Toxicity Assessment of Gas, Solid and Liquid Emissions from Li-Ion Cells of Different Chemistry Subjected to Thermal Abuse

Abstract

Lithium-ion batteries (LIBs) are employed in a range of devices due to their high energy and power density. However, the increased power density of LIBs raises concerns regarding their safety when subjected to external abuse. The thermal behavior is influenced by a number of factors, i.e., the state of charge (SoC), the cell chemistry and the abuse conditions. In this study, three distinct cylindrical Li-ion cells, i.e., lithium nickel cobalt aluminum oxide (NCA), lithium titanate oxide (LTO), and lithium iron phosphate (LFP), were subjected to thermal abuse (heating rate of 5 °C/min) in an air flow reactor, with 100% SoC. Venting and thermal runaway (TR) were recorded in terms of temperature and pressure, while the emitted products (gas, solid, and liquid) were subjected to analysis by FT-IR and ICP-OES. The concentrations of the toxic gases (HF, CO) are significantly in excess of the Immediate Danger to Life or Health Limit (IDLH). Furthermore, it is observed that the solid particles are the result of electrode degradation (metallic nature), whereas the liquid aerosol is derived from the electrolyte solvent. It is therefore evident that in the event of a LIB fire, in order to enhance the safety of the emergency responders, it is necessary to use appropriate personal protective equipment (PPE) in order to minimize exposure to toxic substances, i.e., particles and aerosol.

1. Introduction

Lithium-ion batteries (LIBs) are employed in many fields, such as electric vehicles (EVs), portable devices and energy storage systems (ESSs), due to their superior performance in comparison to previous generations of batteries [1]. However, the primary safety concern associated with these devices is thermal runaway (TR) [2], which occurs when the device is subjected to an external or internal abuse, i.e., electrical, mechanical, or thermal [3], which are conditions that fall outside the safety window of the LIBs [4].

To enhance the safety of commercial Li-ion cells, it is possible to equip them with protective devices, such as a current interrupt device (CID), top and bottom vents [5] and a protection circuit [6]. In the case of cylindrical cells, only the CID and the top vent are mandatory [7]. These devices should be activated once the internal pressure increases above pre-defined values [8], such as 1.0–1.2 MPa for the CID or 2.2–2.3 MPa for the venting [9,10].

Abuse conditions can be considered as a direct cause of the TR of LIBs [11]. They result in the internal decomposition of the components through multiple exothermic reactions across a wide temperature range (from above 70 °C to 300 °C) [12], accompanied by the emission of smoke, gases, liquids, and aerosol products with relative toxicity, the projection of fragments, fire, and/or explosion [13]. Accidents of fire involving EVs and ESSs using LIBs are on the rise, causing losses [14].

The composition of the emitted products is dependent on a combination of several parameters, including the characteristics of the Li-ion cells and the abuse test conditions [15]. The principal parameters associated with Li-ion cells are the chemistry of the electrodes and the electrolyte, the electrical specifications (e.g., capacity and state of charge (SoC)), the geometry (e.g., cylindrical, prismatic, pouch, or coin), and the relative dimensions (e.g., for cylindrical cells, dimensions may be 18650 or 21700). With regard to the abuse conditions, the tests may be conducted by subjecting the cells to thermal, mechanical, or electrical abuse in a variety of experimental settings. These include the atmosphere (e.g., N₂, air) [16], as well as confined or open spaces [17]. In terms of gaseous products, the main species detected are permanent gases (e.g., CO and CO₂), a variety of hydrocarbons (e.g., CH₄, C₂H₂, and C₂H₄), fluorinated compounds (e.g., HF and phosphorus oxyfluoride (POF₃) [18]), hydrogen, and the vapors of electrolyte solvents (such as dimethyl carbonate (DMC), diethyl carbonate (DEC), and ethylene carbonate (EC) [19,20,21,22,23]. For a fully charged cell (100% SoC), oxygen consumption is higher than for other SoC levels. Additionally, temperature increases sharply after the onset of TR, resulting in a higher maximum peak and a variation in gas production [24]. Therefore, it can be concluded that a higher SoC corresponds to a higher amount of gas produced [17]. With regard to the emissions of HF, there is a conflict regarding the trend of this compound. The results presented by Larsson et al. [25] indicate that the greater amount of HF occurs due to low SoC. However, Peng et al. [26] demonstrated that the increase in SoC also increases the production of HF and the toxic or highly toxic properties of the gases. The composition of the gas can be identified and quantified by means of different analytical techniques, including Fourier transform infrared spectroscopy (FT-IR) and gas chromatography (GC). Both techniques can be connected online to the reaction chamber for real-time acquisition [25,26,27]. FT-IR is highly sensitive to HF compounds at low concentrations, exhibiting good linearity. However, the GC method is unable to analyze toxic species, such as fluorinated compounds [27]. The identification and quantification of HF throughout thermal abuse tests is of significant importance, as HF is the most toxic compound in gaseous emissions, with CO being the second most toxic. In fact, the National Institute for Occupational Safety and Health (NIOSH) [28] has established Immediately Dangerous to Life or Health (IDLH) values for CO and HF emitted during TR, with prescribed limits in order to evaluate risk to people exposed to an LIB fire [15,16,26,29]. Furthermore, organic carbonates are regarded as toxic species, although it should be noted that IDLH values are not available for them. Instead, CO and HF have been designated with IDLH values, 1200 ppm [30] and 30 ppm [31], respectively. A further type of product that may be released during TR is solid particles, resulting from the degradation of the electrode foils. The primary physical and chemical characteristics of the solids that must be evaluated are their diameter and chemical composition [32]. The solid ejection is typically composed of fine particles, comparable to an aerosol, with diameters ranging from 8.5 to 300 µm. However, it also includes particulate matter with diameters below 2.5 µm (PM 2.5) [33]. The particles are composed of carbon, organic compounds (e.g., carbonates), transition metals, and transition metal oxides, in accordance with the initial composition of the LIBs [34]. It is also possible for liquid products to be emitted from Li-ion cells during TR, dependent on the temperature reached and the initial composition of the electrolyte. The composition of the carbonate mixture varies according to the manufacturer, such as EC:DEC with a ratio of 1:1 v/v, or EC:DEC:DMC with a ratio of 2:1:2 v/v. The presence of liquid emissions in the aerosol phase is strictly dependent on the temperature reached during the TR. The highly volatile nature of DMC and DEC, with boiling points of 90 °C and 126 °C, respectively, reduces the probability of detecting liquid emission in comparison to EC, which has a boiling point of 244 °C, resulting in its emission as a liquid aerosol.

In order to investigate the effect of the cell chemistry on thermal behavior, three different cylindrical Li-ion cells were considered: lithium nickel cobalt aluminum oxide (LiNiCoAlO₂, NCA), lithium iron phosphate (LiFePO₄, LFP), and lithium titanate oxide (Li₂TiO₃, LTO). The cells were selected for the comparison of two distinct cathode compositions (NCA vs. LFP) and two different anode compositions (NCA vs. LTO). The NCA and LTO cells have the same cathode composition (NCA), but differ in their anodes. NCA and LFP cells use graphite as anodes, while LTO cell use LTO. Conversely, the NCA and LFP cells have different transition metal oxides as cathode materials. Then, each cell with the same SoC of 100% was subjected to thermal abuse (heating rate of 5 °C/min) in a tubular reactor connected to an online FT-IR spectrometer. The heating rate of 5 °C/min has been selected for the evaluation of the thermal behavior of the LIBs when subjected to a constant heating process, which is comparable to an external heat source. The low heating rate has been selected in order to prevent the saturation of the gas instrumental signal and to facilitate observation of the various phases of the TR. In fact, the application of a higher heating rate resulted in a significant reduction in the time interval between venting and TR, with a more destructive effect, already observed in previous work [35]. Prior research demonstrated that thermal abuse at a rate of 5 °C/min was most closely aligned with the established standard for assessing the thermal stability of cells (e.g., UN/ECE Regulation No. 100), which predicts the lowest temperature at which TR is reached. With regard to the SoC, 100% SoC has been selected for the purpose of studying LIBs at an unfavorable charge level. In fact, fully charged Li-ion cells were observed to present a greater risk than lower operational SoC and a higher amount of gas produced.

The cell surface temperature and reactor pressure were monitored in order to establish a correlation between the key events, i.e., venting and TR, and the emitted gases. Furthermore, the solid and liquid collected at the end of the thermal abuse test were characterized in order to evaluate the metallic composition using inductively coupled plasma–optical emission spectroscopy (ICP-OES) and to evaluate the organic components by FT-IR. The objective was to identify the safest cathode and anode materials, considering both TR behavior and emitted products, by comparing different Li-ion cells. To evaluate the toxicity of the gases produced, the amounts of CO and HF were compared with their IDLH levels. Finally, the mass of metals present in the ejected particles was quantified for the different Li-ion cells. A more comprehensive understanding of the hazards posed by cell products can assist in the identification of the personal protective equipment (PPE) required for emergency responders responding to an LIB fire, thereby ensuring their safety. The most efficacious equipment against such cases includes gas masks, or air-purifying respirators, which protect against the inhalation of noxious substances, tailored to the specific hazards and airborne concentrations [36].

2. Materials and Methods

2.1. Li-Ion Cells

Thermal abuse tests were carried out on three cylindrical 18650 Li-ion cells. The chemical composition, as determined by analytical evaluation [37], and the electrical specification, as detailed in the respective Technical Data Sheet (TDS), are reported in

Table 1. Cylindrical Li-ion cells and relative electrical specification according to the TDS.

Li-Ion Cell	Chemical Composition [37]	Electrical Specification
NCA 18650	Anode: Graphite (C); Cathode: Lithium nickel cobalt aluminum oxide (NCA); Electrolyte: DMC:DEC:EC (2:1:1 v/v); Separator: polyethylene (PE).	Parameter	Value
		Nominal Capacity	3250 mAh minimum; 3350 mAh typical
		Nominal Voltage	3.6 V
		Max. operating Voltage Range	2.75 V to 4.2 V
		Standard Charge	1625 mA and 4.20 V for 4.0 h
LTO 18650	Anode: Lithium titanium oxide (LTO); Cathode: Lithium nickel cobalt aluminum oxide (NCA); Electrolyte: DMC:DEC:EC (1:1.5:1.5 v/v); Separator: polypropylene (PP).	Parameter	Value
		Rated Capacity	1300 mAh minimum; 1350 mAh typical
		Nominal Voltage	2.4 V
		Watt-hour rating	3.12 Wh
		Max. operating Voltage Range	1.60 V ± 50 mV to 2.80 V ± 50 mV
		Standard Charge	1.3 A/1 C
		Standard Discharge	1.3 A/1 C
LFP 18650	Anode: Graphite (C); Cathode: Lithium iron Phosphate (LFP); Electrolyte: DMC:DEC:EC (1.5:2:1 v/v); Separator: PP.	Parameter	Value
		Rated Capacity	1300 mAh minimum, 1320 mAh typical
		Nominal Voltage	3.2 V
		Watt-Hour rating	4.16 Wh
		Max. operating Voltage Range	2.3 V ± 50 mV to 3.70 V ± 50 mV
		Standard Charge	650 mA/0.5 C
		Standard Discharge	1300 mA/1 C

.

Prior to testing, the Li-ion cells were subjected to a standard procedure of five charge–discharge cycles using the Battery Test System BaSyTec CTS from Thasar (Milan, Italy) to ensure the correct formation of the solid electrolyte interface (SEI). The cells were initially discharged at a C/2 discharge current, resulting in a fully discharged state corresponding to the minimum operating voltage value, which is indicated by the term 0% SoC. Following an interval of 1 h to allow the cell to stabilize, a first phase of charge at a constant current (CC) of C/2 is applied until the maximum operating voltage value is reached, indicated by with the term 100% SoC. Subsequently, a second phase of charge is applied at constant voltage (CV), with the current decreasing until it reaches a value of 0.5 A. After another 1 h interval, the cycle is repeated 4 additional times. Upon completion of the five charge–discharge cycles, a 100% SoC was achieved, thereby necessitating the identical charging phase that was applied during the SEI formation process. The maximum operating voltage range is directly correlated with the SoC of the Li-ion cells, as indicated in the TDSs. Therefore, the 100% SoC in terms of voltage differs for each cell type—4.2 V for NCA, 2.80 V for LTO, and 3.70 V for LFP—in accordance with the maximum operating voltage range reported in the TDSs (Table 1).

2.2. Methods

2.2.1. Thermal Abuse Tests

The thermal abuse tests were carried out in a stainless-steel reactor placed in the center of a cylindrical furnace (900 W). The reactor was connected to the FT-IR gas cell via a filter unit and a transfer line. The reactor inlet was connected to a gas supply line through a mass flow controller (MFC) to regulate the inlet air flow to 500 NmL/min. Additionally, two thermocouples (TCs) were placed on the cell surface: TC₁ at 0.5 mm from the vent valve (negative electrode side) and TC₂ at 0.5 mm from the cell bottom (positive electrode side). The TCs, MFCs, and pressure transmitter were connected to a data logger for continuous recording.

The gases generated during the thermal abuse test were purged from the reactor by a controlled draft system. They were purified of any particulate matter by passing through a polytetrafluoroethylene (PTFE) filter (47 mm diameter and 1.0 µm pore size, Whatman^TM, Little Chalfont, UK) which was connected to a transfer line leading to the FT-IR gas analysis cell (Spectrum 3, Perkin Elmer, Milan, Italy). The filter, transfer line and cell were heated to 180 °C to prevent condensation of the vapors.

A graphical image of the laboratory plant for the thermal abuse tests is shown in Figure 1a, while Figure 1b shows the internal configuration of the reactor, with the cell holder for the 18650 cell situated at its center and the two TCs positioned on the upper surface at an equal distance from the two poles.

Further details regarding the laboratory system can be found in reference [35].

2.2.2. Emission Characterization Methods

The products emitted during the thermal abuse test were subjected to analysis in order to assess their composition and concentration. The principal products are the gaseous, solid, and liquid effluents. To characterize the different emissions, an online FT-IR spectrometer was connected to the reactor for gas analysis, while ATR-FT-IR and ICP-OES were used for the analysis of the solid and liquid effluents, respectively.

Gas Emissions

An online FT-IR spectrometer (Spectrum 3, Perkin Elmer) was connected to the reactor outlet in order to permit the continuous monitoring of the emitted gases. It is important to underline that this technique does not allow the detection of bimolecular species like H₂. The spectra were collected with a Mercury Cadmium Telluride (MCT) detector with the following spectrometric parameters: a resolution of 4 cm⁻¹, a spectral range between 4500 and 650 cm⁻¹, and a scan/spectrum of 8. The TimeBase, tool of the Spectrum IR software (application version 10.7.2.1630, PerkinElmer), was used for continuous acquisition, while the SpectrumQuant software (application version 10.7.2.1630, Perkin Elmer) was utilized for identification and quantification through the use of standard spectra and calibration lines. The standard spectra and the relative calibration lines for the single species (CO, HF, DMC, DEC, and EC) were obtained through the utilization of technical gases purchased from SIAD (Bergamo, Italy). From the individual spectra of the pure gaseous species, the primary wavenumbers were identified and expressed in cm⁻¹. Subsequently, the standard spectra were obtained at varying concentrations in order to construct the calibration line, utilizing the SpectrumQuant software (application version 10.7.2.1630, Perkin Elmer). Then, the spectra of the various species were compared, and a single wavenumber was selected for each. It is imperative that the selected wavenumber is present and exhibits narrow and well-defined characteristics across all the spectra obtained at varying concentration values.

Table 2. Concentration range and principal wavenumbers of gases.

Species	Concentration Range (ppmv)	Wavenumber (cm−1)
EC	31.72–277.55	1079; 1087; 1096; 1122; 1131; 1141; 1385; 1860; 1868; 1876; 3735
DMC	23.25–271.25	917; 925; 985; 990; 996; 1295; 1455; 1463; 1768; 1780; 2199
DEC	15.6–364.0	791; 862; 1021; 1093; 1258; 1302; 1374; 1409; 1448; 1746; 1742
HF	29.0–811.0	4172-4175 (4110); 4202–4203
CO	1000–900,000	2115; 2173

Note: the selected wavenumber is underlined.

reports the relative linearity range of each gas and the principal wavenumbers of each species, with the unique wavenumber selected for calibration underlined.

Solid and Liquid Emissions

The nature of the emitted product determined the type of analysis performed. The organic components were analyzed using ATR-FT-IR, while ICP-OES was used for the metallic components.

The solid and liquid effluents were analyzed using an ATR-FT-IR, Perkin Elmer Spectrum 3TM FT-IR Spectrometer. The spectra were acquired with the following parameters: a resolution of 4 cm⁻¹, a spectral range between 4000 and 650 cm⁻¹, 8 scans/spectrum, and a TGS detector. The spectra were identified through a comparison with those available in the Perkin Elmer library (Perkin Elmer).

To identify and quantify the metals present in the electrode, an ICP-OES analysis was conducted in accordance with the EPA 200.8-1 method [38]. A sample of 0.150 g was placed in a muffle furnace at 530 °C for 8 h in order to eliminate organic compounds; 10 mL of reverse aqua regia (3:1 v/v HNO₃:HCl) was added to each sample. The digestion process was conducted at room temperature (25 °C) for a period of 3 h with continuous agitation. Upon completion of the digestion process, the samples were filtered through an ultra-slow filter into a flask and subsequently diluted with ultrapure water, resulting in a final volume of 100 mL. Further dilutions were conducted as necessary.

3. Results

For each chemical composition (NCA, LTO and LFP), three Li-ion cells were subjected to a thermal abuse test. The relative temperature and pressure associated with the main events, such as venting, TR, and maximum peak, were recorded. Additionally, the gas emission profiles were analyzed using FT-IR, and the characterization of solid and liquid products was performed using ATR-FT-IR and ICP-OES, as described in the previous section.

3.1. Temperature and Pressure Profiles

During each thermal abuse test, the cell surface temperature and the pressure inside the reactor were monitored continuously, as shown in Figure 2.

The first general observation from the thermal abuse tests is that the temperature and pressure values for the NCA cells are higher than those for the LTO and LFP cells.

Venting occurs at 151 ± 4 °C for NCA, at 181 ± 16 °C for LTO, and at 212 ± 9 °C for LFP, with corresponding pressure increases inside the reactor to 0.125 barg for NCA, 0.123 barg for LTO, and 0.714 barg for LFP. The observed drop in cell surface temperature, recorded by TC₁ and TC₂, is due to the release and evaporation of the electrolyte, which absorbs heat during the endothermic phase change. After the electrolyte evaporates, the temperature begins to rise again. TR onset occurs at 204 ± 1 °C and 0.037 barg for NCA, at 210 ± 4 °C and 1.313 barg for LTO and at 250 ± 12 °C and 1.379 barg for LFP. The maximum temperatures and pressures reached at the peak were 511 ± 96 °C and 5.071 barg for NCA, 381 ± 89 °C and 1.802 barg for LTO and 406 ± 62 °C and 2.202 barg for LFP. The differences in maximum temperature and pressure are significant, with NCA cells reaching over 200 °C higher temperatures compared to LTO and LFP cells. The same trend is observed for the pressure values, with higher values for NCA than the other two chemistries.

From the temperature profiles shown in Figure 2, it can be seem that there is a discrepancy in the recorded temperatures between TC₁ and TC₂. In general, TC₁ monitors higher temperatures than TC₂, which can be attributed to the different locations of the TCs, as shown in Figure 1b. The position of TC₁ in close proximity to the vent valve results in its exposure to the hot gases emitted by the valve and the heat released by the exothermic reactions occurring between the different species. On the other hand, TC₂, in a position opposite to TC₁, is more affected by the temperature reached by the cell surface and less affected by the hot gases emitted. In fact, at the end of the TR and at the end of the phase of increased gas emission from the vent valve, the temperatures recorded by the two TCs become comparable again.

Thermal abuse tests were carried in triplicate for each cell chemistry.

Table 3. Time (s), temperature (°C) and pressure (barg) of the key events during thermal abuse tests.

Phase	Parameters	NCA	LTO	LFP
Venting	t (s)	2192 ± 154	2461 ± 33	2580 ± 130
	T (°C)	154 ± 4	184 ± 11	205 ± 10
	P (barg)	0.127 ± 0.002	0.106 ± 0.024	0.444 ± 0.383
TROnset	t (s)	2666 ± 93	2741 ± 88	2725 ± 94
	T (°C)	206 ± 3	221 ± 14	242 ± 13
	P (barg)	0.037 ± 0.001	0.886 ± 0.605	1.180 ± 0.282
Max	t (s)	2702 ± 76	2772 ± 48	2806 ± 16
	T (°C)	562 ± 96	359 ± 81	358 ± 70
	P (barg)	4.305 ± 1.083	2.050 ± 0.854	1.655 ± 0.774

shows the mean values and the relative standard deviation for time, temperature and the pressure obtained from the three thermal abuse tests carried out for each cell chemistry.

A comparison with the temperature data reported in the literature is affected by the use of different materials and methodologies across the various studies. In fact, the thermal behavior is the result of a complex combination of multiple factors, including the intrinsic properties of the Li-ion cells and the specificities of the abuse test method [15]. It is important to acknowledge that discrepancies in cell characteristics (e.g., chemistry and voltage), equipment (e.g., open or closed system), and experimental conditions (e.g., type of feed gas and heating rate) can significantly influence the observed maximum temperature.

In order to fully characterize the Li-ion cell, it is insufficient to describe the cathode composition alone; the chemical composition of the anode and electrolyte can also vary between manufacturers. Furthermore, the specific solvents utilized (e.g., DEC, DMC, and EC) and their respective ratios may vary, and this information is not typically included in the TDSs. Furthermore, the voltage range may vary depending on the intended use of the cell. Consequently, a 100% SoC does not necessarily correspond to the same voltage level across different cells. Regarding the experimental conditions, the primary parameters that can influence the thermal behavior are the degree of heat exposure and the atmospheric conditions. In particular, an evaluation of the impact of the operating conditions, in terms of heating rate and atmosphere, has been conducted in the same laboratory setup [35]. In particular, thermal abuse tests were conducted on NCA at rates of 5 and 10 °C/min, both in air and in N₂ in order to highlight the differences. The temperature values attained during the venting phase were not significantly different between tests conducted at the same heating rate in air or nitrogen. However, tests performed at a higher heating rate (10 °C/min) exhibited higher temperatures. This behavior is attributable to the fact that the venting phase is defined by the internal pressure set point of the valve. In contrast, the TR onset and maximum temperature values were found to be more influenced by the feed gas employed in the tests than by the heating rate. Indeed, following the venting process, the emitted gases may undergo a chemical reaction with the surrounding atmosphere. In the case of air, the elevated temperature values observed are influenced by the combustion of organic compounds from electrolyte, emitted by the cell. In N₂, combustion reactions are significantly constrained by the oxygen production resulting from the degradation of cathode active materials. Consequently, the atmosphere exerts a profound influence on both the temperature reached by the Li-ion cell and the products emitted, including gases, solids, and liquids.

So, it is imperative to consider the previous considerations for any comparison. In this case, the maximum temperatures recorded in this study were compared to those reported in the literature by Golubkov et al. (2014 and 2015) [21,22] due to the fact that the thermal abuse tests were conducted in a reactor chamber on a single NCA and LFP 18650 cylindrical cell, which was similar to the Li-ion cells presented in this work. In any case, the environmental conditions inside the reactor chamber differed between the present study and Golubkov et al.’s studies, which were conducted in an atmosphere of N₂. Accordingly, Golubkov et al. observed a maximum temperature of 900 °C for NCA cells [22], compared to 404 °C for LFP cells [21].

3.2. Emission Characterization

3.2.1. Gas Emission Profiles

During each thermal abuse test, the concentrations of HF, EC, DMC, and CO were monitored continuously, as reported in Figure 3.

The concentration profiles obtained for the three chemistries, shown in Figure 3, align with the temperature and pressure profiles, shown in Figure 2. Specifically, the emission peaks correspond to the main events, such as venting and TR. The emission peaks are well defined and precise over time, primarily due to the increase in pressure inside the reactor. Once the gaseous species are emitted, their concentrations quickly return to near zero. However, the timing of emissions is affected by the reactions that can take place inside the cell, leading to different emission profiles for the three chemistries studied, despite the presence of the same species. A notable difference is observed in CO emissions at the maximum temperature. For NCA cells, CO is the most abundant species released (2.25 × 10⁵ ppmv). In contrast, for LTO cells, the CO peak value is lower than that of CO₂ (1.45 × 10⁵ vs. 2.75 × 10⁵ ppmv). For LFP cells, CO is entirely converted to CO₂, with the CO concentration being negligible (nil ppm) and a CO₂ concentration at 1.65 × 10⁵ ppmv. This behavior is attributed to the higher stability of the LFP cathode material, which remains stable up to 500 °C [24], a temperature not reached even during TR (maximum cell surface temperature around 310 °C).

The composition of the gas emissions can be measured in various experimental conditions using different instruments, such as GC or FT-IR. Values in the literature typically refer to the average concentration of gases, with the ratio between these species varying depending on the SoC and the chemical composition of the electrodes. A comparison of the main species emitted from the Li-ion cells in this study and the total amount of toxic gases reported in the literature [16,39,40] from thermal abuse tests on Li-ion cells is provided in

Table 4. Toxic gases, expressed in g/kWh, emitted by Li-ion cells at 100% SoC during thermal abuse tests (conducted in air), as obtained in this study and reported in the literature.

Li-Ion Cell	SoC (%)	CO (g/kWh)	HF (g/kWh)	Reference
NCA	100	11	0.06	this study
LTO	100	51	0.38	this study
LFP	100	25	0.30	this study
LCO	100	n.r.	20	[39]
LFP	100	n.r.	170	[39]
LFP	100	n.r.	55	[39]
LFP	100	n.r.	25	[39]
LFP	100	n.r.	55	[39]
NCA	100	n.r.	55	[39]
LFP	100	10–65	40–145	[16]
NMC	100	110–210	5–30	[16]
LMO	100	161	37	[40]

n.r.: not reported.

.

The data reported in Table 4 are normalized by the kilowatt-hour (kWh) of the Li-ion cell to account for differences in cell capacity. At the same SoC of 100%, the total amount of HF released, expressed in g/kWh, is significantly higher for LFP compared to NMC. Conversely, CO is released in greater amounts during the thermal abuse tests on NMC cells than the LFP. The CO data reported in this study are comparable to those reported in the literature. However, the amount of HF quantified in this work is lower than reported in the literature. The discrepancy in the quantity of HF can be attributed to the different geometries and dimensions of the cells employed in this work, which were 18650 cylindrical, in comparison to those documented in the literature, such as pouch and prismatic. Despite the normalization to kWh, the differences observed in the amount of gas produced are due to the internal composition of the cell. The production of HF is attributable to the degradation of the electrolyte, as well as the degradation of the polyvinylidene fluoride binder (PVDF) present in the cathode composition of the cells. The mass weight of each component, as a function of geometry and size, exhibits a considerable degree of variation. For instance, an 18650 cell weighs approximately 45 g, whereas a pouch cell weighs between 600 and 900 g. Therefore, a notable discrepancy in the weight of the initial cell’s components results in a considerable variation in the quantity of products emitted.

To assess the toxicity of the released species, the concentration values must be compared with the IDLH limits. The total amount released for each species was obtained by integrating the area under the concentration–time curve. These values were normalized for a 30 min period to compare with the IDLH limits reported by the NIOSH [28], as shown in

Table 5. Mean concentration (ppm) for the gases emitted during the thermal abuse tests.

Li-Ion Cell	DMC (ppm)	EC (ppm)	HF (ppm)	CO (ppm)
NCA	48 ± 37	222 ± 88	49 ± 1	7564 ± 705
LTO	99 ± 3	235 ± 25	101 ± 12	9018 ± 131
LFP	47 ± 7	56 ± 12	110 ± 23	6060 ± 826
IDLH	n.a.	n.a.	30 [31]	1200 [30]

n.a.: not available.

.

The data in Table 5 indicate that, regardless of the chemical composition, the concentration values exceed the IDLH values of 30 ppm for HF [31] and 1200 ppm for CO [30] over a 30 min period as set by NIOSH. In the worst case, the HF emissions from the LFP cell exceed the limit by an order of magnitude. However, HF and CO are not the only toxic species present in the gas emitted during TR. Organic carbonates are also considered toxic, although the IDLH is not available for these species.

3.2.2. Liquid and Solid Characterization

The residues inside the reactor, both solid and liquid, as well as the condensate phase on the filter, were collected for subsequent characterization analysis through ATR-FT-IR and ICP-OES. The amounts of each product are reported in

Table 6. Weight of the cells and of the products collected, expressed in g, for the thermal abuse tests on NCA, LTO, and LFP at 100% SoC.

Li-Ion Cell	Δcell (g)	Condensate Phase Weight (g)	Solid Weight (g)	Liquid Weight (g)
NCA	17.39 ± 1.38	n.p.	4.010 ± 0.000	n.p.
LTO	8.80 ± 0.00	0.131 ± 0.000	n.p.	n.p.
LFP	8.88 ± 4.94	0.076 ± 0.014	0.045 ± 0.000	0.034 ± 0.016

n.p.: not present.

and shown in Figure 4.

According to the data reported in Table 6, a greater loss of material, calculated as the different between the initial weight and the final weight of the Li-ion cell (indicated with Δcell), was observed for NCA cell, with a value of 17.39, compared to 8.80 g and 8.88 g for and LTO and LFP cells. This loss is due to the release of gases, solids, and liquids in varying amounts depending on the cell chemistry. Liquid was collected only from the LFP cells (Figure 4f), while solids and condensates were collected for all cell chemistries.

The different nature of the residues is shown in Figure 4. It reveals a similar condensate phase, in brown, collected on the filter of LTO (Figure 4c) and LFP (Figure 4e). In contrast, the residue collected from NCA cells is a black powder (Figure 4a).

An ATR-FT-IR analysis was conducted on the filters from the thermal abuse tests of LTO and LFP cells, given the organic nature of these samples in contrast to the inorganic residue from the NCA tests. Furthermore, ATR-FT-IR analysis was performed on the liquid ejected from the LFP cells. The ATR-FT-IR peaks, along with their identifications and intensities (which were rated as very weak (vw), weak (w), medium (m), strong (s), or very strong (vs)), are reported in

Table 7. Wavenumbers (cm⁻¹) for the condensate phase on the LTO and LFP filters and the liquid ejected from LFP with relative intensities (vw, w, m, s or vs) and identification.

LTO—Condensate Phase		Identification		LFP—Liquid		Identification
2950.77	s	PP	s	3530.05	vw	EC	vw
2918.11	vs	PP	vs	2998.09	w	EC	vw
2867.48	m	PP	m	2925.92	w	EC	vw
2839.13	m	PP	m	1960.02	w	EC	w
1452.26	m	PP	s	1797.05	vs	EC	vs
1375.71	s	PP	vs	1771.57	vs	EC	vs
1166.96	m	PP	m	1554.11	w	EC	w
997.40	m	PP	m	1481.90	m	EC	m
972.10	m	PP	m	1455.47	w	EC	w
841.20	m	PP	m	1390.23	m	EC	m
LFP—Condensate Phase		Identification		1260.76	w	EC	w
2952.36	m	Paraffin oil	s	1158.80	vs	EC	s
2922.26	m	PTFE Filter	m	1069.86	vs	EC	vs
2853.87	w	Paraffin oil—PEO	vs	971.76	m	EC	s
2552.80	w	Paraffin oil—PEO	vs	893.60	m	EC	m
1719.98	w	Paraffin oil	vw	845.69	m	EC	m
1682.76	vw	Paraffin oil	w	773.14	s	EC	s
1376.78	s	Paraffin oil—PEO	m	716.12	s	EC	s
1111.39	w	PEO	w	558.54	w	EC	m
729.74	m	Paraffin oil	m	449.05	vw	EC	w

EC: ethylene carbonate; PEO: polyethylene oxide; PP: polyethylene.

for the LTO and LFP filters and the liquid ejected from the LFP.

The peaks identified in the LTO filter are attributed to the separator materials. Indeed, the plastic separator used in this cell is made of PP, as reported in Table 1. The analysis of the LFP filter led to the discovery of more intriguing results. The most intense peaks identified in the LFP filter can be attributed to the presence of paraffin oil and polyethylene oxide (PEO). Such materials may be used in the composition of Li-ion cells with a view to enhancing their properties. The detection of paraffin oil is related to the application of paraffin wax as a protective layer for the anode material or separator [41]. The release of paraffin oil is observed as the temperature of the cell rises, with the melting point of paraffin wax occurring at approximately 55 °C and decomposition occurring above 240 °C. In addition to paraffin oil, the main peaks identified were due to PEO [42]. PEO is typically used as a multifunctional binder to partially replace the commonly used PVDF binder. PEO offers several advantages, including good adhesive quality, high ionic conductivity, high flexibility, low cost, excellent processability, and environmental friendliness. The utilization of high-conductivity binders has been demonstrated to significantly enhance battery cycling performance [43].

The liquid ejected from the LFP was identified as ethylene carbonate (EC), the solvent with the highest boiling temperature among the carbonates in the electrolytic solution, at approximately 244 °C. The presence of this liquid is attributable to the fact that the surface temperature of the cell in question did not exceed 350 °C during the TR event. This prevented the evaporation of the entirety of the EC present in the solution.

Subsequently, the metals were quantified through the use of ICP-OES analysis. The quantities of metals are reported in

Table 8. Quantification of the metals on the filter (mg) for NCA, LTO, and LFP.

Li-Ion Cell	Al	Co	Cu	Fe	Li	Mn	Ni	P	Ti
NCA	324.1	266.9	480.2	n.p.	172.1	0.08	<LOD	<LOD	n.p.
LTO	3.327	<LOD	0.5370	n.p.	0.6418	0.0223	<LOD	<LOD	0.3013
LFP	0.780	n.p.	0.065	0.440	0.016	n.p.	n.p.	0.135	n.p.

n.p.: not present. LOD: limit of detection.

and are expressed in mg.

The relatively low concentration of Mn in NCA and LTO can be attributed to its lower percentage in the initial cathode composition compared to the other metals. Meanwhile, phosphorus (P) was undetected (below the instrument’s LOD) in NCA and LTO due to the nature of this metal. In LIBs, phosphorus is present in the form of a salt dissolved in the electrolyte. Consequently, the majority of P exists in a liquid state and evaporates during heating, resulting in its release as a gas rather than a solid. An exception is observed in the chemical composition of LFP, where P is present in the oxide of the cathode material.

4. Discussion

In order to highlight the differences in thermal abuse data among the three chemistries, it is possible to categorize the results according to their internal composition. The cathode composition of NCA and LTO cells is identical, consisting of NCA, whereas the anode composition differs: graphite in the case of NCA and lithium titanium oxide in LTO. Conversely, NCA and LFP have different transition metal oxides as cathode materials but share a graphite anode.

A comparison of the results for the two different anodes reveals that TR occurs in both chemistries, though with significantly different temperature and pressure values. The maximum temperature for NCA is higher than for LTO (511 °C vs. 381 °C), and the increase in reactor internal pressure is 4 barg higher (5.071 barg vs. 1.802 barg). These values result in a significant weight loss in NCA compared to LTO, with a difference of 10 g, due to the ejection of solids and gases. Of the principal gases emitted, the most toxic are HF and CO, whose concentrations significantly exceed the IDLH limits. Furthermore, the quantity of residues ejected is also higher for NCA compared to LTO (4.010 g vs. 0.131 g), with notable differences in composition attributed to the initial chemical composition of the active materials and the organic fraction. In the case of LTO, the presence of traces of the plastic separator was identified in the condensed phase on the filter, accompanied by metallic particles derived from the active materials. In LTO, the presence of titanium (Ti) is associated with the composition of the anode.

Despite the differing cathode compositions, TR occurs in both cells (NCA and LPF) with comparable emitted products in terms of both type and total amount, with the exception of a minor quantity of liquid collected after the LFP thermal abuse test. The initial venting event occurred at temperature between 151 °C for NCA and 212 °C for LFP. The maximum temperature and pressure reached at the TR peak differed significantly, from 406 °C and 2.202 barg for LFP to 511 °C and 5.071 barg for NCA.

The data may be utilized to assess the most efficacious PPE [36]. Given the nature of the products emitted, which include gases and solid particulates, it is recommended that a gas mask equipped with a combination of gas filters be utilized. Gas filters serve to protect against a range of contaminants, including vapors, chemicals, and toxic gases. They are classified according to contaminant type, use conditions (single or multiple), filtering capacity, and efficiency. Each filter type is identified by a specific code and color, as reported in

Table 9. Filter color, type, filtered contaminant, and terms of use and filter capacity or efficiency [44].

Filter Color		Filter Type	Filtered Contaminant	Terms of Use Filter Capacity or Efficiency
Dark brown		AX	Gases and vapors organic compounds with boiling point < 65 °C such as methyl acetate, acetone, butane, chloroform, methanol, freons, etc.	Use immediately after opening, for single use only. Group 1: 100 ppm max 40 min, 500 ppm max 20 min. Group 2: 1000 ppm max 60 min, 5000 ppm max 20 min.
Light brown		A	Gases and vapor of organic compounds with boiling point < 65 °C, mainly solvents and hydrocarbons such as acetates, acetic acid, acrylics, alcohols, benzene, phenols, styrene, etc.	Class 1: 1000 ppm Class 2: 5000 ppm Class 3: 10,000 ppm With a ventilated system: Class 1: 500 ppm Class 2: 1000 ppm
Grey		B	Inorganic gases and vapor except CO: bromine, cyanide, chlorine, hydrogen sulfide, fluorine, isocyanates, formol, hydrocyanic acid…
Yellow		E	Acid gases and vapors: sulphuric anhydride, sulphur dioxide, hydrochloric acid, hydrofluoric acid, formic acid, etc.
Green		K	Ammonia and organic ammonia derivates: hydrazine, methylamine, aziridine, etc.
Black		CO	Carbon monoxide	Single use (max. 10,000 ppm)
Red		Hg	Mercury vapors	Maximum operating time: 50 h
Blue		NO	Nitrous vapors and nitrogen oxides	Maximum use time: 20 min single use
Orange		Reactor	Radioactive iodine, including radioactive methane iodine	Depending on the level of radioactivity
White		P	Particles	P1: filter efficiency > 80% P2: filter efficiency > 94% P3: filter efficiency > 99.95%

[44].

A potential combined filter may be A1E1P3 + CO, as indicated by the findings of the thermal abuse tests conducted on NCA, LTO, and LFP, as presented in Table 9.

5. Conclusions

In this study, three 18650 Li-ion cells with various chemistries, i.e., NCA, LTO, and LFP, were subjected to identical thermal abuse conditions (heating rate of 5 °C/min) to evaluate the impact of internal chemistry on thermal behavior. The cells were maintained at the same SoC of 100%. Comparisons were made between NCA and LTO, which have the same cathode but have different anode compositions, and between NCA and LFP, which have the same anode but different cathode compositions. The objective of these comparisons was to analyze the temperature during thermal runaway and the emitted products, including gas concentrations in relation to IDLH levels.

A comparison of anode composition (with the same cathode) revealed that LTO exhibited greater stability than graphite. The temperature and pressure reached during the TR were both lower in LTO cells than in NCA cells. The products emitted from LTO were predominantly liquid and condensate phases, whereas NCA emitted metallic powder. Regarding the cathode composition (with the same graphite-based anode), LFP exhibited greater stability than NCA. LFP experienced lower temperatures and pressures during TR, with emissions consisting of liquid and condensate phases, as opposed to the respirable metallic powder ejected by NCA. Regardless of the temperature and pressure reached during TR test, the primary products emitted were gases and vapors. The gases and vapors emitted are mainly composed of CO, HF, and DMC and EC, resulting from the evaporation of the electrolyte. HF and CO are particularly hazardous due to their toxicity; the NIOSH has set exposure limits of 30 ppm for HF and 1200 ppm for the CO. These limits were exceeded during all thermal abuse tests conducted on the three Li-ion cells.

In consideration of temperature and pressure, the safest cell can be identified as LTO, while the least safe cell can be defined as NCA, due to the lower temperature at which venting occurred.

A comprehensive understanding of the hazards associated with cell products led to the identification of the combined gas filter A1E1P3 + CO as the most efficient gas filter for use in gas masks, which is the primary PPE used by firefighters during extinguishing operations.