The Effects of Cell Chemistry, State of Charge, and Abuse Method on Gas Generation in Li-ion Cell Failure

Abstract

We report on the effect state of charge (SoC), cell format, and chemistry have on the volume and composition (H₂, CO₂, CO, CH₄, C₂H₄, C₂H₆, C₃H₆, and C₃H₈) of cell failure gas from Li-ion cells. Nickel manganese cobalt oxide (NMC) 21700 cells with a 5 Ah capacity were externally heated to failure at a 5–100% SoC under an inert atmosphere. This showed that the volume of gas increased with cell SoC (1.8 L at 5% SoC vs. 8.3 L at 100% SoC). The effect of the cell chemistry format and abuse method was also investigated using 18650, pouch, and prismatic cells (2.3–50 Ah) with Ni-based or lithium cobalt oxide (LCO) cathodes or lithium titanium oxide (LTO) anodes. The results showed that at higher SoCs, larger quantities of gas were generated; however, there was no correlation between the cell SoC and the composition of gases produced. Tests on the other cells found that the Ni-based cell generated 1.29–1.89 L/Ah of gas. The main constituents of this were H₂, CO, and CO₂; however, all other hydrocarbons were identified in varying quantities. The LTO cells generated lower volumes of gas, 0.8 L/Ah compared to Ni-based cells, and the gas was found to contain lower H₂ concentrations but higher concentrations of CO₂. The LCO cell was found to generate a gas volume of 1.2 L/Ah. This forms the final of four papers which cover a total of 213 tests on 29 cell types with six different chemistries, all tested using a single robust testing method.

1. Introduction

The increasing popularity of Li-ion batteries for use in electric vehicles and battery energy storage, according to Chen et al., has the potential to lead to an accelerated number of battery failure events—a result of unwanted cell chemical reactions within the cell or exposure to conditions outside of the cell’s normal operating range (thermal, electrical, or mechanical) [1]. Exposure to a range of conditions can lead to the cell entering thermal runaway (TR), which leads to self-heating of the cell [2,3].

An increase in temperature can result in decomposition of the cell electrolyte. The decomposition products are often gaseous compositions and cause an increase in pressure within the cell casing, which eventually causes a failure event, often involving the release of smoke, flame, or sparks, which have the potential to create safety concerns.

Several factors can impact the volume and composition of gas generated during cell failure—including the cell chemistry, state of charge (SoC), and atmosphere in which the cell is failed. There are reports that give data for the volumes of gas generated for different format cells, tested under varying conditions [4,5]. To allow for comparison, gas volumes have been given as standard atmospheric temperature and pressure (SATP).

Studies into cells with nickel manganese cobalt oxide (NMC) chemistry, with various formats and capacities, have been outlined in the literature. Cells that failed via external heating under an air atmosphere at a 100% SoC have been found to be consistent, generating, on average, gas volumes in the range of 1.81 to 3.5 L/Ah [6,7,8]. Gas volumes given in L/Ah refer to the volume of gas generated per capacity (Ah) stored in the cell.

The testing of NMC 18650 cells, via the same method and within an inert atmosphere, by Golubkov and Zhang found that a similar volume of gas was generated (1.96 L/Ah and 1.87 L/Ah, respectively). These gas volumes were higher compared to those reported by Diaz in N₂ and air atmospheres, from 3 Ah NMC 18650 cells. Gas volumes under air generated 0.68 L/Ah and under N₂, they generated 0.66 L/Ah [6]. The opposite was observed by Howard, in 5 Ah NMC pouch cells, who reported an average volume of 1.3 L/Ah under N₂ compared to 0.9 L/Ah under air [9]. These differences were attributed to the O₂ availability for combustion in an air atmosphere.

The gas composition of failed NMC cells is typically found to contain H₂, CO, CO₂, and other small hydrocarbons, with any other trace gases deemed negligible [10]. Work by Laurelle identified the presence of CO, CO₂, C₂H₄, and CH₄ in gas samples of cells that failed under an air atmosphere and also identified that the confinement of the cell reduced the concentrations of these gases [11]. The quantification of gases by Shen and Koch found that most of the gas contained proportions of CO₂ (12–37%), CO (17–28%), and H₂ (13–27%), with small hydrocarbons also identified [4,8]. Similar gas compositions were observed by Zhang. Gas collected from cells under a N₂ atmosphere were composed of H₂ (33%), CO (30%), CO₂ (25%), and other hydrocarbons contributing >5% each [12].

Adding lithium manganese oxide (LMO) to the cathode of a NMC cell appears to affect the gas volume and composition produced during failure. Golubkov observed that NMC/LMO cells generated almost double the gas, 2.63 L/Ah, compared to NMC cells, 1.96 L/Ah, when failed under an air atmosphere [13]. The CO₂ and CO composition also differed; the NMC cells produced 24.9 mol% and 41.2 mol%, whereas the NMC/LMO cells generated 27.6 mol% and 13 mol%, respectively. There was little difference, however, in the concentration of other gases identified (H₂, CH₄, C₂H₄, C₂H₆). CO (1.2 mg/m³), CO₂ (0.9 mg/m³), and CH₄ (0.16 mg/m³) were also reported and quantified by Zhong under similar conditions [14]. Interestingly the gas volume reported by Sturk, from the failure of a 14 Ah five cell block of NMC/LMO cells, reported a total of 1500 L, (21 L/Ah) of gas under a N₂ atmosphere [10]. However, the temperature was not reported, which may explain the large volume of gas compared to similar tests.

Observations by Xu found that the gas volume decreased with the cell’s SoC, suggesting a correlation between these properties. The decrease in SoC also affected the gas composition, with an increase in CO₂ but a decrease in CO concentration. Interestingly, the H₂ concentration was highest at 50% SoC [15]. Slight variations in this general trend were observed by Willstrad, reporting that both CO₂ and CO concentrations increased with SoC [16].

The abuse method can also impact the volume of gas generated. Diaz reported that mechanically failed cells generated a smaller gas volume compared to externally heated cells, regardless of atmosphere [6]. Gas analysis comparisons on different abuse methods (external heat and nail penetration), tested under a N₂ atmosphere, found H₂, CO, and CO₂ to be the most abundant, with CH₄, C₂H₄, C₂H₆, C₃H₆, and C₃H₈ also detected [17]. Quantification of gases generated by a 156 Ah prismatic cell was as follows: CO (25–35 vol%), CO₂ (18–32 vol%), and H₂ (25–35 vol%). The pouch cell produced CO (22–24 vol%), CO₂ (29–39 vol%), and H₂ (20–30 vol%) [18]. The presence of CO and CO₂ was also observed, in air atmosphere, for 40 Ah prismatic cells failed via nail penetration [19].

Gas volume and composition data are more limited for cells with a lithium titanium oxide (LTO) anode. Yuan failed a 1.3 Ah LTO 18650, which generated 2.46 L/Ah, with the gaseous species quantified [7]. The following gas concentrations were given: CO₂ (37.6%), H₂ (8.41%), CO (5.3%), C₂H₄ (1.38%), CH₄ (1.23%), C₂H₆ (0.4%), and C₂H₂ (0.0008%). However, it is not clear what the rest of the gas composition contained.

There are numerous factors that can affect the gas volume and composition measurements, which include the abuse method used to fail the cell, the atmosphere in which cell failure took place, and cell chemistry or SoC. The reactions that generate the gas during failure events can be influenced by the surrounding environment. For example, testing cells under an air atmosphere can result in variation in gas volumes, due to the O₂ concentration available for combustion. Many studies referenced also rely on a single data point, which does not show variation in the gas volume and composition values.

We have previously reported numerous test campaigns that demonstrate the factors that can affect the gas volume generated during cell failure. The first paper indicated that the cell SoC can affect the volume of gas produced—cells at a 100% SoC generated more gas compared to those at a 25% SoC [20]. There was also a difference in the gas composition. Further testing of Ni-based 18650 and pouch cells indicated that the volume and composition of gas was affected by the atmosphere in which the cell was failed. The cells failed under N₂ or Ar often generated a larger gas volume than those failed under air. The gas was generally found to contain less H₂ and CO in air but more CO₂ [9,21,22]. Testing of nickel cobalt aluminium oxide (NCA) cylindrical cells with different SoCs and different cell capacities, given in Ah, found that as these two factors increase the volume of gas generated also increases, and that at increased SoCs, the amount of flammable gas constituents is increased [23].

Lithium iron phosphate (LFP) cells have been tested, including 26,650, pouch, and prismatic cells ranging from 3 to 230 Ah. These tests found that there is little difference in gas volume, when calculated as L/Ah, accounting for the cell capacity (Ah) at the time of failure. The gas composition was also similar between cell formats and the atmosphere in which the cell had failed. These LFP cells typically have a higher percentage of H₂ (40–50%) generated by LFP due to differences in failure mechanisms between LFP cells and Ni-based cells [24].

All our previously reported data provides a comprehensive dataset for the gas generation for cell failure, using the same test set-up, analysis methods, and calculations.

We have continued to expand the dataset through the results presented in the following paper and provide data points for less commonly reported cell chemistries, including lithium cobalt oxide (LCO) and cells with an LTO anode, as well as a broader dataset on a NMC cell. This paper is therefore the fourth in a series which now encompasses over 210 individual tests on 29 different cell types. All tests have been repeated and a variety of SoCs and abuse methods have been used. We believe that this is the largest publicly available collection of such data which uses a single methodology and equipment and therefore provides directly comparable results.

2. Materials and Methods

The test series aimed to understand two key influencing factors that affect the failure gas volume and composition, cell SoC, and failure method:

• External heating of a single cell type at SoC ranging from 5 to 100%.
• External heating and nail penetration of different capacity cells (2.3–50 Ah) with Ni-based, LCO cathodes, or LTO anodes.

2.1. Cell Types Tested

Testing was performed on eight commercially available cells with different capacities, formats, and chemistries. All cells were charged according to the manufacturer’s specification. Each cell was given a cell ID to identify it throughout testing. Prior to the experiment, where possible, the average cell weight, dimensions, and internal resistance (IR) were recorded (

Table 1. The details for the cells failed including the capacity, format, dimensions, chemistry, average weight, and internal resistance (IR). NMC (nickel manganese cobalt oxide), LMO (lithium manganese oxide), LTO (lithium titanium oxide), LCO (lithium cobalt oxide). (Note: where marked as ‘N/A’, data was not relevant and/or unable to be collected.)

Cell ID	Nominal Capacity (Ah)	Format	Dimensions (mm)	Chemistry	Average Weight (g)	IR (mΩ)
A	5	21700	21 (d) × 70	NMC-811	68	12.7 @ 30% SoC
B	6	Prismatic	120 × 75 × 10	NMC	240	0.787 @ 30% SoC
C	10	Pouch	100 × 60 × 10	NMC	128	32.47 @ 30% SoC
D	15	Pouch	N/A	NMC/LMO	381	N/A
E	23	Prismatic	115 × 105 × 21	LTO	552	N/A
F	30	Pouch	223 × 199 × 9.4	NMC	808	N/A
G	50	Prismatic	316 × 105 × 11	NMC/NCA	759	0.737 @ 30% SoC
H	50	Pouch	280 × 100 × 10	NMC	905	0.925 @ 30% SoC
I	2.3	18650	18 (d) × 65	LCO	46	N/A

).

2.2. Pressure Vessel Set-Up

All tests within the experimental series were performed inside a 47 L pressure vessel rated to 10 bar at 200 °C, contained within a blast chamber and remotely operated. Temperature data was collected using type-N thermocouples (accuracy ± 2.2 °C) placed above the cell, inside the pressure vessel (ambient temperature), and on the cell’s surface (cell temperature). Pressure data measurements were taken using a 0–2 bar pressure transducer for lower capacity cells and 0–7 bar for higher capacity cells. The pressure transducer was located on the back wall of the pressure vessel and data measurements were taken every 10 ms. Depending on the test requirements, the pressure vessel has either an air or an inert (N₂ or Ar) atmosphere. To achieve the inert atmosphere, the vessel was purged seven times, bringing the O₂ concentration to >0.05%. When required, a gas sample was collected, via a sampling vent to a 5 L Tedlar bag, once the vessel temperature had returned to ambient temperature.

2.3. Gas Volume Calculations

The volume of gas generated during cell failure was calculated using the ideal gas law. During the test, the ambient temperature and pressure were recorded and used in the gas volume calculations. By calculating and normalising the initial moles, the additional gas volume could be attributed to the cell failure. The number of moles was subsequently used to calculate the final gas volume. The values were standardised to achieve a gas volume at standard atmosphere temperature and pressure (298 K, 101,325 Pa), allowing for easier comparison between tests [9,20].

2.4. Gas Composition Analysis

The composition of gas collected after testing was analysed by mass spectrometry using a Hiden Analytical™ HPR-20 quadrupole mass spectrometer with a heated capillary tube (Hiden, Warrington, UK). Raw data was converted into % volume by subtracting the contribution of overlapping fragment ions and converting them using the instrument’s relative sensitivity factors for the target gas. A background subtract was applied to the data prior to conversion. Data was collected over a 0–150 m/z mass range and specifically quantified for the following target gases based on our literature review: H₂, CO₂, CO, C₂H₆, C₂H₄, C₃H₆, C₃H₈, and CH₄. If any other gases were present in the samples, it was assumed to be at very low concentrations and therefore negligible [10]. Only permanent gas was included in the final analysis results, which therefore excluded vapour concentration. Gas analysis was carried out on the same day as testing. Further details for the gas composition analysis calculations have previously been reported. Although we present gas composition data, we do not explore the mechanisms behind this formation, which has been previously reported and assessed to be due to the decomposition reactions of the solid electrolyte interphase (SEI) layer or electrolyte [25,26].

2.5. External Heat Test Set-Up (All Cells)

All cells were heated to failure by the attachment of an adhesive heater to the cell’s surface. The exact size and power supplied to the heaters can be found in the supplementary data (Table S4). All cells were tested under an inert (Ar or N₂) atmosphere at a 100% SoC. The cylindrical cells (Cell A and I) were held in place by a cell holder placed inside a baffle box to prevent flame impingement on the pressure vessel’s inner surface. The pouch and prismatic cells were placed between calcium silicate fireboard and stainless-steel plates, which were bolted together to prevent cell swelling. At least two tests were performed for each cell type, with gas samples collected from each of these tests.

Additional tests were performed on Cell A, under Ar at a 5, 25, 50, and 75% SoC and at a 100% SoC under an air atmosphere. It should be noted that the cells at each SoC were new and had only undergone 3 ‘wake-up’ cycles to get them to their required SoC; therefore, this cannot be correlated to cell state of health (SoH).

2.6. Nail Penetration Test Set-Up

Cell C, Cell D, and Cell F were also failed using nail penetration inside the pressure vessel with an inert atmosphere. Tests were performed using a custom-built rig, which used an actuator with a nail moving at 0.1 m/s to puncture the centre of the cell. The cell was held in place using steel cable ties and the nail entered the cell about half of the overall depth. The cell surface temperature was measured throughout the test using a type-K thermocouple, and once the cell had failed, gas samples were collected. Three tests were performed on Cell C and Cell F, and two on Cell D.

The test set-up for each cell type is shown in (Figure 1).

A summary of all the tests performed in this test series is given in

Table 2. A summary of the number of each type of test performed for this test series.

Cell ID	SoC (%)	Abuse Method	Atmosphere	Number of Tests
Cell A (5 Ah NMC 18650)	100	External Heat	Air	2
Cell A	100	External Heat	Inert *	6
Cell A	75	External Heat	Inert *	3
Cell A	50	External Heat	Inert *	3
Cell A	25	External Heat	Inert *	3
Cell A	5	External Heat	Inert *	3
Cell B (6 Ah NMC Prismatic)	100	External Heat	Inert *	3
Cell C (10 Ah NMC Pouch)	100	External Heat	Inert *	3
Cell C	100	Nail Penetration	Inert	3
Cell D (15 Ah NMC/LMO Pouch)	100	External Heat	Inert *	3
Cell D	100	Nail Penetration	Inert	2 +
Cell E (23 Ah LTO Prismatic)	100	External Heat	Inert *	3
Cell F (30 Ah NMC Pouch)	100	External Heat	Inert *	3
Cell F	100	Nail Penetration	Inert	3
Cell G (50 Ah NMC/NCA Prismatic)	100	External Heat	Inert *	2
Cell H (50 Ah NMC Pouch)	100	External Heat	Inert *	2
Cell I (2.3 Ah LCO 18650)	100	External Heat	Inert *	3

* An inert atmosphere refers to tests performed in either an Ar or N₂ atmosphere. ⁺ Both cells in these tests did not fail; therefore, gas volume and composition data are not provided.

.

3. Results and Discussion

3.1. 21700 NMC (Cell A) 100% External Heat

At a 100% SoC, Cell A (21700 NMC), cells tested under air generated half the gas, resulted in 4.2 L, compared to those under an inert atmosphere, 8.3 L. The results for Cell A have been previously reported but are included here for easier comparison [20]. This has been seen in previous tests on other Ni-based cells and is likely due to the combustion of H₂ resulting in the formation of water [9,21]. This is supported by the gas composition, which shows a H₂ concentration of 4.9% in air compared to 22.0% in inert conditions. The difference in O₂ availability also results in a much higher percentage volume of CO₂ in air due to combustion compared to inert atmospheres. The gas volume and composition for Cell A at 100% SoC is shown in (

Table 3. Average gas volume and composition data for Cell A (NMC 21700) at 100% SoC failed via external heat under air and inert atmosphere.

Atmosphere	Average Gas Volume (L)	Average Gas Volume (L/Ah)	Average Gas Composition
			H2 (%)	CO2 (%)	CO (%)	C2H6 (%)	C2H4 (%)	C3H8 (%)	C3H6 (%)	CH4 (%)
Air	4.2	0.83	4.9	67.5	24.7	0.2	0.3	0.4	0.2	1.9
Inert	8.3	1.67	22.0	27.8	37.8	0.9	3.9	2.3	0.5	4.9

).

3.2. 21700 NMC (Cell A) Other SoC External Heat

3.2.1. Temperature Data and Failure Characteristics (Cell A)

The failure characteristics for NMC 21700 (Cell A) showed no distinct pattern in the maximum temperature reached by the cell during failure, with maximum temperatures falling between 350 and 560 °C. The temperature reached by the cell did also not seem to be impacted by the amount of jelly roll ejected during failure. In general, the jelly roll was partially expelled from the cell, except for one test at a 25% SoC (A-15). The results for all Cell A tests are summarised in (

Table 4. Characteristics of Cell A failure including maximum temperature, if jelly roll was ejected, and mass loss for each state of charge (SoC) tested.

Cell ID	SoC (%)	Atmosphere	Maximum Temperature (°C)	Jelly Roll Ejected?	Mass Loss (%)
A-1	100	Air	492	No	68
A-2	100	Air	345	Partial	72
A-3	100	N2	413	No	37
A-4	100	N2	565	No	64
A-5	100	N2	414	Partial	71
A-6	100	Ar	710	Partial	65
A-7	100	Ar	552	Partial	69
A-8	100	Ar	332	N/A	N/A
A-9	75	Ar	568	Partial	72
A-10	75	Ar	572	No	59
A-11	75	N2	345	Partial	63
A-12	50	Ar	433	No	37
A-13	50	Ar	197	Partial	50
A-14	50	N2	416	No	37
A-15	25	Ar	264	Yes	81
A-16	25	Ar	297	N/A	72
A-17	25	Ar	353	Partial	71
A-18	5	Ar	332	No	10
A-19	5	Ar	385	No	12
A-20	5	N2	353	No	9

).

The different visual failures of the cell are summarised in Figure 2. Figure 2a shows the complete ejection of the cell jelly roll. The vent cap can also be seen next to the holder, which became detached from the cell due to the amount of pressure released during the failure event. The jelly roll remains partially contained and joined to the cell casing, as demonstrated in Figure 2b. Figure 2c shows a cell where no jelly roll was ejected, and metal deposits can also be seen around the vent cap. These events were all in cells of the same type, and the jelly roll’s position was unpredictable after cell failure.

3.2.2. Gas Volume and Composition (Cell A)

Gas volume appeared to be affected by the cell SoC, with the volume of gas generated increasing with the cell SoC. A seemingly linear trend was observed until the lowest SoC, 5%, was reached, which is consistent with the results observed by Xu. Analysis of the gas composition for the cells at different SoCs under inert conditions shows no consistent pattern for H₂, CO₂, CO, C₂H₆, and C₂H₄; however, there is an increase in the percentage volume of C₃H₈, C₃H₆, and CH₄ with decreasing SoC. The average values for gas volume and composition for these tests are found in Figure 3.

The gas produced during failure was collected for the cells at different SoCs and the composition analysed. The results showed no consistent pattern for H₂, CO₂, CO, or C₂H₄. However, there was an increase in percentage volume for C₃H₈, C₃H₆, and CH₄ as the SoC decreased. Figure 3 shows the average gas volume and compositions for the test series. A full breakdown of all tests performed on Cell A is provided in the supplementary data (Table S1).

The data presented in Figure 3 showed general consistency between cells of the same type and configuration, with the gas composition values falling within a 5% range. The Cell B dataset showed a higher variation between CO levels. Two of the tests gave 21.8 and 21.3%, but the third was lower at 14.5%. The supplementary data gives the full data for individual tests.

3.3. Ni-Based Pouch and Prismatic External Heat and Nail Penetration

3.3.1. Temperature Data and Failure Characteristics

A record of the maximum observed temperature during all external heat and nail penetration tests on Ni-based pouch and prismatic cells is summarised in

Table 5. Maximum observed temperatures reached across all external heat and nail penetration tests performed on Ni-based pouch and prismatic cells. (* Maximum temperature was much lower due to thermocouple becoming detached during cell failure.)

Cell ID	Abuse Method	Maximum Temperature (°C)	Cell ID	Abuse Method	Maximum Temperature (°C)
B-1	External Heat	323	G-2	External Heat	451
B-2	External Heat	351	H-1	External Heat	794
B-3	External Heat	406	H-2	External Heat	744
C-1	External Heat	754	C-4	Nail Penetration	736
C-2	External Heat	830	C-5	Nail Penetration	55 *
C-3	External Heat	908	C-6	Nail Penetration	584
D-1	External Heat	429	F-4	Nail Penetration	142
D-2	External Heat	379	F-5	Nail Penetration	261
F-1	External Heat	484	F-6	Nail Penetration	209
F-2	External Heat	486	D-3	Nail Penetration	14
F-3	External Heat	452	D-4	Nail Penetration	36
G-1	External Heat	593

.

When failed via external heat, the pouch cells generally reached higher temperatures compared to the prismatic cells. After failure, the prismatic cells remained intact, with only the vent cap appearing to burst on both Cell B and Cell G (Figure 4a,c). Despite Cell G being constrained, there was also slight swelling to the case (Figure 4a). The outer casing of all the pouch cells was damaged during the external heat tests, often having been melted, leaving the electrode foil exposed, which crumbled very easily when moved (Figure 4b).

Exposing cells to external heat provided an insight into the failure events in different cell types. The pouch cells appeared to generally reach a higher temperature than prismatic cells. The consequences of a failure event were also variable. The prismatic cells remained intact, which was demonstrated by Cell B and Cell G (Figure 4a,c). The vent caps burst, and Cell G had slight swelling to the casing, despite being constrained (Figure 4a). The pouch cells appeared to sustain more damage, with the outer casing melting to expose the electrode foil that crumbled when disturbed.

Utilising nail penetration demonstrated that pouch cells can undergo failure using this as a mechanism. Cell F’s outer foil showed signs of splitting along the side seam (Figure 4d); however, the smaller, lower capacity pouch cell (Cell C) had opened completely (Figure 4e).

Cell C and Cell F were tested using both external heat and nail penetration as failure mechanisms to compare the effects. The cells reached a higher surface temperature when subjected to nail penetration compared to external heat tests. Localised heating around the entry point, as the temperature measurements were taken from the outer foil layer, could be reasons for the observation. Interestingly, Cell D was also tested but did not appear to fail and only reached a maximum temperature of 36 °C across both tests.

3.3.2. Gas Volumes and Composition

Ni-based pouch (Cell C, Cell D, Cell F, and Cell H) and prismatic (Cell B and Cell G) cells that failed using an external heat source demonstrated consistent gas production, which ranged from 1.29 to 1.89 L/Ah (Figure 5). Analysis of the composition found that CO₂ (19.3–29.3%), CO (19.2–28.5%), and H₂ (23.2–33.3%) were the most abundant gases (Table 4). Small hydrocarbons were also identified within the mixture and typically contributed <5% of the total volume. C₂H₄ was detected at a higher level, particularly for Cell G, which could be attributed to variations in electrolyte mixtures between the cells.

Cell C and Cell F, both pouch cells with NMC chemistries, were subjected to external heat and nail penetration to observe the effects of these stimuli. These were the only cells to reach failure where gas samples could be collected. Cell C generated a large gas volume when heated, 1.3 L/Ah, compared to nail penetration at 1.09 L/Ah, which was consistent with the results observed by Diaz [6]. In contrast, Cell F generated a larger gas volume in nail penetration tests (3.08 L/Ah) than the external heat tests (1.88 L/Ah) (Figure 5). There was an observed variation between tests with different failure mechanisms, but the gas composition percentage remained constant (

Table 6. Gas composition for Ni-based pouch and prismatic cells. All pouch and prismatic cells were failed via external heat (EH), with additional nail penetration (NP) tests performed on ‘Cell C’ (10 Ah pouch) and ‘Cell F’ (30 Ah pouch) cells.

Cell ID	Abuse Method	Average Gas Composition
		H2 (%)	CO2 (%)	CO (%)	C2H6 (%)	C2H4 (%)	C3H8 (%)	C3H6 (%)	CH4 (%)
Cell B	External Heat	28.7	29.3	19.2	2.0	6.6	6.4	2.9	4.7
Cell C	External Heat	30.4	19.2	28.4	3.5	9.2	2.1	3.3	3.9
Cell D	External Heat	28.5	27.2	25.5	1.9	5.7	3.0	3.7	4.4
Cell F	External Heat	33.3	28.0	19.2	2.8	8.2	3.2	2.2	3.0
Cell G	External Heat	23.2	23.3	27.8	3.1	13.0	1.6	2.1	5.8
Cell H	External Heat	32.5	21.5	29.5	2.0	7.4	1.8	1.4	4.1
Cell C	Nail Penetration	30.3	24.2	16.9	4.1	11.8	3.4	5.5	3.8
Cell F	Nail Penetration	30.0	26.1	26.6	2.1	7.0	3.3	1.6	3.3

). All full breakdown of all external heat tests for Cells B–I are given in the supplementary data (Table S2).

3.4. Cell E (23 Ah LTO Prismatic) External Heat

Three external heat tests were performed on the 23 Ah LTO prismatic cell inside the pressure vessel.

3.4.1. Temperature Data and Failure Characteristics

When removed from the pressure vessel after a failure event, each cell tested appeared to have slight swelling of the outer casing and a burst vent, which was similar to what was seen with Cell G during the external heat tests; the maximum temperatures recorded were 365 (Cell E-1), 377 (Cell E-3), and 383 °C (Cell E-2). These temperatures are lower than those recorded for 50 Ah Ni-based pouch and prismatic cells (Cell G and Cell H). A comparison of the temperatures for each of the cells is shown in Figure 6.

3.4.2. Gas Volume and Compositon

Gas volume calculations for the LTO cells were shown to generate smaller volumes of gas, producing on average 0.8 L/Ah compared to the average 1.09 L/Ah, the lowest volume of gas produced for Ni-based cells. This is consistent with tests performed by Yuan [7]. They also had slightly different gas compositions. One observation was that less H₂ was produced, 12.9%, compared with 33.3%, the maximum concentration recorded in the Ni-based cells. More CO₂ was also observed 47.7% compared with a maximum of 29.3% in the Ni-based cells (Figure 7). According to Haung, the difference in the gas formation is understood to be due to the lithium intercalation differences between LTO anodes and NMC cathodes, which affects how this reacts with the electrolyte during thermal runaway [27]. The LTO cells appeared to generate smaller gas volumes (giving an average of 0.8 L/Ah), compared to the lowest value for Ni-based cells (1.09 L/Ah), a finding that was consistent with Yuan [7]. The gas compositions also varied between cell types, with the maximum percentage of H₂ (LTO: 12.9% vs. Ni-based: 33.3%) and CO₂ (LTO: 47.7% vs. Ni-based: 29.3%) being distinctly different (Figure 7).

The effect of the cell failure method on the compostion of gas has been discussed in more detail in our previously published paper, which found that OC made the biggest difference to gas composition [22]. Only two cells in this test series were tested using more than one failure method, Cell C and Cell F, which were subjected to external heat and nail penetration, both of which did not seem to affect gas composition. All individul test data for the nail penetration tests on Cell C, Cell F and Cell D is given in Table S3.

3.5. LCO 18650 (Cell I) Gas Volume and Composition

When failed via external heating, ‘Cell I’ generated an average gas volume of 2.8 L, with the main gas composition comprising CO₂, 38.6%, and H₂, 29.1% (

Table 7. Average gas volume and composition and temperature data for Cell I. Same data is also included for 3 Ah NMC 18650 cell taken from our previous publication [21].

Cell ID	Average Gas Volume (L)	Average Gas Volume (L/Ah)	Max Temperature Range (°C)	Average Gas Composition
				H2 (%)	CO2 (%)	CO (%)	C2H6 (%)	C2H4 (%)	C3H8 (%)	C3H6 (%)	CH4 (%)
Cell I	2.8	1.2	133–433	29.1	38.6	0.5	2.6	12.3	5.4	4.7	6.9
3 Ah NMC 18650	3.9	1.3	283–606	20.4	35.4	24.4	1.8	1.6	0.9	-	15.4

). Gas analysis and temperature data from our previously reported external heat tests on a 3 Ah NMC 18650 are included for comparison (note: these tests did not account for the presence of C₃H₆). The full test results for the NMC cell are presented in a paper by Abbott et al. [21].

Cell I, a LCO 18650 cell, was subjected to an external heat test and 2.8 L of gas was collected at the failure event. The main composition consisted of CO₂ (38.6%) and H₂ (29.1%), which is outlined in Table 7; the results for a similar 3 Ah NMC 18650 cell are included for comparison.

The NMC cell was tested using the same test method and generated an average gas volume of 1.3 L/Ah, which is like the LCO cell. The H₂ and CO₂ concentrations were also comparable; however, the CO proportion was significantly larger for the NMC cell (24.4%) compared to the LCO (0.54%). Generally, hydrocarbon ratios vary between the cell types, demonstrated by the higher percentage of CH₄ but lower C₂H₄ for NMC when assessed against the LCO values.

4. Conclusions

An experimental campaign was conducted to study the failure events when a series of Li-ion cells, with different properties and chemistries, were subjected to external heat or nail penetration. The data highlights the changes to the gas volume and composition when different failure mechanisms are used; however, we have no further evidence to report why this has the potential to affect failure mechanisms. Other results, within the published literature, corroborate the findings presented within this paper. A full dataset was compiled for all cell cathode and additional LTO anode chemistries that are commonly used in commercial applications. Completing the tests using the same methodology as we have previously reported allowed for a direct comparison between the 29 different cell types covered withing this and our previous papers [21,22,23,24]. These cover all currently available common Li-ion cell chemistries and formats and cells between 2 and 230 Ah. Over 210 individual tests are reported, each experiment has been repeated at least once, and for the most part, results from three tests are given. We believe this to be the largest and broadest such dataset within the open literature.

Within this paper, we show that Ni-based cells gave a gas volume of 1.3–1.8 L/Ah when exposed to external heat, which is higher than the tests on LCO cells (1.2 L/Ah) and LTO (0.8 L/Ah). Interestingly, the gas mixtures also varied between the cell chemistries. For example, the sample collected for LTO failure contained a lower concentration of H₂ compared to similar Ni-based cells.

The method impacts the overall volume of gas generated when a cell fails, which is highlighted by the lower values for nail penetration tests when compared to external heat. The ratio of gas remained proportional in both cases.

The SoC for a cell at the time of failure also impacts the gas volume, with both showing a linear decrease. However, clear correlation was not observed for SoC and gas composition.

Overall, it is important to understand that many factors can influence the volume and composition of gas generated when different Li-ion cells fail; therefore, it should not be assumed that a single cell failure will give a result that is completely representative of all cells.