification

the string and the number of strings in parallel. Individual cells with a high internal resistance in series with other cells may be discharged to negative voltage levels during a high rate discharge and experience much higher overvoltages during charging. It may be appropriate for testing of modules or battery systems to select cells with much higher internal resistance and lower capacity than others to take premature ageing or deviations of cell properties as a result of manufacturing into account.

A considerable amount of TR, TRP and TP test data have been published in the literature. Table 3 classifies examplary test data according to the classification scheme that has been developed above. Although examples for the different test categories can be found, many researchers concentrate on either abuse tests or tests that simulate internal defects. The effects of ageing are less well investigated, although in the field it is most likely to expect TR or TRP events with aged cells and batteries. Also, most examples published concentrate on triggering a TR and investigating its effects and potentially their TRP. The potential benefits of mitigation measures and their limits (i.e. the maximum cooling rate to prevent a TRP) need also further investigation. When more examples for the different test categories become available, the full matrix from Table 2 could become filled. However, one normally will be interested in the outcome for a given system configuration including all TR and TRP mitigation measures (i.e. cooling, current interruption etc.). Thus, a set-up of 12 different TR and TRP tests is not expected to become a standard procedure. However, Table 3 makes clear that a search for a one-and-only test procedure is neither reasonable nor possible due to the different TR categories (Table 4).

3.4.2. TR driven by short circuits Short circuits cannot always be terminated by an external switch or a fuse. External short circuits may occur for groups of cells if there is no fuse between the group of cells. Obviously internal short circuits cannot be terminated by an external switch or fuse and not necessarily by internal fuses or shutdown separators. The information to be established is: What are the risks of a cell in its protective environment (e.g. internal fuse, venting, enclosure of cell with fire proof materials) suffering a short circuit? What are the risks of one and more cells directly connected to each other in a module suffering from a short circuit? For such tests it would be necessary to establish what the minimum short circuit resistance can be in the case of an individual cell and in the case of cells within a module. The assumption that the power and energy released by bridging the terminals of cells or modules with a zeroresistance connection will always exceed the power and energy of cells with an internal short circuit needs to be verified.

3.6. Limitations of test possibilites However, the test possibilities are limited by several factors such as 1) lack of suitable (i.e. reproducible, reliable or simply feasible) test methods for obtaining data that are necessary to perform an analysis like those shown in Fig. 3, 2) lack of suitable test devices due to e.g. extensive cost, 3) overlapping of several phenomena (e.g. chemical reactions) in one test without possibilities to separate their effects.

3.4.3. TR driven by external heat sources The information to be established is, how long a cell or module can withstand thermal power before a TR takes place and which voltage and temperature measurements from a BMS are capable of providing warning signals before the onset of a TR. Cells and modules need to be tested in their protective environment which may be extremely limited for individual cells but quite extensive in the case of cells in a module. In the case of external cooling and heating systems their functioning during external heating needs to be clarified. Is it, for instance, realistic to presume that no cooling liquid circulates in the module at the beginning of testing and is it realistic to presume that full cooling power

Of course, in certain cases, modelling might assist in obtaining the desired data. But also modelling has its limitations. So there will remain an area that cannot be covered directly by tests. That is the reason why complimentary solutions such as a formalized, documented approach [12] using industry standards like ISO 26262 [28] for functional safety or the hazard mode and risk mitigation analysis [29] as already referenced in SAE J2464 [30] also have their justification and a sound

Table 4 Summary of TR and TRP tests and possible limitations for them. TR category driven by external currents driven by short circuits driven by external heat sources driven by mechanical impact

possible test set-ups a) b) a) b) a) b) c) a)

overcharge test cyclic aging … external short circuit test internal short circuit test … overheating using a microheater chemical heating laser-heating … nail penetration (ceramic or metallic nail) b) crushing c) shock impact …

test limitations/drawbacks BMS functionality must be known and tests could be easy to tamper with special DUTs are necessary which might be very difficult to produce, manage and move very difficult to find a heating source location and design that generally fits all possible battery designs and gives unambiguous results reproducibility is low because in general, more than one contact between anode and cathode or corresponding current collectors will result


Journal of Energy Storage 24 (2019) 100649

A. Börger, et al.

Fig. 3. Possible sources of thermal runaways and their consequences including a classification of events as thermal runaways (following [18]) only in case a significant damage or harm has been done (prot. act. = passive protection activated).

Fig. 4. Sequence of application of Semënov stability criteria and corresponding tests.

combination of test data and documented risk detection and mitigation analyses will be necessary to solve the battery TR and TRP problem for most cases.

it is quite probable that science-based and practical (as in IWG GTR EVS) definitions will both continue to be used. As long as everybody knows what he talks about, this is not a problem – this contribution intends to intensify the discussion and to broaden the understanding in the field.

4. Summary and outlook Defining several types of thermal runaways for batteries and analyzing the heat generation situation in (lithium-ion) batteries gives a much better insight in what we talk about when we discuss undesirable thermal events in lithium-ion batteries. This together with a closer inspection of heat generation and removal gives more science-based criteria for the discussion about thermal runaway phenomena. Nonetheless, since test results require clear and fast evaluation criteria,

Acknowledgments We thank Dr. Ralf Benger (EFZN Goslar) for discussions as well as Jacob Klink and Anna Meyfahrth (both TU Clausthal) for assistance during the literature collection.


Journal of Energy Storage 24 (2019) 100649

A. Börger, et al.


TRANS-WP29-GRSP-2017-02e.docx (29.01.2018). [15] V.M. Schmidt: Elektrochemische Verfahrenstechnik, Wiley/VCH, Weinheim, 2003 [in German]. [16] ISO, Risk Management - Principles and Guidelines (ISO 31000:2009), (2009). [17] DIN 820-12:2014-06: Normungsarbeit - Teil 12: Leitfaden für die Aufnahme von Sicherheitsaspekten in Normen, (2014) [in German]. [18] D. Hoffmann, Sicherheit von Lithium-Ionen-Batterien, in: E. Fahlbusch (Ed.), Batterien Als Eenrgiespeicher. Beuth-Verlag (Berlin), 2015 p. 265. [in German]. [19] Q. Wang, P. Ping, X. Zhao, G. Chu, J. Sun, C. Chen, J. Power Sources 208 (2012) 210. [20] P. Huang, P. Ping, K. Li, H. Chen, Q. Wang, J. Wen, J. Sun, Appl. Energy 183 (2016) 659. [21] X. Feng, M. Fang, X. He, M. Ouyang, L. Lu, H. Wang, M. Zhang, J. Power Sources 255 (2014) 294. [22] D. Ren, X. Feng, L. Lu, M. Ouyang, S. Zheng, J. Li, J. Power Sources 364 (2017) 328. [23] D. Ouyang, Y. He, M. Chen, J. Liu, J. Wang, J. Therm. Anal. Calorim. 132 (2018) 65. [24] P.T. Coman, E.C. Darcy, C.T. Veje, R.E. White, J. Electrochem. Soc. 164 (2017) A587–A593. [25] S. Santhanagopalan, P. Ramadass, J. Zhang, J. Power Sources 194 (2009) 550. [26] W.A., Hermann, N.P., Manov, P., Karplus, Moraga, A., Prilutsky, Method and apparatus for electrically cycling a cell to simulate an internal short, US Patent US 8. 421,469 B2 (2013). [27] L.S. Guo, Z.R. Wang, J.H. Wang, Q.K. Luo, J.J. Liu, J. Loss Prevent. Process Ind. 49 (2017) 953. [28] ISO, Road Vehicles – Functional Safety (ISO 26262:2011), (2011). [29] C.N. Ashtiani, ECS Trans. 11 (2008) 1. [30] SAE, Electric and Hybrid Electric Vehicle – Rechargeable Energy Storage System (RESS): Safety and Abuse Testing SAE J2464, (2009).

[1] J. Steinbach, Chemische Sicherheitstechnik, Wiley/VCH, Weinheim, 1995 [in German]. [2] P.G. Balakrishnan, R. Ramesh, T. PremKumar, J. Power Sources 155 (2006) 404. [3] B. Barnett, D. Ofer, S. Sriramulu, R. Stringfellow, Lithium-ion batteries, safety, in: R.J. Brodd (Ed.), Batteries for Sustainability. Selected Entries from the Encyclopedia of Sustainable Science and Technology, Springer, New York, 2013. [4] D.H. Doughty, see: Batteries_FRM-Panel1d-Doughty.pdf (29.01.2018). [5] H. Lundgren, P. Svens, H. Ekström, C. Tengstedt, J. Lindström, M. Behm, G. Lindbergh, J. Electrochem. Soc. 163 (2016) A209. [6] H. Maleki, J.N. Howard, J. Power Sources 191 (2009) 568. [7] C.J. Orendorff, E.P. Roth, G. Nagasubramaniam, J. Power Sources 196 (2011) 6554. [8] S. Santhanagopalan, P. Ramadass, J. Zhang, J. Power Sources 194 (2009) 550. [9] R. Spotnitz, J. Franklin, J. Power Sources 113 (2003) 81. [10] T.G. Zavalis, M. Behm, G. Lindbergh, J. Electrochem. Soc. 159 (2012) A848. [11] A. Ahlberg-Tidblad, Thermal Runaway and Propagation in Li Ion Batteries Literature Review, see: (2018) (29.01.2018), rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved= 0ahUKEwjpo5W30tTYAhWEEiwKHa8OBJkQFggoMAA&url=https%3A%2F pdf%3Fapi%3Dv2&usg=AOvVaw0wpoDIC5nPbUPXI75_TWmF. [12] C.F. Lopez, J.A. Jeevarajan, P.P. Mukherjee, J. Electrochem. Soc. 162 (2015) A2163. [13] C.F. Lopez, J.A. Jeevarajan, P.P. Mukherjee, J. Electrochem. Soc. 162 (2015) A1905. [14] see