On the Effectiveness of Aerosol Extinguishing Agents for Battery Vent Gases and Hydrogen

The use of new energy carriers, such as hydrogen and lithium-ion batteries (LIB), presents new safety challenges that require updated guidance for the prevention of fire incidents. An important step in addressing these challenges is to investigate the efficiency of established suppression methods when applied to fire events involving these new energy carriers. Although widely used in fire suppression systems, potassium-based aerosols have not been extensively studied by the scientific community. Studies indicate that alkali metal carbonates, such as sodium and potassium carbonates, are effective substitutes for halons in suppression systems [1, 2]. Experimental results indicate that sodium carbonate has a superior mass-based extinguishing efficiency compared to halon 1301 [1]. Likewise, potassium carbonate aerosols are highly effective at suppressing diffusion flames, with an efficiency per unit weight of 3–10 times greater than that of halon 1301 [2]. Because of their proven effectiveness, potassium compounds are the primary constituents of several commercial fire suppression compositions, either with potassium bicarbonate, KHCO₃, or dipotassium carbonate, K₂CO₃. The aerosols, when heated to approx. 900⁰C decomposes into gas phase products, and the fire suppression is achieved as a result of a radical scavenging process consuming OH radicals but potentially also H and O [3]. The sequence of events leading to flame extinguishment with aerosols begins with the distribution of the aerosol within the enclosure, followed by heating and subsequent decomposition of the aerosols, and finally, interactions with radical chemistry.

According to the literature [4], the decomposition of K₂CO₃ takes place with reaction 1, where the product in a moist environment forms KOH through reaction 2. Both K₂O and KOH can then react with a range of radicals and stable components in the flame, to form K-radical which has a flame-retardant effect through efficient consumption of OH radical in the chain termination reaction 3.

Following these reactions, the KOH will further react with radicals OH and H, decreasing the overall pool of reactive radicals sustaining the fire. The inhibition effects are both physical and chemical. The physical effect is primarily achieved through heat absorption [5]. The chemical effect of potassium carbonates is highly efficient in flame inhibition in the gas phase, as the process is auto-catalytic. Mechanistic studies on the homogeneous gas phase process include the chemical kinetics mechanism [3], its effect on strain rate [6], and time-scale analysis [7]. An important step before homogeneous flame inhibition is the thermal decomposition of the particles, as investigated for the analogous Na₂CO₃ by Dounia et al. [8]. The effect of aerosol particle size on KHCO₃ has been studied by Zhou et al. [9], and it was concluded that a decrease in particle size improved fire suppression. The smaller particles decomposed at lower temperatures and absorbed more heat than the larger particles, resulting in larger amounts of reactive radicals available for chemical fire suppression. It is also noteworthy that an experimental study on bicarbonates indicates that potassium carbonates are superior to sodium carbonates [10]. A review by Zhang et al. [11] focuses on various fire extinguishing agents that they categorize as “hot aerosols,” which include potassium-based aerosols. In a review by Rohilla et al. [12] covering common agents from a general fire perspective, potassium is both mentioned as an ingredient in advanced “Aerosol Forming Composition” and as an extinguishing agent on its own belonging to a group named “chemical powders”.

The fire risk related to Li-ion batteries is associated with when the batteries enter a state known as thermal runaway, which involves uncontrolled heat and gas production that can lead to fire or, in instances of delayed ignition, gas explosions. Li-ion batteries contain a range of chemical substances, a large proportion of which are flammable, such as dimethyl carbonate (CH₃OC(O)OCH₃), and smaller amounts are toxic substances such as lithium phosphorus hexafluoride (LiPF₆). During the thermal runaway, uncontrolled chemical reactions occur between different chemical substances in the battery, which results in the formation of new flammable and toxic reaction products as well as a large amount of heat. The breakdown of large stable molecules into smaller and more volatile ones leads to an increase in pressure inside the battery cell, which is further exacerbated by the increased temperature. In thermal runaway, not only are the flammable gases hydrogen (H₂) and carbon monoxide (CO) and the inert carbon dioxide (CO₂) primarily produced, but a significant proportion of methane (CH₄), ethylene (C₂H₄), and other gaseous hydrocarbons is also formed [13]. Among the toxic components, hydrogen fluoride (HF) can be mentioned in particular.

The proportion of different types of gases formed in a lithium-ion battery during thermal runaway depends on several factors, such as the type of cell chemistry and the state of charge (SOC). LIB with a lower SOC emits smaller volumes of gas with a less flammable composition as compared to highly charged batteries [14]. The variation in the composition of the vented gases suggests that the effectiveness of extinguishing agents may also vary, and their effectiveness likely depends on whether the agent is applied to a fire (diffusion flame) or to prevent gas explosions (premixed flame). Generally, there is also a notable variation in the extinguishing efficiency of different agents when applied at the cell level compared to the module and rack levels [15]. Comprehensive review studies on fire-extinguishing agents for suppressing lithium-ion battery fires have been conducted by Majeed et al. [16] and Yuan et al. [17], highlighting the essential design principles for effective fire-extinguishing agents and evaluating various options. However, these reviews do not address potassium carbonates. The present work investigates the effects of applying dipotassium carbonate K₂CO₃ to a premixed flame, simulating a gas explosion due to delayed ignition. This compound is more efficient than potassium bicarbonate, KHCO₃ [3]. The experiments are performed with aerosols with a diameter of around 1 µm since a recent literature review suggests that aerosols with a size in the order of 10 µm are generally ineffective for pure hydrogen flames, as they do not evaporate before reaching the flame unless hydrocarbons are added [18].

Combustion characteristics of gases vented from Li-ion batteries were recently investigated in a simulation study by Nilsson et al. [13]. As part of that study, the compositions of battery vent gases were reviewed, and typical cases were selected for ignition and flame simulations. Batteries with LFP (Lithium iron phosphate) and NCA (Lithium nickel cobalt aluminum oxide) chemistries, studied experimentally by Lammers et al. [19] were found to produce the least and most flammable vent gas mixtures, respectively, which was demonstrated by the determination of laminar burning velocities by Henriksen et al. [20]. The variation in flammability of the vent gases depends on the balance between the inert CO₂ on one hand, and the highly flammable CO and H₂ on the other hand. At 100% SOC, the gases from an LFP battery were found to consist of approximately 48% CO₂, 9% CO, and 30% H₂, while a high-energy NCA battery has less CO₂ and more flammable gases, namely 10% CO₂, 37% CO and 43% H₂. The highly reactive gas mixture from NCA has a burning velocity that is about three times higher than that of the gas mixture from the LFP battery. The high hydrogen content also leads to the flame having a high burning velocity even at low oxygen concentrations (at a high equivalence ratio number). The combustion is driven by radicals that arise in the flame front, mainly O, H, and OH. The hydroxyl radical, OH, is counted in most combustion cases as the most important radical for maintaining flame propagation. However, the study clearly shows that for the highly reactive mixture with a high hydrogen content, the H radical also has great importance when there is a lack of oxygen.

Considering the high amount of H₂ in the most flammable gas mixtures from lithium-ion batteries, and the fact that H₂ itself is an important energy carrier becoming increasingly used, the efficiency of flame extinguishing media on high-H₂ gas mixtures needs to be studied. In the present work, an experimental study on the effect of K₂CO₃ aerosols on the flames for the mixtures between air and a gas mixture mimicking vented gases from an NCA battery¹ as well as pure H₂ is presented. To increase the understanding of the effect of the aerosol-generating fire extinguisher on premixed flames, the experiments were designed to simulate a gas explosion using a well-controlled laminar setup as suggested by [18]. Experiments with methane (CH₄) are also performed as a reference as well as CO₂ as an alternative suppressant. To further understand the chemical effects in the flame-extinguishing process, chemical kinetics simulations were performed for both the NCA-, LFP-, and pure H₂-mixture to elucidate the pure chemical gas phase effects under the assumption that the aerosol is completely evaporated.

The distinct behavior of micro-sized aerosol suppression agents (based on K₂CO₃) toward H₂ and H₂-rich battery vent gas flames, specifically, the increase in laminar burning velocity observed in experiments and the discrepancy with simulation results, can be attributed to the phase of the aerosol agents, the properties of H₂, and the interaction between the aerosol agents and the H₂ flame.

As previously discussed, both thermal and chemical effects can significantly influence flame suppression as aerosol concentration increases. First, these micrometer-sized aerosols absorb heat to initiate evaporation, lowering the flame temperature. Then, they exhibit chemical effects by enhancing recombination reactions and inhibiting chain-branching reactions in the combustion process, leading to successful suppression [12, 30, 31].

For the CH₄ flame, the reduction in the flame cone angle and the expansion of the flame reaction zone area, as observed in the OH-PLIF images in Fig. 4, along with the consistent decrease in laminar burning velocity with increasing aerosol concentration shown in Fig. 7, are attributed to the suppressing effects of aerosols. Figure 10 presents a schematic comparing the interaction between (a) aerosol agents and the CH₄ flame, and (b) aerosol agents and the H₂/battery vent gas flame. The CH₄ flame has a low burning velocity, as indicated by previous simulation results, and exhibits significantly high radiative heat transfer (RHT) efficiency due to the presence of CO₂, H₂O, CO, and possibly soot [32, 33]. The high RHT, combined with extended heat transfer time, promotes the conversion of aerosol agents from solid or liquid to vapor or gas phases in the pre-flame zone, ensuring the effective availability of potassium radicals (K) for gas-phase chemical reactions. The simulation results, assuming complete evaporation of the aerosol, clearly show that K₂CO₂ has strong flame suppression capabilities through reactions involving various potassium compounds and combustion radicals, primarily OH, H, and O. For lean mixtures, even a concentration of 10 g/m³ of K₂CO₃ can reduce laminar burning velocities to less than one-third of their original value. These findings for the CH₄ flame are fully consistent with the simulation results discussed in Sect. 3.2.

Bild vergrößern

In the experimental study, however, applying the same suppressing aerosols to the H₂ flame or high-reactive battery vent gas flame yielded significantly different results. These differences can be attributed to the high hydrogen content in the fuel mixture and the phase conversion process of the aerosols, from solid/liquid to vapor or gas phase when interacting with the flame. Although the H₂ flame burns at a higher temperature than a CH₄ flame [34], it is less efficient at radiative heat transfer due to the absence of radiatively active species (e.g., CO₂, CO) that are present in methane flame. When aerosols are introduced into a H₂-rich flame, local heat loss occurs due to their evaporation. This heat loss is more pronounced in the H₂-rich flame compared to a hydrocarbon flame, likely due to its lower radiative heat transfer (RHT) and higher burning velocities [12, 35]. The faster burning rate of the H₂ flame reduces the time available for heat transfer to the aerosols in the pre-flame zone. In the flame where no aerosol evaporation occurs, the burning rate remains constant. However, in regions where evaporation takes place, the local cooling reduces the burning velocity, causing different sections of the flame to burn at varying speeds. These instabilities, driven by variations in burning velocity, lead to an increase in the flame area. This explanation can be supported by experiments in spray combustion [36]. In this work by Atzler et al. [36], it was shown that the injection of non-evaporating hollow glass spheres in a laminar gaseous propane–air mixture did not alter the flame surface, which was smooth and similar to that without glass spheres. However, introducing the water aerosol in a gaseous propane/air mixture induced cellular surface structures that initiated instabilities, as compared to the smooth flame in a similar experiment without water aerosol [36]. This work indicates the heat loss from the flame due to the evaporation of water droplets has a significant role in the development of instabilities and cellular surface. Bradley et al. [37] also showed that this cellularity results in an increase in the flame surface area followed by a relative increase in laminar flame. Although previous studies [18] have identified the challenges of evaporating K₂CO₃ for hydrogen gas mixtures, no indication of an increase in the burning velocity has been identified. This could be due to the significantly larger particle sizes (~ 5 µm), which might reduce heat loss in the flame and result in insufficient formation of cellular structures to increase flame surface area.

As mentioned earlier, the burning velocity for both the battery vent gas and H₂ flame decreases slightly at lower aerosol concentrations but increases at higher concentrations. This can be explained by the fact that when the aerosol concentration is sufficiently low, the local heat loss due to aerosol evaporation is not significant. As a result, the suppression effect of the aerosols can dominate and thereby reduce the burning velocity. However, at high aerosol concentrations, significant heat loss occurs due to aerosol evaporation, leading to local cooling. This reduces the local burning velocity, creating instabilities through burning velocity gradients along the flame surface, thereby increasing the flame area and laminar burning velocity. The development of instabilities and cellularity due to heat loss by aerosol evaporation in the flame is connected to the Lewis number, \(Le=\frac{\alpha }{D}\), which is defined by the ratio of thermal (\(\alpha\)) to mass diffusivities \((D)\) of the reactant. Clavin et al. [38] and Harper et al. [39] demonstrated that the Markstein number increases linearly with \(Le\) for a wide range of mixtures. The dimensionless Markstein number (\(Ma\)) characterizes the effect of local heat release of a propagating flame on variations in the surface topology along the flame and the associated local flame front curvature [40]. The larger the Markstein length, the greater the effect of curvature on localized burning velocity. Droplets near the flame front will absorb heat as they evaporate; this reduces the heat diffusion from the reaction zone but is unlikely to change the molecular diffusion. Hence the evaporation process of droplets reduces \(Le\) and, so reduces the \(Ma\) which results in a greater propensity to cellularity [41]. For a larger droplet as compared to smaller ones, the reduction in \(Ma\) is expected to be greater. Besides, due to the rich content of H₂ in battery vent gas, the \(Ma\) is also significantly small as compared to as compared to the methane/air flames as H₂ has a small Lewis number \((Le=0.3)\) due to having higher molecular diffusivity as compared to thermal diffusivity. Therefore, the investigated battery vent gas is more prone to develop instabilities and so increase in laminar burning velocity as a consequence of heat loss by the aerosol evaporation [42, 43].

For the pure hydrogen flames, shown in Fig. 6, due to preferential diffusion or differences in thermal and molecular diffusion and negative flame stretch, the flame appears open near its tip [44]. If the thermal diffusivity of the premixture sufficiently exceeds the molecular diffusivity (Le > 1) the flame assumes a continuous luminous conical shape. However, in the opposite case, when molecular diffusivity is high enough (Le < 1), the flame appears open near its tip [45, 46]. The unbalance between fuel and oxidant diffusion is known as preferential diffusion. Kozlovsky et al. [47] numerically observed and identified that the Bunsen flame tip opens at mixture conditions having a Lewis number less than unity (Le < 1). Ern et al. [4] investigated the flame structure of lean and rich hydrogen/air, methane/air Bunsen flames to understand the effect of thermal diffusion of the species on the flame structure.

As noted previously, the effects of CO₂ gas injection on the battery vent gas flame are quite opposite to those of aerosols. With increasing CO₂, the flame cone angle decreases, and the laminar burning velocity drops (see Figs. 5 2a–l and 7). This is because, unlike aerosols, CO₂ does not cause heat loss through evaporation, nor does it create local instabilities through burning velocity gradients along the flame surface. Literature [21, 48, 49] suggests that the laminar burning velocity decreases with increasing CO₂ concentration. For increasing CO₂, the chemical reactivity of the mixture becomes low due to the reduced heat of the flame, which leads to a reduced adiabatic flame temperature and lower concentrations of radicals. High CO₂ enhances the recombination reaction of H + O₂ + M → HO₂ + M, thus reducing the concentration of H radical levels in the flame, affecting the branching reaction H + O₂  ↔ O + OH [50, 51]. Therefore, it can exhibit significant local extinction in the reaction zones, affecting the flame's sustainability.

Experimental studies, complemented by 1D flame simulations, were performed to understand the effect of K₂CO₃ aerosols on the flames for a gas mixture mimicking vented gases from an NCA battery² as well as pure H₂.

The simulation study demonstrates a satisfactory extinguishing effect of gasified aerosols; however, aerosol gasification has been identified as a limiting factor from experimental studies in the effectiveness of the extinguishing agent for premixed gas mixtures of highly reactive battery gases (equivalent to NCA at 100% SOC) and hydrogen gas. For those gases, the lower radiative fraction and the higher burning velocity of the hydrogen-rich mixtures, reduce heat transfer to the aerosol in the pre-flame zone, preventing effective evaporation of the aerosols. When the condensed aerosol reaches the flame, it was found to induce a local cooling resulting in a corrugated flame increasing the flame area and thereby the total burning velocity. This is likely to be further enhanced by the preferential diffusion of hydrogen leading to the growth of the cellular structure which should be further investigated.

Less reactive battery gas mixtures (e.g., those associated with lower SOC or less reactive chemistries) were not tested in the current study. However, the results from simulation and experimental studies on CH₄ flames indicate that aerosols are more effective for such mixtures. The lower flame burning velocity, combined with higher radiative heat transfer (RHT), facilitates the essential evaporation of the aerosol agent and effectively suppresses the CH₄ flame. Similarly, the aerosol suppressant may also be more effective for less reactive battery gas mixtures, but experimental validation is necessary.

In summary, this indicates that, although the aerosol agent has been found highly effective for diffusion flames, a beneficial effect cannot be guaranteed for situations, where a hydrogen-rich gas is allowed to mix with air prior to ignition such as in the case of a gas explosion in gasses emitted from an NCA-battery at 100% SOC or pure hydrogen. To improve the ability of the aerosol to inert gas mixtures with high concentrations of hydrogen, modification of the aerosol such as reduced particle size and higher initial temperature could be considered.