Study of the Role of LiNi_1/3Mn_1/3Co_1/3O₂/Graphite Li-Ion Pouch Cells Confinement, Electrolyte Composition and Separator Coating on Thermal Runaway and Off-Gas Toxicity

LiNi_1/3Mn_1/3Co_1/3O₂ (NMC 111)/graphite flat pouch prototype cells of a capacity of around 0.6 Ah and of dimensions 5 × 34 × 37 mm were assembled at CEA (Grenoble-France). The positive electrode represents about 31%, the negative electrode 20%, the current collectors 18%, the electrolyte 18%, the separator 4% and the remaining compounds as packaging materials and connectors, 9% of the prototype cells total mass (13.5 ± 0.3 g).

The cells were activated with different compositions of electrolytes. The reference electrolyte purchased from Solvionic (99.9% H₂O < 20ppm) is composed of 1 M LiPF₆ dissolved in a mixture EC:DMC:EMC (1:1:1 vol.). Electrolyte with vinylene carbonate (VC) as additive is based on reference electrolyte composition in which VC purchased from Sigma Aldrich (99%) was added as 2 wt% of the overall electrolyte composition. Electrolyte with a mixture of LiPF₆ and LiFSI salts is composed of 2/3 M LiPF₆ (Aldrich, battery grade ≥ 99.99%) and 1/3 M LiFSI (Suzhou Fluolyte Co., > 99.9%) dissolved in EC:DMC:EMC (1:1:1 vol.) with 2 wt% VC as additive.

The thickness of the reference PP/PE/PP tri-layer separator is 25 μm. The separator with ceramic coating is composed of a polyolefin PP/PE/PP tri-layer separator with a ceramic coating of 4.5 μm each side with a total thickness of 25 μm. Both types of separator have similar porosity (∼40%).

After cell formation at 45 °C with a charge at C/10 up to 4.15 V followed by a constant voltage step until C/20, prototype cells were discharged to 50% of state-of-charge (SOC) (at C/5) for the transportation and then charged at 100% SOC with a VMP system (Biologic, Claix, France), less than 12 h before conducting the thermal stability tests.

A specific testing device designed for cells of low capacity and small size was developed. This device (Fig. 1) is composed of a transparent cylinder in which the prototype is placed with thermal insulation. The controlled heating process is provided by a heating wire connected to a regulator (JUMO DICON). The prototype cells were submitted to a continuous ramp of 5 °C min⁻¹ up to 300 °C. All tests were performed under air. During the test, the cell voltage and the surface temperature, through thermocouples directly placed on the cell surface, were recorded. A video recording was performed in order to observe visually track cascading effects.

Zoom In Zoom Out Reset image size

Online gas sampling is carried out through a heated line (180 °C) positioned on the upper part of the device and connected to a Fourier-transform infra-red (FTIR) spectrometer (Thermo Scientific Nicolet Antaris IGS Analyzer, gas cell of 2 m). The flow rate of gas sampling during the experiment was 0.55 Nm³.h⁻¹. The online FTIR apparatus provides quantitative information regarding gases release from battery thermal runaway such as organic carbonates (EC, DMC, EMC, etc), hydrocarbons (CH₄, C₂H₄, etc), aldehydes (OCH₂, CH₃CHO, etc), carbon oxides (CO₂, CO), fluorinated species as HF and POF₃, and other species as HCN, NO_x and SO₂ responding in the infrared domain, according to adequate calibration processes. For pertinent exploitation of obtained FTIR spectra, characteristic wavenumber ranges for each component were selected with the aim of limiting as far as possible interferences. Further details regarding FTIR analysis conditions and relating calibration processes have been added in the supplementary information section.

Thermal stability tests of sample cells were performed in both loosely and tightly confined conditions, allowing (respectively, not allowing) sample pouch cell swelling.

For thermal stability tests of loosely confined cells, a purpose built cell holder was manufactured by INERIS (Fig. 2a) to allow a potential cell swelling in a free volume while ensuring a homogeneous heating. This holder is composed of two aluminum plates separated by wedges that maintain the prototype cell. The assembly is screwed, and the heater wire is wrapped around the aluminum plates. Thermocouples for heating regulation are placed between aluminum plates and the heater wire. Since cell opening was systematically observed on the connectors side during preliminary tests, the holder was mounted vertically inside the testing device so that gases release could most likely be directed to sampling area (upper part of the device).

Zoom In Zoom Out Reset image size

For thermal stability tests of tightly confined cells, aluminum plates positioned on each side of the cell are placed directly in contact with the prototype faces. The heater wire, wrapped around the aluminum plates, allows a homogeneous heating at the cell surface (Fig. 2b). It maintains a certain pressure to prevent the cell from swelling upon heating.

For each studied parameter (cell confinement, electrolyte composition, type of separator), each test was reproduced at least two times.

As used in a previous study,²⁶ the state-of-the-art fire-induced toxicity indices relating to given critical conditions, developed by ISO TC92 SC3 were used for the toxicity assessment. ISO 13571:2012 standard²⁷ is intended to address the consequences of human exposure to the life-threat components of fire and can be used for the estimation of the time at which individuals may be expected to experience compromised tenability. With care (in particular since some debate on the validity of the underpinning equations defining incapacitation has recently popped among ISO TC92 SC3 experts in the matter²⁸), this guidance can also be applied to estimation of the time limit for rescuing people who are immobile due to injury, medical condition, etc. If exposed individuals are able to perform cognitive and motor-skill functions at an acceptable level when exposed to a fire environment, the exposure is said to be tenable. If not, the exposure is said to result in compromised tenability.

Toxic-gas models of ISO 13571 are well suited when the time-dependent concentrations of fire effluents are known. For all thermal stability tests, these data were obtained thanks to the FTIR spectrometer. Concentrations of pollutants resulting from the experiments were converted into state-of-the-art fire-induced toxicity indices of the ISO 13571.

Because they are physiologically unrelated, and mechanistically independent, asphyxiant and irritant toxicants are treated separately, as referred to fire toxicity engineering state-of-the-art in the latest version of ISO 13571. Namely so-called fractional effective doses (X_FED) and fractional effective concentration (X_FEC) are computed for considering additive effects of mostly asphyxiant pollutants (e.g. CO, HCN...), respectively additive effects of essentially irritant fire gases (e.g. inorganic acids...). These models additionally account for a dose effect response on exposure to asphyxiants and a concentration effect response on exposure to irritant gases. X_FED and X_FEC can be obtained from the evolution of pollutant concentrations in a given enclosure using Eqs. 1 and 2, respectively:

$$\begin{eqnarray}&&{X}_{FED}=\displaystyle \sum _{{t}_{1}}^{{t}_{2}}\displaystyle \frac{\left[CO\right]}{35000}\,{\rm{\Delta }}t+\displaystyle \sum _{{t}_{1}}^{{t}_{2}}\displaystyle \frac{{\left[HCN\right]}^{2.36}}{1.2\times {10}^{6}}\,{\rm{\Delta }}t\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&\begin{array}{l}{X}_{FEC}=\displaystyle \frac{\left[HCl\right]}{{F}_{HCl}}+\displaystyle \frac{\left[HBr\right]}{{F}_{HBr}}+\displaystyle \frac{\left[HF\right]}{{F}_{HF}}+\displaystyle \frac{\left[S{O}_{2}\right]}{{F}_{S{O}_{2}}}+\displaystyle \frac{\left[N{O}_{2}\right]}{{F}_{N{O}_{2}}}\\ \,\,\,+\,\frac{\left[acrolein\right]}{{F}_{acrolein}}+\frac{\left[formaldehyde\right]}{{F}_{formaldehyde}}+\sum \frac{\left[{irritant}\right]}{{F}_{{C}_{i}}}\end{array}\end{eqnarray} \tag{ 2 }$$

where F_i is the critical concentration of each irritant gas that is expected to seriously compromise occupants' tenability.

No exposure limit was found for POF₃ but we may think that the toxicity of POF₃ might act through other poisoning mechanisms than HF by comparison with chlorine analog POCl₃/HCl and critical limits of exposure might be lower for POF₃ than for HF.²⁹ However, without consolidated exposure limit for POF₃, we considered as a reasonably conservative hypothesis, that the critical concentration of POF₃ was equivalent to that of HF (i.e. 500 ppm).

The terms containing [CO] and [HCN] in Eq. 1 at each time increment are to be multiplied by a frequency factor V_CO2 (Eq. 3) to account for the increased rate of asphyxiant uptake due to hyperventilation.

$$\begin{eqnarray}&&{V}_{CO2}=\exp \left(\displaystyle \frac{\left[ \% C{O}_{2}\right]}{5}\right)\end{eqnarray} \tag{ 3 }$$

As "tenability" may be defined according to different types of potential impacts to exposed people, critical values chosen here in Eqs. 1 and 2 refer to escape impairment that is supposed to be reached for X_FED or X_FEC equal to 1 for ordinary sensitive people.

Although ISO 13571 standard is the state-of-the-art to address the consequences of human exposure to the life incapacitation threat components of fire, few limits exist, such as the non-consideration of the effects of aerosols and particles and their interactions with gases emitted during fire, or the hypothesis that asphyxiant and irritant gases are acting separately whereas some interactions between the effects of different gases (synergy, antagonism), even below lethality threat are reported.^28,30 Another obvious limitation is that the results of the proposed modeling shall only be used in terms of comparisons of studied cells under same thermal abuse conditions since they are scenario-dependant.²⁶

Influence of confinement (allowing or not cell swelling) on thermal stability was studied on reference prototype cells (Fig. 3). Under loose confinement, prototype cell opens around 130 °C under the pressure of vaporized linear carbonate solvents DMC and EMC, then, at 180 °C, undergoes a self-heating process with a detected temperature rise up to around 420 °C. As DMC and EMC displayed the same profile, for clarity purpose, only DMC was plotted in figures of this paper. The heat generation was accompanied by the emission of opaque gases and the release of the cyclic carbonate solvent EC. This thermal runaway phenomenon was clearly attenuated when cells are tightly confined, providing by design, an external counter-pressure to thermally abused test cells. In these conditions, a much lower self-heating temperature peak is observed and this peak was detected 65 °C higher (i.e. at 245 °C) as compared to loosely confined cell case. As confinement, in this configuration, prevented the cell from swelling, its opening occurred at lower temperature, around 115 °C. When cell temperature reaches 240 °C, EC vaporized along with the low exothermic reaction observed but no opaque gas was emitted. The attenuation of the thermal runaway phenomenon and its detection at a higher temperature for test with tighting confinement is explained by the earlier volatile solvents release and the covering of electrodes surface by the melted separator. Thus, the electrolyte accessibility at the surface of the electrodes is limited: (i) gaseous PF₅ and volatile linear solvents do not come in contact with lithiated graphite material, consequently, first exothermic reactions initiated by SEI cracking and solvents reduction are prevented. (ii) Oxygen released from NMC material cannot oxidize vaporized EC. By contrast, without tighting confinement, the separator shrinkage followed by its melting makes the electrodes surface accessible to the electrolyte, leading to cascading reactions producing a thermal runaway.

Zoom In Zoom Out Reset image size

Gas analysis by FTIR showed that a lower quantity of gases was released during the test with tightly confined cell (figure 4b). CH₄ and OCH₂ were hardly detected. C₂H₄ coming from the reduction of EC^31,32 was detected from 165 °C and its concentration gradually increased until the detection of the low exothermic reaction around 245 °C, temperature at which other gases (CO, CO₂, HF and POF₃) were also detected. The very weak detection of gases coming from linear carbonates reduction (OCH₂ and CH₄) and the C₂H₄ concentration profile confirm the difficult accessibility of the electrolyte to the surface of the negative electrode. At 245 °C, gaseous EC reacts with the oxygen released by the cathode to form CO₂, CO and water, the latter promoting the formation of HF and POF₃ from PF₅. Without tighting confinement (Fig. 4a), larger amount of all the gases coming from electrolyte reduction, oxidation and reaction with water are detected from the beginning of the exotherm around 180 °C.

Zoom In Zoom Out Reset image size

Fire induced toxic-gas models of ISO 13571 applied on these experiments showed that, for asphyxiant gases, within the enclosure inside our testing device, the critical threshold value of X_FED (Fig. 3c) equal to 1 was never reached for both configurations. Of course, it shall not be taken for granted that none of those cells configurations would not lead to any toxic threat to people from asphyxiant gases, as this observation only deals with the test configuration enclosure, and subsequently do not directly reflect a plausible scenario of interest in field use of these cells in given application. More interestingly therefore is the fact that maximal X_FED value is significantly lower for test on tightly confined cells, as compared to loosely confined cell test configuration owing to the fact that the production of CO is significantly lower in this condition (Fig. 4). Regarding irritant gases, the critical threshold value of fractional effective concentration, X_FEC (Fig. 3a), is never reached for test on confined cells contrary to the test performed without tighting confinement, confirming a real safety advantage on pouch cell anti-swelling mounting arrangement. Whereas X_FEC profile in loosely confined case is driven by the production of fluorinated compounds (HF, POF₃) and formaldehyde (OCH₂), it is only governed by the release of fluorinated compounds (HF, POF₃) in the case of the test on tightly confined cells. These modeling results confirmed the positive effect of a cell mounting inducing tighting confinement as a result of module design on the thermal runaway and induced off-gas toxicity.

DSC measurements performed on graphite-based negative electrode film with different electrolyte compositions had showed, in a previous study,¹⁷ a synergistic effect of VC (2 wt%) addition and partial substitution of LiPF₆ (1/3 M) by LiFSI resulting in an improvement of the thermal behavior, i.e. a shift of the SEI related exothermic reactions to higher temperature (+50 °C) and a decrease of the heat energy release of more than 30%. Considering these encouraging results, shown at SEI level, thermal stability tests on loosely confined pouch prototype cells with same graphite electrode and electrolyte compositions were performed.

The reference and VC containing electrolyte cells opening (Fig. 5a), characterized by linear carbonates emission, occurred at the same temperature of 130 °C, whereas the thermal runaway phenomenon accompanied with gases (EC, CO, CO₂, CH₄, C₂H₄, OCH₂, HF and POF₃) release (Fig. S1 is available online at stacks.iop.org/JES/167/090513/mmedia) of the prototype cells activated with the VC containing electrolyte occurred at a higher temperature of 20 °C compared with the reference electrolyte prototype cell. The nature and the quantity of emitted gases were close for both types of cells (Fig. 6), indicating that reaction mechanisms are similar. Nevertheless, certain trends are observed for VC containing electrolyte prototype cells such as a relatively lower emission of EC and gaseous products coming from the reduction of the carbonates as well as a higher emission of CO₂. It is supposed that, upon heat-induced secondary SEI generation,¹⁷ the remaining VC additive reduces before solvents, lowering related gas emission. Additionally, the delayed thermal runaway gives EC time to react with oxygen released from NMC, leading to lower EC and higher CO₂ emissions. HF and POF₃ profiles are similar for both prototypes, indicating water presence related LiPF₆ salt degradation mechanisms are logically not affected by VC addition.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Regarding toxicity assessment, the critical threshold value of X_FED (Fig. 5c) obtained in our test conditions equal to 1 was never reached whatever the electrolyte composition, indicating the low contribution of the asphyxiant gas. CO was the only asphyxiant gas detected during the thermal stability test on these prototypes. Figure 5b shows that fractional effective concentration X_FEC rises over the critical threshold value when the thermal runaway occurs, after 111.2 min for prototype with reference electrolyte and after 115.6 min for prototype with VC electrolyte. These results are consistent with the thermal profiles. FEC profile is driven by the production of fluorinated compounds (HF, POF₃) and formaldehyde (OCH₂), whatever the electrolyte composition. The improvement of the thermal behavior in terms of onset thermal runaway temperature in presence of VC, detected from DSC measurements¹⁷ (+50 °C) is then confirmed, though to a lesser extent (+20 °C), at a larger scale, in thermal tests on 0.6 Ah pouch cells. The effect of the presence of an additive acting on SEI formation demonstrates the predominant role of the negative electrode/electrolyte interface in the early stages of the thermal runaway.

This 20 °C onset thermal runaway temperature shift detected in case of VC containing electrolyte prototype cells is maintained when LiPF₆ salt is partially substituted by LiFSI (Fig. 7a). In addition to gases coming from electrolyte solvents and LiPF₆ salt decomposition (Fig. S2), other gases such as SO₂, NO and HCN are likely to be emitted from LiFSI decomposition/combustion.²⁰ In the experiments of this study, SO₂ (Fig. 7b) was detected in a low amount (few milligrams) compared with LiFSI initial weight in the electrolyte (128 mg), whose total transformation (combustion and decomposition) would lead to 87 mg of SO₂. As already observed by DSC, the reduction of LiFSI into LiSO₂N(Li)SO₂Li salt and LiF³³ is favored over its combustion/decomposition at higher temperature, explaining this low proportion of SO₂. As shown in Fig. 7d, the critical threshold value of X_FED equal to 1, in this case again, is never reached whatever the electrolyte composition, indicating the low contribution of the asphyxiant gas in toxicity assessment. For 1 M LiPF₆-based reference electrolyte prototype, CO (Fig. S2) was the only asphyxiant gas that contributed to the FED indice whereas for LiPF₆/LiFSI-based electrolyte, X_FED profile was largely driven by the production of CO with a very slight contribution of HCN, since this latter gas was emitted in very low amount (<1 mg in total). X_FEC rises over the critical threshold value after the same time for both prototypes (Fig. 7c). For reference electrolyte prototype, X_FEC profile was driven by the production of fluorinated compounds (HF, POF₃) and formaldehyde (OCH₂). For LiPF₆/LiFSI-based electrolyte prototype cells, SO₂ and OCH₂ contributed to X_FEC profile. Fluorinated compounds also contributed, but, to a lesser extent, since the maximal concentration peak of these gases was around 2 to 2.5 times lower than those observed with reference electrolyte prototype cells. These results lead to the conclusion that replacing 1/3 M LiPF₆ by LiFSI does not significantly modify the thermally-induced toxicity threat from potentially released asphyxiant and irritant gases as a result of thermal runaway of concerned cells.

Zoom In Zoom Out Reset image size

Thermal stability tests on prototype cells containing VC electrolyte and polyolefin PP/PE/PP separator with or without ceramic coating were performed. Whichever the type of separator, a first sudden voltage drop (Fig. 8a) was observed at 180 °C due to successive melting of the polyolefin layers, and then a second voltage drop until 0 V occurred during the thermal runaway at 200 °C. The nature of emitted gases and the quantity of CO₂, CO, CH₄, C₂H₄ (Fig. S3), and carbonates (DMC, EMC, EC) released were similar for both types of cells. The total quantity of fluorinated compounds (POF₃ and HF) as well as of OCH₂ was found to be lower for ceramic-coated separator prototype cells (12, 11 and 14 mg vs 26, 20 and 22 mg respectively). This difference can be explained by the fact that fluorinated compounds react with the ceramic part of the separator through acid-base reactions. The maximal concentration peak of HF, POF₃ and OCH₂ was respectively 5.8, 3.7 and 3.9 times lower than those measured with prototype cells without ceramic coating. Although X_FEC value rises over the critical threshold value after the same time for both prototypes (Fig. 8b), it is noticeable, in the case of prototype cells with ceramic coating, that a significantly lower room flow rate would be required to bring X_FEC value inferior to one. X_FED equal to 1 was never reached in our test conditions whatever the separator composition, indicating again the low contribution of the asphyxiant gas (Fig. 8c) in toxicity assessment with CO as the only asphyxiant gas detected during the thermal stability test on these prototypes.

Zoom In Zoom Out Reset image size