Six commercial LIB cathode chemistries account for the majority of the present-day market. Their thermal-runaway onset temperatures and specific energy densities are summarized in Figure 1 and Table 1. The data reflect a fundamental trade-off: cathode chemistries that maximize specific energy (NCA at 280 Wh/kg, NMC at 220 Wh/kg) operate close to the limit of cobalt-oxygen bond stability and undergo cathode oxygen release at 150-210°C, while olivine-structured LFP retains its phosphorus-oxygen framework up to 500-800°C [3,16]. The LFP onset of thermal runaway is correspondingly the highest of the six chemistries (270°C); LFP fires also burn cooler (peak cell-face temperature ≈620°C) than NMC fires (≈800°C) and eject only 20-25% of the cell mass against 40-50% for NMC [4,16].

Thermal runaway in LIBs is a three-stage cascade of exothermic decomposition reactions that release the chemical and electrochemical energy stored in the cell as heat (Figure 2). In stage 1, an abuse condition like internal short circuit, mechanical penetration, overcharging, or external heating raises the cell temperature above the onset value of the solid-electrolyte-interphase (SEI) film, typically 80-120°C. The SEI decomposes and exposes the lithiated graphite anode (Li_xC₆) to the electrolyte, releasing additional heat. In stage 2, between roughly 130°C and 200°C, the polyolefin separator (polyethylene, polypropylene) melts and ruptures; the cathode and anode short internally, and the electrolyte solvents (ethylene carbonate, dimethyl carbonate, ethyl methyl carbonate) begin to volatilize and react. In stage 3, above the cathode decomposition temperature, the metal-oxide cathode releases lattice oxygen, which combines with the volatilized electrolyte and any released hydrogen and methane in a violent combustion event that may take the form of an open flame, a high-velocity jet, or, in confined cylindrical formats, a missile-effect explosion [5,17,18].

Three commercial cell formats dominate automotive and stationary applications: cylindrical hard-case cells (typically 18650 or 21700 form factor), prismatic hard-case cells, and pouch cells with a flexible aluminum-laminate envelope (Figure 3). The format strongly influences the thermal-runaway propagation mechanism and the suitability of different suppression strategies. Cylindrical cells distribute internal pressure evenly across their circumference and are equipped with safety vents that release accumulated gas; thermal runaway in cylindrical cells is therefore typically characterized by jet venting and, in confined packs, by missile-like ejection of the cell can. Prismatic cells, particularly LTO formulations, are mechanically constrained and tend to extinguish more quickly than NMC pouch cells; however, they may undergo violent rupture if their casing is over pressurized. Pouch cells, with their high packaging efficiency (90-95% cell volume utilization) but low mechanical rigidity, propagate thermal runaway most rapidly due to their planar geometry and require active thermal management even in nominal operation [19,20].

Within each cell, four functional components (cathode, anode, electrolyte and separator) operate in concert (Figure 4). The cathode and anode set the cell capacity and voltage, while the electrolyte and separator set the safety margin. Damage to the separator is the most common direct precursor of thermal runaway: a 100 µm-scale defect in the polyolefin separator allows direct cathode-anode contact, generating a localized internal short circuit with current densities exceeding 100 A/cm² and sustained heat-release rates above 100 W/cell [21].

Three principles underpin all fire-suppression chemistry:

Practical extinguishing agents combine these three principles to differing degrees, and a brief comparison frames why a solid salt hydrate is of interest for LIB fires. Water and water mist act chiefly by cooling and by latent-heat absorption during evaporation, but they conduct electricity and react with reactive metals (detailed below); carbon dioxide and inert gases act by oxygen displacement, giving clean, residue-free knock-down but negligible cooling and no re-ignition barrier - a decisive drawback for cells that release their own lattice oxygen; foams form a smothering, oxygen-excluding blanket with some cooling, but drain and collapse on hot, near-vertical cell surfaces; and conventional dry chemical powders (mono-ammonium phosphate, sodium bicarbonate) interrupt the radical chain reaction and smother the flame, yet deliver little sustained cooling and, for mono-ammonium phosphate, decompose into corrosive, toxic phosphorus and nitrogen oxides. Against this background a hydrated-salt solid powder is attractive because it couples the electrical safety and smothering action of a dry powder with a cooling capacity - supplied by endothermic dehydration - that approaches that of water (Section 1.5).

Conventional water-based extinguishers act predominantly through cooling: a liter of water absorbs approximately 313 kJ in heating from 25°C to 100°C, plus a further 2,300 kJ on completely vaporizing at 100°C, yielding a total cooling capacity of ≈2.6 MJ/L [22]. However, water suffers from three intrinsic limitations when applied to LIB fires:

and the hydrogen so released combusts with atmospheric oxygen to regenerate water vapor, but only after passing through an explosive intermediate state:

The water-based cooling capacity Q_cool for an applied mass mw of water at initial temperature T₀ heated to a final temperature T_f (with full vaporization) is given by the standard sensible-plus-latent expression:

where ΔT = T_f - T₀, C_p is the specific heat capacity (4.18 kJ/kg·K for liquid water), and L_v is the latent heat of vaporization (2,260 kJ/kg at 100°C). Equation (3) sets the upper bound on the cooling that water can provide: even when fully consumed, 1 L of water cannot extract more than ≈2.6 MJ from a burning cell. A 60 kWh NMC pack contains approximately 216 MJ of stored chemical energy, of which an estimated 30-40% may be released during full thermal runaway. Even allowing for radiative and convective losses, the bare cooling capacity of water is therefore only marginally sufficient to absorb the heat of a single major thermal-runaway event in a typical EV pack [9,10].

Hydrated (crystal-water-bearing) inorganic salts have emerged as a distinct class of LIB cooling and fire-suppression agents precisely because they combine a large endothermic capacity with electrical non-conductivity. Li et al. [23] systematically screened seven salt hydrates with different crystal-water contents as dry-powder extinguishants and showed, by differential scanning calorimetry, that the endothermic enthalpy released below 150°C can exceed 800 kJ/kg - sufficient to interrupt thermal-runaway propagation - while electric breakdown voltages of 4-60 kV confirm that, unlike liquid water, these powders do not provide a conduction path between live pack components. Borax pentahydrate (Na₂B₄O₇·5H₂O), the active component of the present agent, belongs to exactly this family: it carries five molecules of crystal water that are released in a stepwise endothermic dehydration cascade (Section 2.2).

Several recent studies have exploited salt hydrates for battery protection in adjacent configurations. Lin et al. [24] integrated latent-heat and thermochemical storage in an inorganic salt hydrate that both delayed thermal runaway and inhibited cell-to-cell propagation, and composite phase-change materials built on Na₂SO₄·10H₂O and the alum-type hydrate KAl(SO₄)₂·12H₂O have reported latent heats near 227 J/g and thermal conductivities of ≈1.4 W/(m·K) when blended with expanded graphite [25]. These works, however, deploy the salt hydrate as an inter-cell barrier plate, an encapsulated phase-change composite, or an external spray, and most are evaluated on single cells or two-cell propagation pairs. The specific configuration examined here - a cylindrical 18650 pack fully immersed in loose boron-salt-hydrate powder so that every cell face is in direct conductive contact with the agent - has not, to the authors’ knowledge, been reported. It is this in-pack immersion geometry, rather than the salt-hydrate chemistry itself, that constitutes the novel contribution of the present study. It is acknowledged that the cited works provide the material-level characterization (DSC/TG enthalpies, conductivity) that the present proof-of-concept does not, a point returned to in Sections 2.4 and 4.4.

Six cell formats from four chemistries (NMC, LTO, NCA, LFP) were tested. The principal characteristics are summarized in Table 2. Pouch cells (types 1, 2, 3) were of NMC chemistry, prismatic cells (type 4) were of LTO chemistry, and cylindrical cells were of NCA (type 5) and LFP (type 6) chemistries. All cells were tested in a fully charged state (state of charge 100%).

To ensure reproducibility, the following details apply to all tests. Every cell and pack was tested at 100% state of charge (SoC). Thermal runaway was forced by external heating only: a 400-kW propane gas burner positioned beneath the specimen, applied continuously until visible runaway initiated (typically 30-90 s after ignition); no nail penetration, overcharge, or mechanical abuse was used, so the heat source in Tests T1-T8 is external rather than internally generated. Two suppression tools were used: a manual low-pressure water lance for single cells, and a Cobra-type high-pressure water-with-additive lance, capable of perforating thin metallic enclosures, for packs. Cell-surface and flame temperatures were recorded by the host laboratory using K-type thermocouples and infrared instrumentation. The boron-cooling experiment (Test T9), described in Section 2.3, used a separate setup at Çukurova University and a single continuously logged in-pack K-type thermocouple. The two campaigns were conducted on different platforms and are therefore compared qualitatively rather than as a controlled head-to-head trial; this limitation is stated explicitly in Section 2.4 and revisited in Section 4.4.

Ignition was provided by a 400-kW propane gas burner placed beneath the cell or pack. Cells were heated until thermal runaway became visible (typically 30-90 s after ignition). For the eight water-suppression tests (Tests T1-T8), a manual extinguisher (low-pressure water lance) and, for cylindrical-cell packs, a Cobra-type extinguisher (high-pressure water-with-additive lance capable of penetrating thin metallic enclosures) were used. The full test matrix is given in Table 3.

The boron-based extinguishing chemical used in the present work is a B-C-class fire-extinguishing powder developed by Demircan Fire (commercial name Fireborex) [31]. Its principal active component is borax pentahydrate (Na₂B₄O₇·5H₂O), with the composition reported in Table 4. Borax pentahydrate is obtained from the Eti Maden tincal mine (Türkiye), which holds 73% of the world’s reserves of refined boron compounds [15,32]. The substance is non-carcinogenic, has a longer shelf life than ammonium-phosphate-based dry chemical extinguishers, and forms a glassy film at temperatures above ≈600°C that suppresses re-ignition.

On heating, borax pentahydrate undergoes a cascade of three endothermic dehydration steps:

followed by the further loss of water to borax monohydrates at 290-299°C and to anhydrous borax at 400-450°C [31,33]. Each step is endothermic, with an enthalpy of dehydration approximately equal to the latent heat of vaporization of the released water (≈42 kJ/mol H₂O at the relevant temperature). The total endothermic heat absorbed by a mass m_B of boron pentahydrate is:

where n_j is the number of moles of water released in step j and ΔH_v,j is the corresponding latent heat. With three dehydration steps releasing a total of 5 mol H₂O per mol Na₂B₄O₇·5H₂O, the total endothermic capacity is approximately 210 kJ/mol of borax pentahydrate, distributed across the three thresholds shown in Figure 5.

Fifteen 18650-format cylindrical NMC cells (nominal capacity 2,000 mAh each, total stored energy ≈110 Wh) were connected in series by a thin steel wire and placed inside a steel box of internal volume 700 cm³. The box was then filled with boron-based solid cooling powder until the cells were entirely covered, and the assembly was placed on top of a 1.4 kW resistive heating furnace (Figure 6). A K-type thermocouple was inserted into the center of the box, in direct contact with one of the cells. The full setup was operated continuously for 23 minutes; the heating furnace was switched on at t = 0, and the temperature was logged at 1 s intervals.

Thermophysical properties and barrier-formation mechanism. The cooling action of the powder rests on three thermophysical mechanisms, whose governing parameters are summarized here together with an explicit statement of which were measured and which were taken from the literature. (i) Endothermic dehydration: each of the three dehydration steps of borax pentahydrate releases crystal water with an enthalpy close to the latent heat of vaporization of water (≈42 kJ/mol H₂O), giving ≈210 kJ/mol of borax pentahydrate, i.e. ≈0.55 MJ/kg from dehydration alone; the step temperatures (160-170°C, 290-299°C, 400-450°C) and enthalpies are taken from the thermal-analysis study of Akbay and Altıokka [33], not measured here. For comparison, salt hydrates with higher crystal-water content can reach DSC endothermic enthalpies above 800 kJ/kg below 150°C [23], indicating clear room to optimize the agent. (ii) Sensible heat: with a specific heat capacity of the powder of order 1.0-1.4 kJ/(kg·K) for sodium-borate solids and a bulk density of ≈1.15 g/cm³, the ≈800 g fill provides additional sensible storage; these values are nominal, drawn from material handbooks rather than measured on the present batch. (iii) Insulation: the granular packing has a low effective thermal conductivity (estimated 0.2-0.4 W/(m·K) for a porous sodium-borate powder bed), which limits radiative and conductive heat transfer from the heater to the cells.

Barrier formation. Above ≈600°C, the dehydrated residue (sodium borate/boron oxide) softens and forms a glassy, oxygen-excluding film over exposed surfaces, which is the mechanism proposed to suppress re-ignition. In the present work, this barrier is inferred from the observed absence of re-ignition and from the known glass-forming behavior of borate melts; it was not independently verified. We did not perform SEM, XRD, or TG-DSC on the post-test residue, and we did not measure thermal conductivity, specific heat, or phase-change kinetics on this specific powder. These measurements are necessary to confirm the barrier mechanism and to quantify the kinetics, and they are identified as priority future work in Section 4.4. The values quoted above should therefore be read as literature- and handbook-based estimates that bound the expected behavior, not as characterized properties of the tested batch.

This study did not involve human participants, human-derived tissue, or animal subjects, and therefore no institutional ethical-review-board approval or informed-consent procedure was required. All experimental activities were conducted under controlled fire-laboratory conditions at the University of Poitiers/ISAE-ENSMA Pprime Institute (Poitiers, France) and at the Çukurova University Fire-Test Laboratory (Adana, Türkiye), in full compliance with the host institutions’ standard operating procedures for high-voltage lithium-ion battery fire testing. Tests were performed in dedicated outdoor or naturally ventilated areas with the on-site presence and operational support of the Departmental Fire and Rescue Service of the Vienne Region (France) and the Departmental Fire and Rescue Service of the Adana Metropolitan Municipality (Türkiye). All test personnel wore certified personal protective equipment, including self-contained breathing apparatus where toxic gas release (HF, CO, hydrocarbons) was anticipated, and minimum exclusion radii were enforced to mitigate the risk of cell ejection (missile effect) and thermal-runaway propagation. Test residues, including burnt cell debris and electrolyte residues, were collected and disposed of as hazardous waste in accordance with EU Directive 2008/98/EC and the corresponding Turkish hazardous-waste regulations. The authors declare that the experimental data, figures, and temperature recordings reported in this paper are original, accurate, and have not been manipulated. The work has not been submitted, in whole or in part, for publication elsewhere.

Water-suppression tests were divided into single-cell heating tests (Tests T1-T3), pack-level Cobra-extinguisher tests on prismatic and pouch packs (Tests T4-T6), and Cobra-extinguisher tests on real cylindrical EV-scooter and leisure-vehicle batteries (Tests T7-T8). The principal experimental observations are summarized below.

Single-cell NMC pouch fires (Tests T1-T2, Figure 7) progressed rapidly. Once the surface temperature reached the manufacturer-quoted thermal-runaway onset of 200°C, the aluminum-laminate envelope ruptured, and the cell entered open-flame combustion within approximately 15 s. Flame temperatures of 600-800°C were observed; the entire cell was consumed within 90 s. By contrast, the LTO prismatic cell (Test T3, Figure 8) burned much more slowly: the cell case did not rupture during the test, and combustion was confined to electrolyte vapor leaking through the safety vent. Manual water extinguishing was effective on the LTO cell within 30 s, although a residual smolder persisted for several additional minutes.

Pack-level water-suppression tests on pouch packs (Tests T4-T5, Figure 9) demonstrated the importance of the casing material. The aluminum-and-plastic pack of Test T4 melted progressively, eventually allowing water from the Cobra extinguisher to reach the cells; the steel-cased pack of Test T5 required mechanical deformation or melting of fusible parts before water could be injected. The pouch-cell pack of Test T6 underwent a violent rupture as the molten aluminum of the cell case explosively combined with the released electrolyte. After approximately 9 minutes of total exposure, all three packs were eventually extinguished, although the residual cell debris continued to off-gas for several hours.

The most challenging tests were the cylindrical-cell scooter and leisure-vehicle batteries (Tests T7-T8, Figure 10, Figure 11). Cylindrical cells, equipped with safety vents, repeatedly ejected gas and partial cell contents under the missile effect characteristic of pressurized-can rupture. The 2 kWh NCA scooter battery (Test T7) underwent the first thermal runaway at approximately 3 minutes of heating, was apparently extinguished, then re-ignited 4 minutes later in a second runaway accompanied by abundant white smoke and explosive cell ejections. The 20 kg LFP leisure-vehicle battery (Test T8) entered a sustained 9-minute combustion punctuated by repeated explosions, with cells launched several meters from the test bench (missile effect). The visual records of the cylindrical-cell tests are shown in Figure 10 and Figure 11.

The boron-pentahydrate-cooled 18650 pack (Test T9) was operated for 23 minutes with continuous 1.4 kW heat input. The temperature evolution measured at the center of the steel box, in direct contact with the cell pack, is plotted in Figure 12.

Three observations are noteworthy. First, the cell-pack temperature plateaus at 65-75°C throughout the heating exposure, well below not only the NMC thermal-runaway threshold of 200°C, but even the boiling point of water (100°C). Second, the only significant temperature excursion is the transient peak at t ≈ 16 min, which corresponds to a deliberate withdrawal of the thermocouple from the cell pack and direct exposure to the heating element underneath the box; once the thermocouple is reinserted into its in-pack position, the temperature returns to the equilibrium plateau within approximately 60 s. Third, no fire, no flame, no gas venting, and no audible cell rupture occurred during the entire 23-minute test. Post-test inspection confirmed that all 15 cells retained their original geometry, that the safety vents had not opened, and that no electrolyte leakage was visible on the exterior of the cells.

The mechanism underlying this performance is a combination of the three effects identified in Section 2.2:

where h_phase is the additional effective heat-transfer coefficient associated with the dehydration phase change. With h_phase ≈ 200 W/(m²·K) [characteristic of phase-changing porous media], the equilibrium pack temperature predicted by equation (14) for the imposed heat flux of approximately 9 kW/m² is 70°C, in close agreement with the observed plateau.

Table 5 consolidates the present tests for cross-comparison, while Figure 13 and Table 6 compare the peak temperature observed during the boron-cooling test with the values reported by Mrozik et al. [12] for water-mist, encapsulator-water-mist, ammonium-bicarbonate carbonate-agent and mixed-agent (boron-and-surfactant) methods on a small 8 kWh BESS, and with the no-suppression reference; The peak temperature reached during the present boron-cooling test (102.5°C, itself a transient artifact of thermocouple withdrawal, with the plateau at ≤75°C) is roughly an order of magnitude below the 780-890°C peaks of the water-based methods. We deliberately avoid quoting a precise “factor of seven/nine” here: as detailed in Sections 3.3 and 4.4, the present test differs from the BESS data in scale, energy content and heat-input mode, so this contrast is indicative of the qualitative difference in behavior rather than a controlled effect size. It is important to acknowledge that the test conditions differ in three respects:

The present results, set against the recent (2024-2025) BESS-suppression literature, suggest that solid boron-based cooling agents occupy a distinct and complementary niche relative to water-based, encapsulator-water-mist, and microcapsule-based strategies. Mrozik et al. [12] characterized five suppression methods on an 8 kWh domestic BESS and found that water mist and encapsulator-water-mist methods extended the time to thermal-runaway propagation by 179% and 167%, respectively, but that none of the tested methods prevented eventual destruction of the system, with peak temperatures of 780-890°C reached within 5-15 minutes of fire onset. Bai et al. [13], working with thermosensitive microcapsules of in-situ-polymerized polyurea encapsulating a ternary core (water, ammonium polyphosphate, boron-based agent), reduced the peak thermal-runaway temperature of a single NMC cell by 14.0 K and the heating rate by 0.17 K/s, but were unable to suppress the runaway event itself fully.

By contrast, the present boron-cooled 18650 pack remained below the thermal-runaway threshold for the entire 23-minute test. The mechanistic explanation lies in the granular, in-pack distribution of the suppressant: each cell is in direct conductive contact with the boron pentahydrate matrix on all sides, and the high effective heat capacity of the matrix (combined sensible heat, latent heat of dehydration, and trapped water vapor) absorbs the heat input before it can raise the cell temperature to the SEI-decomposition threshold of approximately 130°C. This intimate-contact configuration is qualitatively different from the external spray-application methods reported in [12,34]. It is more analogous to the immersion-cooling concepts investigated for high-power EV battery thermal management [35].

Figure 14 shows the qualitative heat-release-rate profile predicted by the Arrhenius model of Section 1.6 for NMC, LFP, and LTO cells subjected to a constant external heat input. The NMC cell reaches a peak heat-release rate of approximately 9 kW within 12 minutes of heating onset, while the LFP and LTO cells release 4 kW (at 26 min) and 2 kW (at 22 min), respectively. The total endothermic capacity of the 700 cm³ boron-pentahydrate-filled box used in the present experiment, computed from equation (13) with mB ≈ 800 g (powder density ≈ 1.15 g/cm³) and the 210 kJ/mol total dehydration enthalpy, is approximately 580 kJ. This is sufficient to absorb the heat-release rate of a single NMC cell during thermal runaway for approximately 60 s. To extend protection for the full duration of a multi-cell propagation event would require either a larger total mass of boron powder (≈5 kg per cell for 5 minutes of protection) or a hybrid configuration combining boron powder with active water mist.