Fire Hazards Associated with the Use of Water and Glycol as Coolants for Li-Ion Battery Systems

As society moves towards increased electrification, the fire hazards related to use of Lithium-Ion Batteries (LIB) are given increasing attention. With expanded use of LIB in applications such as electric vehicles (EV) and Battery Energy Storage Systems (BESS), also the number of incidents increase. LIBs are generally safe energy carriers with a failure rate estimated to be about 1 in 10 million cells [1]. However, internal cell failure is not the only potential source of incident, but failure can originate from the control system or any of the physical elements of the system, such as busbars, cabling, transformers, or liquid cooling systems [2]. Using BESS as an example, information on incidents are collected in the BESS Failure Incident Database [3] where today 87 BESS fire incidents are included. In an analysis of 26 incidents, it was found that only 3 (11%) originated in the cells or modules, while 12 (43%) were related to the Balance of System (BOS), i.e. the surrounding physical systems. Out of these 12 incidents, 9 occurred during integration, assembly and construction of the BESS. The 2021 incident in Australia at the Victoria Big Battery facility is an example of BOS failure due to leaking coolant as a result of assembly quality issues [4]. In an EPRI report from the present year, it is concluded that “Integration-related failures have become more common.”, and the vast majority of these failures are related to the BOS and components like AC or DC wiring, cooling systems, or safety systems such as water suppression piping [2].

The cooling side of the battery system has many interfaces and seals for coolant lines inside the battery casing, where also DC/DC converters and other high-voltage components are placed. Also, the coolant itself can be aggressive to material, resulting in cracks appearing in the seals. Coolant leakage can occur early in the lifetime of the battery system, as a result of assembly failure, or later due to ageing of materials.

Post-fire analysis of battery systems is challenging due to the simple fact that the evidence has been consumed in the fire, and it is often difficult to identify the root cause of the incident. One example where the cooling system was indeed identified as the highly likely cause of the fire was onboard the ship “MF Ytterøyningen” in Norway in 2021, where leaking coolant resulted in a short circuit or arcing [5]. Hours after the fire was extinguished, there was a severe explosion from the same battery, and it has been speculated that it could be a result of arcing triggered by water from the sprinkler system. There are also anecdotal descriptions from our industrial partners who experienced explosions and subsequent fires in LIB packs due to leakage in the coolant system before a thermal runaway (TR) was triggered in the experiments. These incidents resembled gas explosions, which indicate rapid formation of gases prior to gas release from the LIB cell itself.

Considering the above indications that leaking cooling liquid may cause battery fires and gas explosions, the authors would like to formulate the following hypothesis:

The electrical environment in a LIB system may result in that criteria for dissociation of cooling media into hydrogen and other flammable gases are fulfilled.

In this Short Communication, we outline available information about the common coolants water and ethylene glycol (EG), the electrical environment of a typical LIB in an EV, and conditions for electrically induced dissociation of water and EG.

Water is a good cooling medium due to its high thermal conductivity but has the drawback of a high freezing point. Glycol was first synthesized in the late nineteenth century and gained wide use as a cooling and anti-freeze agent already in the 1920’s. Glycol’s properties as a heat transfer fluid and its ability to lower the freezing point of water made it useful to mix with water, commonly in 50/50 mixtures.

The most commonly used glycol for cooling is ethylene glycol (EG) with the chemical name ethane-1,2-diol (C₂H₆O₂), a small hydrocarbon with two alcohol functionalities. The main purpose is to extend the operational temperature range of the cooling system, but a drawback of EG compared to pure water is its lower thermal conductivity. Heat transfer properties of glycol can be improved by addition of surfactants or nanofluids, the latter can be metal oxides or nano-based carbon materials. The drawback of nanofluids is that they enhance the electrical conductivity [6].

Glycol-based coolants are agressive to many materials and known to degrade some metals for example copper and brass, elastomers and some plastics. EG in mixtures with water tend to produce acids over time, which is one of the main reasons for material degradation. Leakage of coolant can thus result from materials degradation, like penetration through porous alloys or drying and subsequent cracking of plastics.

The traditional use of EG as a coolant for combustion engines has been considered relatively safe from a fire perspective. In a thorough analysis based on a combination of literature review and experiments, Hull et al. [7] concluded that EG can indeed be combusted in a vehicle fire, but it will not be the first fuel ignited. Several highly unlikely criteria would have to be fulfilled simultaneously for EG to ignite. They do bring up the hypothesis that an electrolysis process producing hydrogen could act as a cause of fire, but, due to the lack of high enough voltages in a regular combustion engine, this cause of fire was ruled out.

The potential risk of using EG as a coolant or de-icing liquid for aircraft or spacecraft was investigated after the Apollo accident, and it was found that a DC voltage, in the range of 20 to 30 V, from a silver-covered copper anode can result in electrolysis of EG [8‐10].

Modern energy carriers such as LIB and fuel cells require cooling just like combustion engines. Among different cooling strategies it has been found that indirect liquid cooling with a coolant like EG is most efficient from aspects of design and energy consumption [11]. However, these modern systems operate with high voltages making the electric environment quite different compared to combustion engines.

External short circuits of a single LIB cell can result from physical contact between cell tabs or via electrically conductive solid or liquid. An efficient external short circuit will occur at low resistance and lead to rapid discharge of the battery cell, under which heat is produced [12], which is considered a common cause of TR [13]. The main concern raised by battery manufacturers related to the use of liquid coolants for LIB-systems, is that leaking coolant can result in external short circuit, which has promoted coolants with low electrical conductivity. Therefore, present-day coolants are designed to not be conductive enough to create a rapid short circuit of single cells, but they can result in a weak current through the liquid and a local production of gases through electrolysis. Regarding conductivity, it is also plausible that, in case of coolant leakage, particles in the air or on surfaces are dissolved in the liquid, increasing the conductivity. At high currents, ohmic heating results in a temperature increase, which increases the conductivity of the liquid and thus results in even higher current. In this environment electrolysis of water and EG can occur rapidly, as further discussed in the next section. An even more extreme possibility is an electric breakdown of liquid or gas between two points or surfaces acting as electrodes, i.e. when a conductive plasma channel form an electrical connection, leading to formation of an arc [14].

Electrical conductivity through a gas or liquid between electrodes, whether it is a modest current or a rapid discharge, depends on a number of parameters: electrode materials, applied voltage, gap distance, electrode size and shape, temperature, pressure, and conductivity of the liquid.

While electrolysis of coolant was not considered a likely risk scenario in combustion engine vehicles [7], our hypothesis stated earlier is that the electrical environment in an EV can fulfill the criteria for dissociation of water or EG into hydrogen and other flammable gases, which can occur via electrolysis. For this to happen the liquid needs to conduct electricity, which indeed occur to some extent even with coolants of low conductivity. An important complication related to conductivity of coolants is, as mentioned previously, that the conductivity can be expected to increase during operation (depending on system material) and when exposed to external environment as in a leakage. The conductivity is temperature dependent, which means that if the internal temperature increases in the LIB cell, or if it is exposed to a warm external environment, the electrolysis rate may increase. In addition, when a current flow through the liquid, ohmic heating can result in reduced local resistance and, thus, further increase current.

The slow gas production can be easily demonstrated by simply immersing a regular 3.6 V cylindrical battery in water or water/EG. Within seconds or minutes, small bubbles are formed at both anode and cathode. For the case of water, the underlying theory is simple: two electrons are required to split a water molecule, and gas forming is H₂ and O₂ at the cathode and anode, respectively. Electrolysis of EG is more complex and, following the mechanism for ethanol reforming, it may result in formation of significant amounts of H₂ but also hydrocarbons such as C₂H₄, depending on the conditions [15].

While the formation of hydrogen and potentially other flammable gases happen via electrolysis, the amounts and the formation rates are likely not sufficient to induce the gas explosions observed. Figure 1a symbolize the electrolysis in a narrow free volume between conducting plates, filled with water, where small amount of gas bubbles is formed via electrolysis.

Bild vergrößern

The free volumes are small in the battery pack, with distances between conducting surfaces potentially being in the mm size range. Conductive media in the free volumes can therefore, beside the electrolysis described above, also induce the formation of plasma known as arcing [16]. Arcing is well described for gaseous media but can also occur in water which has since long been described in the literature [17] and illustrated in openly available videos [18]. The local production of gas through electrolysis, filling the gap between electrodes with small bubbles will change the potential for electric breakdown, increasing the risk of formation of a high-temperature plasma [14]. What is called the “bubble mechanism”, as illustrated in Fig. 1b, result in a localized reduction in liquid density making electron multiplication possible. Joshi et al. [14] point at local heating and evaporation, or chemical (electrolysis) production of gases, as sources of bubbles. The result of the subsequent electric breakdown is complete gassing and potentially a gas explosion, illustrated in Fig. 1c.

With voltages in the range 400 to 800 V in modern EVs, the discharge power can reach MW levels, with currents of thousands of amperes [12, 16]. Ledinski et al. [16] performed calculations, following the methodology by Doan [19], on a battery pack with 3.6 V and 66 Ah pouch cells arranged with 2 cells in parallel and 110 in series, giving a total of 396 V. Assuming an internal resistance of 1 mΩ per cell, they determined a maximum discharge power of about 720 kW.

While extensive and important research on thermal runaway is currently performed at many research institutions, not least illustrated by this special issue, we argue that the risks associated with the battery support systems, and in particular the cooling system, have not received the attention it warrants from a safety perspective. Considering experience from fire testing on LIBs, a correlation between the presence of coolant and sudden ignition before the thermal runaway of the battery is triggered, is evident. In addition, reports from real world battery fires, mentioned in the background section, point at coolant as the trigger for fire, even though the exact chain of events is not known.

A combination of facts from literature studies in several research fields, presented above, and experience from fire incidents, strongly point at a mechanism for rapid gas formation and gas explosion: slow electrolysis result in formation of gas bubbles in the coolant, which lowers the breakdown voltage of the liquid leading to arcing. In the plasma formed in the arc, the temperature reaches thousands of degrees causing evaporation and dissociation of the liquid molecules into hydrogen and other flammable gases. Increased conductivity of the coolant upon contamination from either surface within the system or in the surrounding, may be a first trigger for electrolysis, and thus an indirect cause of the sequence of events leading to the gas explosion observed.

A first step in mitigation of the risks described in this work is to prevent leakage of coolant. This can include choosing appropriate materials for tubing and gaskets, to avoid degradation by EG, and improved procedures for regular monitoring of potential corrosion and leakage. Potentially novel coolant with more advantageous properties can also be developed.

The potential for electrical breakthrough in case of a leakage can be decreased by improved design of the battery system and its high voltage system. However, choosing the best design require a better understanding of the effects of the physical geometry on the electrical environment, including the effect of different battery cell types (pouch, cylindrical, prismatic). Future research should target the steps in the sequence described in Fig. 1, to gain a fundamental understanding of each step. Electrolysis efficiency needs to be studied at different conditions with respect to, for example, distance and temperature. In addition, the effect of bubble formation on electric breakthrough in the relevant geometries need to be understood.

In this short communication, we have illustrated the potential hazards from a theoretical perspective as well as from anecdotal experience. The findings strongly motivate in depth experimental and modeling studies in the area to verify and quantify the identified hazards.