Characterization of Lithium-Ion Battery Fire Emissions—Part 3: Gas Emissions

Abstract

Lithium-ion batteries (LIBs) release significant amounts of toxic, corrosive, and flammable gases when they enter thermal runaway (TR). These emissions can be hazardous to human health, damage nearby equipment, pose fire and explosion risks, and degrade air quality. This study measured concentrations for a range of hazardous gases released from TR-driven combustion of cylindrical lithium iron phosphate (LFP) and pouch-style lithium cobalt oxide (LCO) LIB cells. Gas emissions were measured by dedicated analyzers and Fourier transform infrared spectroscopic (FTIR) analysis, and emission factors were calculated. Dangerous concentrations of hydrogen fluoride (HF) were observed, reaching up to 50 ppm from the combustion of single LIB cells. Large amounts of combustible electrolyte solvents and light hydrocarbons were released in some cases, depending on cell combustion behavior. Electrolyte solvents, hydrogen chloride (HCl), and particles were released earlier than other species and should be targeted for early TR detection. Gas emissions were correlated with cell state of charge (SOC) and combustion behavior. Cells at high SOCs had higher peak concentrations of HF, HCl, CO, and flammable hydrocarbons, and these peaks happened sooner after cell failure than for low-SOC tests.

1. Introduction

Lithium-ion batteries (LIBs) can combust rapidly and generate significant gaseous emissions when they experience thermal runaway (TR), a process in which the cell’s energy is released through uncontrolled combustion [1,2]. Emitted gases can be toxic, corrosive, flammable, and pose significant risks to human health and nearby property [3,4,5]. Many studies have reported gaseous emissions of failing LIB cells, but cell chemistry, energy capacity, failure trigger, gas detectors, and observed results vary significantly [1,6]. The dependence of gas emissions on state of charge (SOC) and combustion behavior is also underrepresented in the literature [6].

Hazardous gas emissions from failing LIB cells are of specific concern due to their immediate and acute risks to nearby people, including first responders [5]. These emissions can also be transported over distances, potentially causing further harm. The most concerning species are acidic gases such as hydrogen fluoride (HF), hydrogen chloride (HCl), and hydrogen cyanide (HCN). Other toxic air pollutants include carbon monoxide (CO), nitrogen dioxide (NO₂), and sulfur dioxide (SO₂). Flammable gases such as light hydrocarbons, evaporated electrolyte solvents, and hydrogen gas (H₂) are also released [1]. Of great concern are HF emissions by the dissociation and hydrolyzation of the electrolyte lithium hexafluorophosphate (LiPF₆) salt through the following sequences [6]:

$$\mathrm{L}\mathrm{i}\mathrm{P}{\mathrm{F}}_6\to \mathrm{L}\mathrm{i}\mathrm{F}+\mathrm{P}{\mathrm{F}}_5$$

(1)

$$\mathrm{P}{\mathrm{F}}_5+{\mathrm{H}}_2\mathrm{O}\to \mathrm{P}{\mathrm{O}\mathrm{F}}_3+2\mathrm{H}\mathrm{F}$$

(2)

$$\mathrm{L}\mathrm{i}\mathrm{P}{\mathrm{F}}_6+{\mathrm{H}}_2\mathrm{O}\to \mathrm{L}\mathrm{i}\mathrm{F}+\mathrm{P}{\mathrm{O}\mathrm{F}}_3+2\mathrm{H}\mathrm{F}$$

(3)

Bugryniec et al. synthesizes gas emission composition, total volume, and toxicity for different LIB types and sizes from over 60 studies [1]. Rappsilber et al. focuses on heat release and total gas emission amounts from 37 studies [6]. Total emission amounts depend on the number and size of LIBs burned, which necessitates the normalization of emissions to the cell’s nominal energy capacity [5]. Gas emission factors (EFs) are therefore reported in mg/Wh to allow for comparison between experiments.

Complementing the particle compositions and size distributions presented by Claassen et al. [7,8], the objectives of this paper are to (1) evaluate gas concentration evolution during cell combustion, including peak concentrations of toxic and flammable compounds; (2) quantify gas EFs; and (3) evaluate the dependence of peak gas concentrations and EFs on cell chemistry, combustion behavior, and SOC. This study improves upon existing studies by examining detailed emission characteristics, correlation with combustion behavior and cell SOC, and how aggregated emissions compare to peak concentrations.

2. Materials and Methods

Experimental methods and data analysis have been described in the companion papers [7,8], with a brief overview given here. Two LIB types were tested: a cylindrical 18650-style lithium iron phosphate (LFP) cell and a pouch-style lithium cobalt oxide (LCO) cell [9,10]. The nominal energy capacities for the two cell types were 3.63 Wh and 9.25 Wh, respectively. Combustion tests of the two cell types were conducted in an ~8 m³ burn chamber [11] with an exhaust flow rate (Q) of ~3300 L/min, corresponding to ~25 air changes per hour (ACH). The air supplied to the chamber consisted of conditioned room air at a typical temperature of 26 °C and a relative humidity (RH) of 20%. The chamber and instrument layout is shown in Figure 1. Each LIB cell was charged to the desired SOC (5 levels between 0 and 100%) by a programable charger before being placed in a ceramic crucible and heated to cell failure and TR by an electric hot plate. A type K thermocouple was placed adjacent to the vent port for LFP cells, and on the cell face for LCO cells, to measure emitted flame temperatures. Each cell type and SOC combination was tested at least three times. A total of 16 and 18 tests were performed for LFP and LCO cells, respectively.

LIB combustion emissions were sampled from the chamber exhaust duct and directed to a suite of gas analyzers. Airborne particulates were filtered from the gas sampling stream by placing Teflon filters upstream of all instrument inlets. The sampling stream was not conditioned or diluted in any other way. The sampling line inlet was positioned roughly 2.5 m above the tested LIBs, and far enough from the stack bottom to ensure uniform flow. The distance from the sampling inlet to the instruments was 2.5–4 m. All instruments were set to a 1 Hz sampling frequency, although not all instruments used are able to make measurements at this rate. CO₂ was measured by a nondispersive infrared analyzer (Model 840A; LI-COR Environmental, Lincoln, NE, USA), while CO, SO₂, and NO_x were measured by U.S. Environmental Protection Agency (EPA) Federal Equivalent Method (FEM) analyzers (Model 48iQ, 43i, and 42iQ, respectively; Thermo Fisher Scientific, Waltham, MA, USA). The NO_x instrument measured NO_x and NO directly and calculated NO₂ from the difference between these two measurements. Number concentrations of emitted 6 nm–10 µm particles were measured by an electrical low-pressure impactor (ELPI+; Dekati Ltd., Kangasala, Finland) [8].

A portable Fourier transform infrared (FTIR) spectroscopic gas analyzer (Model GT5000 Terra; Gasmet, Vantaa, Finland) was used to measure toxic and combustible gases. This analyzer was placed closer to the tested LIBs to limit emissions condensation prior to measurement. The distance from the test LIB to the FTIR measurement cell was roughly 2 m. Combustible gases were grouped into two distinct categories, electrolyte solvents and total hydrocarbons (THC). These groupings are useful as whole-spectrum FTIR analysis poses challenges in differentiating the IR absorption patterns of similar gases, making the exact composition and relative concentrations less certain than their aggregated total.

The hydrocarbons most closely matching the observed IR absorption were methane (CH₄), propane (C₃H₈), hexane (C₆H₁₄), and styrene (C₈H₈). All electrolyte solvents with available reference spectra were included in the analysis, consisting of ethylene carbonate (EC), diethyl carbonate (DEC), dimethyl carbonate (DMC), and ethyl methyl carbonate (EMC). These four species may not be present in both cell chemistries; however, due to similar absorption spectra, a much lower spectral residual was obtained by including them. Likely, DEC was absent from LFP cell tests and DMC was absent from LCO cell tests due to the species being detected only occasionally. Detailed cell chemical composition information is proprietary and not available.

3. Results

3.1. Combustion Behavior

Modified combustion efficiency (MCE) can be used to describe the combustion intensity and phase (i.e., smoldering or flaming) throughout each test:

$$\mathrm{M}\mathrm{C}\mathrm{E}={\displaystyle \frac{\Delta {\mathrm{C}\mathrm{O}}_2}{\Delta {\mathrm{C}\mathrm{O}}_2+\Delta \mathrm{C}\mathrm{O}}}$$

(4)

where ΔCO_x represents the concentration above background levels [12]. ΔCO, ΔCO₂, and MCE time series, as well as the portions of each test with visually observed flaming combustion, are shown in Figure 2 for each LIB type and SOC.

Combustion and resulting emissions in low-SOC (0 and 30%) tests were less vigorous and more prolonged than for higher-SOC (50–100%) tests. LFP tests with 0% and 30% SOCs resulted in CO and CO₂ concentrations barely above background due to a lack of flaming combustion. On the other hand, LCO tests at the same SOCs had visible flaming and generated higher emission peaks.

As expected, flaming combustion corresponded to higher MCEs, generally exceeding 0.9 during the primary emissions phase of each test. However, flames were observed for shorter periods than elevated MCEs were. This indicates that either the large amount of emissions released during flaming combustion took time to be vented from the chamber, or that combustion efficiency was still high within the cell despite no visible flaming. MCE values began to drop as CO and CO₂ concentrations returned to background levels, most notably in LFP tests, indicating prolonged smoldering combustion.

MCE was averaged for each test during the period where CO₂ values were >1% of their maximum. As shown in Figure 3, the average MCE for LFP tests remained under 0.94 for SOCs below 50%, then increased to above 0.96 at higher SOCs. In contrast, MCE values for LCO tests exceeded 0.95 across all SOCs. These results align with visual observations that LFP tests did not flame for SOCs below 50%, whereas LCO tests flamed at all SOCs. This difference is likely due to the higher thermal stability of LFP chemistry and the sealed cylindrical casing, which limits atmospheric O₂ ingress.

3.2. Toxic Gas Emissions

Table 1. Exposure limits for selected air pollutants.

Limit a	CO b (ppm)	CO2 c (ppm)	SO2 (ppm)	NO (ppm)	NO2 d (ppm)	HF (ppm)	HCl (ppm)	HCN (ppm)
REL	35	5000	2	2	1 STEL	3	5 C	4.7 STEL
IDLH	1200	40,000	100	100	13	30	50	50
PEL	50	5000	5	2	5 C	3	5 C	10
AEGL-1	NR	NA	0.2	NR	0.5	1	1.8	2
AEGL-2	83	NA	0.75	NR	12	24	11	7.1
AEGL-3	330	NA	30	NR	20	44	26	15
ERPG-1	200	NA	0.3	NA	1	2	3	NA
ERPG-2	350	NA	3	NA	15	20	20	10
ERPG-3	500	NA	25	NA	30	50	150	25

Note. ^a REL by NIOSH: Recommended exposure limit, typically for time-weighted average (TWA) concentration for up to a 10 h workday during a 40 h workweek. IDLH (30 min exposure) by NIOSH: Immediately dangerous to life or health concentration. PEL (8 h TWA) by OSHA: Permissible exposure limit. AEGL-1 (1 h exposure) by U.S. EPA: Acute Exposure Guideline Level for notable discomfort, irritation, or certain asymptomatic non-sensory effects. However, the effects are not disabling and are transient and reversible upon cessation of exposure. AEGL-2 (1 h exposure) by U.S. EPA: Acute Exposure Guideline Level for irreversible or other serious, long-lasting adverse health effects or an impaired ability to escape. AEGL-3 (1 h exposure) by U.S. EPA: Acute Exposure Guideline Level for life-threatening health effects or death. ERPG-1 (1 h exposure) by AIHA: Emergency Response Planning Guideline for the maximum airborne concentration below which nearly all individuals could be exposed without experiencing more than mild, transient adverse health effects or without perceiving a clearly defined objectionable odor. ERPG-2 (1 h exposure) by AIHA: Emergency Response Planning Guideline for the maximum airborne concentration below which nearly all individuals could be exposed without experiencing or developing irreversible or other serious health effects or symptoms that could impair an individual’s ability to take protective action. ERPG-3 (1 h exposure) by AIHA: Emergency Response Planning Guideline for the maximum airborne concentration below which nearly all individuals could be exposed without experiencing or developing life-threatening health effects. ^b NR: Not recommended due to insufficient data. ^c NA: not available. ^{d STEL}: A short-term exposure limit, typically a 15 min TWA exposure that should not be exceeded at any time during a workday. ^C: A ceiling REL that should not be exceeded at any time.

lists recommended exposure limits (RELs) and Immediate Danger to Life and Health (IDLH) concentrations published by the U.S. National Institute for Occupational Safety and Health (NIOSH) [13], the permissible exposure limits (PELs) by the Occupational Safety and Health Administration (OSHA) [14], the three tiers of Acute Exposure Guideline Levels (AEGLs) by the U.S. EPA [15], and the three tiers of exposure limits from the Emergency Response Planning Guidelines (ERPGs) by the American Industrial Hygiene Association (AIHA) [16] for selected toxic gases emitted from LIB combustion.

Both CO₂ and CO levels stayed below their respective RELs of 5000 and 35 ppm, respectively, for all tests, with only one test showing CO concentrations reaching 30 ppm. Similarly, as shown in Figure 4, SO₂ remained below its REL of 2 ppm, while neither NO nor NO₂ exceeded their RELs. However, the rapid rise in concentrations, especially for LCO tests, where significant emissions occurred for all SOCs, from the combustion of individual, small LIB cells indicates that larger battery packs could quickly emit hazardous levels of these species, especially in enclosed spaces with limited ventilation.

Significant emissions of acidic gases are a major concern in LIB fire safety. Concentration time series for HF and HCl are shown in Figure 5, while HCN concentrations remained below detection limits for all tests. Peak HF concentrations exceeded the REL and PEL (3 ppm) in all tests. The 1 h ERPG-2 (20 ppm), 1 h AEGL-2 (24 ppm), and 30 min IDLH limits (30 ppm) were exceeded in 57%, 45%, and 36% of tests, respectively.

All three 75% SOC LFP tests reached the AEGL-3 of 44 ppm, at which life-threatening health effects are expected. The fact that every 75% SOC LFP test had higher peak HF concentrations than any other test condition is critical, indicating significant differences in HF exposure can result from relatively small SOC differences. HF emission peaks during LCO tests were generally narrower than those in LFP tests, likely indicating more intense combustions, and had the highest HF concentrations occurring at 60% SOC. All but two LFP and LCO tests had HCl concentration peaks below the REL and PEL of 5 ppm, and 30% of tests reached the ERPG-1 of 3 ppm. As previously discussed, these concentrations resulted from the combustion of individual, small LIB cells in a well-ventilated chamber. Combustion involving multiple and/or larger LIB cells in an enclosed area would produce substantially higher, and possibly lethal, HF concentrations and harmful HCl concentrations.

3.3. Flammable Gas Emissions

Evaporated electrolyte solvents were abundant in LIB emissions, as shown in Figure 6. For LFP tests, solvent concentrations were highest at low SOCs and dropped substantially at 50–75% SOCs. This pattern is likely due to increased solvent combustion at higher SOCs compared to lower SOCs, where flaming combustion did not occur. In contrast, LCO tests showed consistent solvent concentrations across SOCs, with all tests except one exceeding 50 ppm. This behavior is likely attributable to the absence of a solid casing, which allowed evaporating solvents to escape without combusting.

THC concentrations were lower than those of evaporated solvents, but they can pose a greater fire and explosion risk due to their higher flammability. Among the THC components, propane had the highest concentration peaks, while styrene was detected sporadically and mostly in LCO tests. Propane was also detected earliest among the four hydrocarbons, being emitted alongside electrolyte solvents. THC concentrations were variable, reaching 38 ppm for one LCO test (80% SOC); however, three-quarters of all tests recorded maximum concentrations below 9 ppm. These concentrations do not represent an immediate explosion hazard (Supplementary Table S1) [17] but indicate that flammable gases could build up in confined areas if many LIB cells undergo TR. THC concentrations were generally higher in LCO tests, possibly due to more venting routes. The commonly reported H₂ gas was not measured in this study but can represent the largest explosion risk during TR [18].

Figure 7 shows the timing of individual solvent species (a and c) and other species (b and d) emitted from 0% and 80% LCO tests. Solvent emissions were detected earlier than other species, occurring during the initial outgassing phase prior to TR. The four solvent species (EC, DEC, EMC, and DMC) also exhibited different onset times, with the less volatile EC released later than the other three. This observation is consistent with the findings of Bertilsson et al. [19], who reported that cyclic carbonates such as EC are substantially less volatile than non-cyclic carbonates such as DMC, with evaporation onset temperatures of 140 and 28 °C, respectively. Clear differences in the release timing of other species were also seen for the 0% SOC LCO test (Figure 7b), with emissions occurring in the following order: total solvents, HCl, HF, THC, CO, and CO₂. Particle emissions followed the CO emission profile. At higher SOCs, peaks occurred more simultaneously (Figure 7d); however, solvent, HCl, and particle emissions started rising earlier than other species.

3.4. SOC Effect on Concentration

Figure 8 shows that peak concentrations and the timing of these peaks depend on cell SOC. With increasing SOC, peak concentrations often increased and occurred sooner after TR was triggered. When cell SOC increased from 0% to 100%, HF concentration peaks increased by 23 and 10 ppm for LFP and LCO tests, with the peak occurring 6 and 3 min sooner after TR start, respectively. The reduction in time between TR starting and peak emission concentrations is important as it leaves less time to extinguish a combusting cell, vent combustion emissions, or evacuate the area before potentially lethal concentrations of HF are present. HCl and CO also had increased peak concentrations at high SOCs and had peak concentrations occurring 2–8 min sooner after TR start. NO_x and SO₂ concentration peaks occurred 3–4 min sooner after TR start, but only NO_x showed increased peak concentrations. Peak THC concentrations increased five- to eight-fold from 0% to 100% SOC and occurred 3–4 min sooner after TR start, increasing the risk of ignition or explosion.

3.5. Gas Emission Factors

While peak concentrations and emissions evolution are useful when determining local and acute impacts, EFs allow for easier extrapolation, comparison, and evaluation of overall plume impacts. EFs for selected gas species from LFP tests are shown in Figure 9 and are also reported in Supplementary Table S2. CO₂ EFs varied considerably for LFP tests, ranging from 42 mg/Wh at 0% SOC to 3157 mg/Wh at 75% SOC. Solvent EFs were lower for 50% and 75% SOCs but were consistently higher at 555–924 mg/Wh for other SOCs. Among the detected solvents, EC and DMC had the highest EFs, followed by EMC and then DEC. THC EFs were also lower at 50% and 75% SOCs, with larger within-SOC variability. Overall, solvent EFs were 4–40 times higher than THC EFs. HF EFs ranged from 67 to 73 mg/Wh at all SOCs except 100% SOC, which showed a lower EF of 34 mg/Wh. This consistency indicates that a large fraction of the available LiPF₆ is converted to HF via Equations (1)–(3) at most SOCs. HCl EFs were below 12 mg/Wh for all SOCs and showed the lowest EFs at 50% SOC. CO EFs were consistent (14–26 mg/Wh) across SOCs. SO₂ and NO_x EFs were minimal due to their small and short-lived emission peaks, reaching up to 0.2 and 1 mg/Wh in individual tests, respectively.

EFs for LCO tests are shown in Figure 10. Compared to LFP tests, LCO tests exhibited fewer variabilities across SOCs and generally had lower EFs for all species except SO₂, NO_x, and CO₂ at some SOCs. LCO solvent EFs dropped at SOCs above 60%, although SOC averages were more similar to one another than in LFP tests. Solvent EFs ranged from 160 to 485 mg/Wh and were up to 100 times higher than THC EFs. HF EFs were consistent at 9–15 mg/Wh, and HCl EFs ranged from 1 to 5 mg/Wh. HF, HCl, and THC EFs were lower than those observed in LFP tests at all comparable SOCs, with LFP EFs being higher by 3–8-, 1.5–3-, and 1.4–8-fold, respectively. Therefore, although LFP cells are thermally more stable than LCO cells, they exhibit higher EFs for HF, consistent with an earlier study [20]. CO EFs were similar between the two cell types. In contrast, average SO₂ and NO_x EFs were higher than LFP tests for most comparable SOCs, with SO₂ increasing by 9–23-fold and NO_x increasing by up to 3-fold.

Acidic gas emissions may be affected by ambient RH due to reactions between the species. For example, Zhou et al. found that increased RH limited the diffusion of gaseous HF [21]. Figure S4 shows that HF emissions increased with H₂O emissions in most tests. It is not known, however, whether this is a causal relationship or the result of generally increased emissions.

Table 2. Comparison of EFs in mg/Wh from this study and aggregated literature values. Literature values are reported as medians, with the inter-quartile range in parentheses.

Test	CO	CO2	HCl	HF	NOx	SO2	Solvents	THC
This study
0% LFP	14	42	8	68	0.2	0.1	924	62
50% LFP	15	2089	2	72	0.5	0.1	84	11
100% LFP	17	930	5	34	0.6	0.1	555	37
LCO	12–17	939–1557	1–5	9–15	0.5–0.9	0.8–2.3	164–484	3–17
Bugryniec et al. [1]
0% LFP	35 (23–46)	1435 (1101–1768)	NA	103 (55–110)	NA	50	NA	NA
50% LFP	28 (3–69)	195 (39–685)	68 (42–94)	79 (13–125)	82 (78–85)	36	NA	NA
100% LFP	19 (6–72)	84 (24–629)	149 (7–591)	52 (37–128)	183 (38–477)	30	11	10
LCO	189 (128–207)	475 (294–744)	NA	15 (10–15)	NA	NA	279 (228–319)	161

compares the EFs in Figure 9 and Figure 10 with aggregated literature values from Bugryniec et al. [1], including three LFP SOC categories and overall LCO EFs. Because of the wide spread in published data, literature values are reported as the medians, with the inter-quartile range shown in parenthesis. For LFP tests, CO EFs from this study agree reasonably well with the literature; however, literature CO EFs for LCO tests are substantially higher than those measured here. CO₂ EFs in this study are generally higher than literature values. Notably, the literature reports higher CO₂ EFs at 0% SOC but lower EFs for 50 and 100% SOC in LFP tests, which is unexpected as the smoldering of 0% SOC tests should produce less CO₂. HF EFs measured in this study are within the range reported in the literature, while literature values for HCl, NO_x, and SO₂ were much higher than our findings. Solvent emissions compare well for LCO tests but were much higher in this study for LFP tests. The opposite is true for THC emissions, with LFP tests comparing much better than LCO tests.

3.6. Emission Factor Correlations

Gas EFs vary with cell SOC, MCE, and other combustion metrics. Figure 11 shows the relationship between the ratio of detected gas mass to cell mass loss and flame temperature for LFP tests. The proportion of mass released as gaseous emissions increased from approximately 40% to 140% of total cell mass loss as flame temperature increased. This trend indicates that, although total cell mass loss during combustion remains relatively consistent [8], a greater fraction of that mass is converted to gaseous rather than particulate emissions as temperature increases. Total gas emissions also increased with maximum detected temperature for LFP tests, increasing fivefold to ~5 g/Wh with an R² of 0.64 (Supplementary Figure S3c). LCO tests did not show these relationships, likely due to challenges in accurately measuring flame temperatures for those tests.

Figure 12 shows EF variation with cell SOC for selected gases. For LFP tests, HF, solvent, and THC EFs were highest at low SOCs and decreased at mid or high SOCs. HF EFs declined by ~50% at 100% SOC, while solvent and THC EFs decreased by more than 80% at 50% and 75% SOCs, before increasing again at 100% SOC. HCl EFs were also lowest at 50% SOC but showed less variation with SOC. EFs for LCO tests showed less SOC-dependent variability than LFP tests, with HF, HCl, and solvent EFs generally decreasing as SOC increased. The HF trend is particularly notable: although HF concentrations were lower at low SOCs (Figure 5), HF EFs did not follow the same trend. The decrease in HF EF at high SOCs is consistent with several earlier studies [22,23,24], although one study reported the opposite trend [25]. This suggests that storing LIB cells at a lower SOC, a common safety practice, may reduce HF concentrations only in well-ventilated areas, where emissions do not accumulate. Conversely, storing LIB cells at 100% SOC may lead to more vigorous combustion, potentially limiting HF formation or promoting rapid neutralizing reactions upon release. Total gas emissions increased at mid-range SOCs for both cell types (Supplementary Figure S3a,e) and more dramatically for LFP tests, matching results from earlier studies [20,24].

Figure 13 shows EF variation with MCE, providing insights into the dependence of emissions on combustion conditions. HF did not show a clear overall correlation with MCE; however, when 50% and 75% SOC data were excluded, a trend of decreasing EF with increasing MCE became apparent for LFP tests. HCl, solvent and THC EFs all decreased with increasing MCE for both cell types, suggesting that these species are consumed through combustion or other reactions during hotter, more complete combustion. Solvent EFs also showed strong negative correlation with the proportion of emissions from flaming combustion (Figure S2), with R² values of ~0.9 for both cell types. This indicates that evaporated solvents were not released during flaming combustion and were instead completely combusted. Total gas emissions increased with MCE for both cell types (Supplementary Figure S3d,h).

4. Limitations and Discussion

Several limitations of this study should be acknowledged:

First, only two cell types—a cylindrical LFP cell and a pouch-style LCO cell—were evaluated, with TR initiated by overheating. Emissions are expected to vary with cell chemistry, form factor, capacity, and failure trigger. Systematic testing across a broader range of conditions is needed to establish more comprehensive emission characteristics.

Second, caution is warranted when extrapolating data from single-cell tests to multi-cell or full LIB pack configurations. As HF exceeded inhalation limits in some tests, more severe toxic conditions would likely arise if larger LIB packs were to combust in enclosed environments. However, combustion behavior and emissions from enclosed LIB packs may differ significantly from open burning of individual cells due to variations in heat transfer, confinement, and oxygen availability. Direct measurements of LIB pack combustion under realistic configurations, such as those used in electric vehicles, would provide more representative emissions data.

Third, several important gases, including hydrogen and other potentially toxic fluoride- or phosphate-containing species beyond HF, were not measured. More comprehensive quantification of these species would improve assessments of toxicity, fire, and explosion hazards.

5. Conclusions

TR of LIBs is an important fire safety risk. LFP cells only outgassed and smoldered at lower SOCs (0 and 30%), while LCO cells had more vigorous flaming combustions at all SOCs, indicating that LFP cell chemistries are safer than LCO. Peak concentrations, time between TR start and maximum concentrations, and EFs depended on cell SOC, especially for LFP tests. These SOC-dependent changes can be linked back to combustion efficiency, which increased with SOC for both cell types, causing combustible species to be converted to CO and CO₂.

Both LFP and LCO cell combustion released significant amounts of hazardous gases. Among the measured gases, HF presents the highest toxicity risk, and reached lethal concentration levels during some tests. Even though LFP cells do not combust as vigorously as LCO cells, HF EFs were higher than those of LCO tests by a factor of four to eight among the tested SOCs. LFP tests also had higher EFs for HCl, solvents, and THC than LCO tests, while LCO tests had higher EFs for NO_x and SO₂. However, these measured toxic or flammable gases were below their respective inhalation or explosion limits.

Gas and particle emissions showed different onset and peak times. For lower SOC tests, electrolyte solvents, particularly DEC, EMC, and DMC, were emitted during the outgassing phase before TR, followed by HCl, HF, THC, particulate matter, CO, and CO₂. Less peak separation was detected during flaming combustion; however, solvent and particle emissions onset still preceded other emissions. The fact that solvent and particle emission occurred up to one minute before other species suggests that these species should be targeted for early detection of LIB failure.

This study seeks to provide a detailed analysis of gaseous LIB combustion emissions in companion with previous publications on particulate emissions [7,8]. This analysis will assist first-responders, LIB manufacturers, and civic authorities to plan for and respond to LIB fires by allowing for a better understanding of released emissions characteristics, concentrations, and EFs, as well as associations among SOC, combustion behavior, and peak concentrations.