Experimental and Numerical Study of the Impact of Pressure During the Pyrolysis of Diethyl Carbonate and Ethyl Methyl Carbonate

Abstract

During a thermal runaway, Lithium-ion battery cells are subjected to a large increase in temperature, which will vaporize and potentially thermally degrade their liquid electrolyte. The formation of gas in the battery cell will increase the pressure until the flammable gases vent and potentially lead to a fire incident. While the pyrolysis chemistry of the electrolyte components has been studied near atmospheric pressure, the effect of pressure has not been investigated. This study was undertaken to better understand the effect of pressure on the thermal dissociation of two common linear electrolyte components, diethyl carbonate (DEC) and ethyl methyl carbonate (EMC). The pyrolysis of DEC and EMC was studied in the gas phase, in 99.75% He/Ar, and was carried out at high temperatures and for pressures near 5.5 atm. The time-resolved CO formation was measured using a quantum cascade laser, providing a unique experimental dataset. A detailed chemical kinetics analysis was performed to understand the effect of pressure on DEC and EMC, with CO time-history results obtained in similar conditions at near-atmospheric pressure for DEC and EMC serving as baselines for comparison. Numerical predictions using detailed chemical kinetics mechanisms from the literature were carried out, and reaction pathways at different pressures were highlighted to emphasize the effect of pressure on the pyrolysis chemistry.

1. Introduction

The electrification of vehicles has been largely increasing globally and is recognized as an incredibly popular, viable option in some countries [1]. This growth in electric vehicle sales was partly due to strong legislative pushes motivated by mitigating air pollution associated with internal combustion engines and also as a key aspect of sustainable transport, with transformative technical developments leading to an ease for its adoption and its high competitiveness in the automotive industry [2]. Additionally, smartphone technology development has led to a phenomenal increase in its use worldwide, triggering the need for efficient and durable Li-ion batteries (LIBs) that are free of significant safety hazards. These fire hazards are due to thermal runaways generally induced by an abuse of the LIB. During a thermal runaway event, the temperature increases rapidly, and the liquid electrolytes in the battery cell—which allow the movement of the Li-salt ions during charge and discharge cycles [3]—will vaporize and undergo thermal breakdown. In some cases, if sufficient pressure is generated in the cell, the flammable thermal runaway gases will vent to the surrounding environment [4] and ignite, leading to a fire that is difficult to control.

In essence, the fire hazard associated with LIBs is due to the flammability of the electrolyte [5] and its vent gases [6]. Linear carbonates such as diethyl carbonate (DEC) and ethyl methyl carbonate (EMC) are among the most common constituents of the liquid electrolytes in LIBs. Understanding the pyrolysis kinetics of these components at higher pressure to observe the effect on the type of flammable gases generated during a thermal runaway is of interest to developing strategies to minimize the risks of LIB thermal runaways and ultimately fire events that are difficult to extinguish [7,8]. The chemical structures of DEC and EMC are presented in Figure 1.

This paper is a continuation of a pyrolysis study of DEC and EMC [9] by the present authors, where CO time-history profiles were obtained behind reflected shock waves for temperatures ranging from 1300 to 1800 K, and at near 1.3 atm. The impact of pressure on DEC and EMC is investigated herein at similar conditions found in [9], with pressure increased by approximately a factor of four-to-five, from 1.3 atm to approximately 5.5 atm, to represent the higher-pressure conditions found in the battery cell during a thermal runaway event and before the venting of the gases. In previous work on DEC [10,11,12,13,14,15], only a couple of studies focused on varying the pressure to highlight its effects. Shahla et al. [11] observed the impact of pressure in oxidative conditions by measuring laminar flame speeds (LFS) of DEC in air at 393 K, from 1 atm up to 3 atm. It can be seen that the increase in pressure decreases the LFS for equivalence ratios between 0.7 and 1.3. Sun et al. [13] measured species mole fractions by gas chromatography (GC) using a flow reactor at three different pressures (0.04, 0.2, and 1.03 atm) at 700–1200 K, and under pyrolysis conditions. The formation of the intermediate species C₂H₄, C₂H₅OH, and final product CO₂ accelerates when the pressure is augmented (i.e., 0.2 and 1.03 atm).

Previous work on EMC [14,15,16,17,18,19,20] shows that Yu et al. [16] carried out rapid compression machine (RCM) testing at 810–980 K, equivalence ratios (ϕ) of 0.5, 1.0, and 2.0, for two different pressures, namely 19.7 and 39.5 atm, where a drastic reduction in their measured IDTs was exhibited at 39.5 atm when compared to 19.7 atm. Feng et al. [17], Zhang et al. [18,19], and Luo et al. [20] focused on LFS for temperatures ranging from 373 to 543 K, ϕ = 0.6–1.6, where the effect of pressure was tested between 0.7 and 3 atm. As expected, the LFS decreased as the pressure was increased, with the same conclusions found in the work from Shahla et al. [11] for DEC. Conversely, for another linear carbonate—the dimethyl carbonate (DMC)—the effect of pressure was studied more extensively, e.g., Hu et al. [21] with ignition delay times (IDTs) in a shock tube (ST) at pressures ranging from 1 to 10 atm (1100–1600 K, ϕ = 0.5–2.0); Alexandrino et al. [22] with IDTs in a ST and a RCM (791–974 K, ϕ = 0.5–1.0 and 954–1585 K, ϕ = 0.5–2.0, respectively) from 2 to 40 atm; and Sun et al. [23] using a burner (quantification of CO, CO₂, CH₄, C₂H₂, C₂H₄ to cite a few). These studies of oxidation at high pressures were performed because linear carbonates are also considered as biofuels for internal combustion engines [10,24]. To summarize, except in the study of Sun et al. [13] with DEC, the effect of pressure was not measured under the pyrolysis conditions that are encountered in battery cells. Moreover, the investigation of the pressure effect was conducted for sub-atmospheric pressures that are essentially not realistic for common applications in the Sun et al. [13] study (0.04 and 0.20 atm).

The aim of the present study was to provide new experimental results for DEC and EMC pyrolysis at higher pressures to complement the current database. The effect of pressure was investigated by comparisons with the results from an earlier effort in [9]. This paper is organized as follows: the experimental methods and conditions are first introduced. Then, the new pyrolysis measurements are presented, and finally, chemical kinetics analyses are shown. This paper compares the reactivity of DEC and EMC pyrolysis at high pressure, utilizing CO time-history profiles for the first time.

2. Method

2.1. Shock-Tube Facility

Experiments were carried out in a stainless-steel shock tube, for which only a brief description is provided here, as further details are available in the literature [25]. The shock tube consists of a 6.78-m long driven section and a 3.0-m long driver section, with inner diameters of 16.2 cm and 7.62 cm, respectively. The two sections were separated by a 0.508 mm-thick polycarbonate diaphragm. Prior to each experiment, the driven section was evacuated to a pressure of approximately 10⁻⁸ atm or lower using a combination of a mechanical pump and a turbomolecular pump. Five piezoelectric pressure transducers, positioned along the final 2 m of the driven section, were employed to monitor the propagation of the incident shock wave and extrapolate its velocity to the endwall. This velocity, along with the initial gas temperature and pressure in the driven section, was used to calculate the temperature (T₅) and pressure (P₅) behind the reflected shock wave. The estimated uncertainties for T₅ and P₅ are 0.8% and 1.0%, respectively [26].

Mixtures were prepared in a separate stainless-steel tank using the partial pressure method. The DEC and EMC were introduced in vials and degassed several times to ensure no air was introduced into the mixture. Then, the DEC or EMC in the vapor phase was introduced into the mixing tank well below its maximum vapor pressure (9 Torr at 23.8 °C and 45 Torr at 25 °C for DEC and EMC, respectively) to guarantee that no condensation occurs during the mixture preparation. The mixtures investigated are presented in

Table 1. Mixture compositions and experimental conditions covered during this study.

${\mathit{X}}_{\mathit{D}\mathit{E}\mathit{C}}$	${\mathit{X}}_{\mathit{E}\mathit{M}\mathit{C}}$	${\mathit{X}}_{\mathit{H}\mathit{e}}$	${\mathit{X}}_{\mathit{A}\mathit{r}}$	T5 (K)	P5 (atm)
0.0025	0	0.2	0.7975	1219–1459	5.06–6.06
0	0.0025	0.2	0.7975	1326–1656	5.14–6.02

. The experimental conditions correspond to temperatures ranging from 1218 to 1656 K and pressures from 5.06 to 6.06 atm. Due to the aforementioned limitations with the vapor maximum pressures of the components considered herein (e.g., 10 Torr of EMC was used to prepare a 524 kPa mixture highly diluted in He/Ar, which allowed for three to four experiments), multiple mixtures were produced to investigate the ranges of temperatures, ensuring reproducibility and reliability. Mixtures were highly diluted in 0.2 He/0.7975 Ar to minimize the temperature change due to chemical reactions. Since the main component of the mixtures in the driven section is argon, the non-ideal boundary-layer effects (i.e., dP/dt) are minimal [27]. The DEC and EMC were purchased from Sigma Aldrich with a purity of 99%, and the gases, He and Ar, were provided by Praxair (Bryan, TX, USA), both with 99.999% purity.

2.2. CO Laser Absorption Diagnostic

The CO time-history profiles were measured using a continuous quantum cascade laser (Alpes Lasers) producing light near 4.8 μm to monitor the P(20) transition line of the 1 $\leftarrow$ 0 fundamental band of CO. The maximum absorption strength of the P(20) line is ensured by centering the laser wavelength using a removable cell containing a mixture of 10% CO in Ar. This quantification technique works in relative isolation from H₂O and CO₂ absorption [28], and the uncertainty in the measurements is estimated to be approximately 5.5% [29]. The laser light is split into two beams, and their intensities are collected using infrared, cryogenically cooled, photovoltaic detectors. The time-resolved incident intensity ($I_0$) is directed to the first detector as a referenced intensity, while the time-resolved transmitted intensity ($I_t$) is going through the sapphire windows of the shock tube and the reacting gases to terminate in the second detector and follow CO absorption signals during the experiment. Before the beams pass bypass filters, irises, and lenses (a direct-absorption setup) and reach the detectors, careful attention to the dark current phenomenon from the detectors themselves was made, and the associated offsets were collected to correct the CO profiles during the post-processing [30]. Time histories of CO mole fractions are processed with these two intensities based on the Beer-Lambert relation, defined as

$$I_t/I_0=exp\left(-S\left(T\right)\mathsf{\varphi }\left(\nu \right)PL{X}_{CO}\right)$$

(1)

where $S\left(T\right)$ is the linestrength, $\varphi \left(\nu \right)$ is the lineshape function, $P$ is the pressure, $L$ is the path length, and $X_{CO}$ is the CO mole fraction. The linestrengths were taken from the HITEMP database [31], while the lineshape function was modeled using a Voigt profile per the method of Liu et al. [32]. The collisional broadening coefficients of the bath-gas components He and Ar were taken from Grégoire et al. [29], and 20% of the diluent is helium to accelerate the vibrational relaxation of CO. A representative example of a CO time-history profile for the pyrolysis of EMC in 99.25% He/Ar is presented in Figure 2, and the uncertainties are shown in the mole fraction levels.

3. Results and Model Comparison

3.1. DEC Pyrolysis

New experimental measurements were carried out in a shock tube to obtain CO time-history profiles for DEC pyrolysis in 99.75% He/Ar near 5.5 atm, see Figure 3. The spikes observed before and at 0 µs are due to temporary laser beam steering, which results from the density change caused by the passage of the incident and reflected shock waves. Laser absorption measurements in a shock tube at elevated pressures are more challenging than at pressures near 1 atm, and care was taken to best align the optical components to minimize any negative impacts on the S/N of the measured beam intensity due to facility vibrations and increased turbulence-induced fluctuations within the beam path, as in Mulvihill [33]. As can be seen from these profiles, the CO mole fractions increase across all temperature cases studied. At lower temperatures (1219 and 1335 K), the CO profile increases relatively slowly with time, with a steeper gradient for the intermediate temperature of 1354 K. At higher temperatures (1386–1459 K), the CO growth rate is more noticeable with a rapid increase in the CO mole fractions from the start of the experiment. The CO time-history profiles at higher temperatures show significant formation of CO from the pyrolysis of DEC, followed by a plateau at approximately 250 and 1200 µs at 1459 and 1386 K, respectively. The DEC decomposes via a pyrolysis process where approximately 20% of the O-atoms introduced in the fuel are converted to CO in our conditions, at the plateau. Additionally, the CO mole fraction reaches a plateau at the highest temperatures due to the absence of oxygen, which would allow the CO-to-CO₂ production reaction mechanism [34,35].

These new experimental results were compared to detailed chemical kinetics models from the literature, namely Nakamura et al. [10], Shahla et al. [11], Sun et al. [13], Sela et al. [12], Takahashi et al. [15], Grégoire et al. [9], Feng et al. [17], Yu et al. [16], Khan-Ghauri et al. [36], and Luo et al. [20]. Three representative CO time-history profiles at high (a), intermediate (b), and low (c) temperatures are exhibited in Figure 4. At 1459 K and 5.06 atm (Figure 4a), all models are under-reactive: when comparing the time at which 50% of the experimental CO formation occurs, at approximately 40 µs, the numerical simulations predict this parameter at approximately 150 µs for most of the models, with the Shahla et al. [11] mechanism (solid orange line) being slightly slower with a 50% CO formation reached near 190 µs. Significant discrepancies are also observed at the plateau, where all models are 8–20% above the measured 1470 ppm, except that the Grégoire et al. [9] (dashed magenta line) and Luo et al. [20] (solid red line) models are underpredicting by 4% the final CO mole fraction, which is within the experimental error. Overall, at this high temperature, the Luo et al. model appears to be the most accurate both in rate of CO formation and CO level at the plateau.

At 1386 K and 5.38 atm, Figure 4b, a similar under-reactivity is seen for the models, i.e., at 50% of the CO production (720 ppm at 270 µs experimentally), all models predict 720 ppm of CO at a slower time near 490 µs, with Shahla et al. (solid orange line) the slowest model at 600 µs and Luo et al. [20] (solid red line) the fastest model observed at 400 µs. Note that no model reaches the plateau within the conditions tested; however, the Grégoire et al. [9] (dashed magenta line) and Luo et al. [20] (red solid line) models fall behind with a 170 ppm difference at 2 ms. Overall, the shape is better predicted by the Luo et al. [20] model under this condition. Conversely, the lower-temperature case at 1335 K and 6.06 atm shows that models are over-reactive compared to the experimental profile, except the Shahla et al. [11] mechanism, which predicts the CO formation with high accuracy.

3.2. EMC Pyrolysis

Figure 5 presents the new CO time-history profiles for the pyrolysis of EMC in 99.75% He/Ar near 5.5 atm at temperatures ranging from 1326 to 1656 K, where similar observations can be made on the general feature of the profiles (referring to Figure 3). However, the CO concentrations reached at the plateau are much more important, i.e., near 2250 ppm for the high-temperature conditions (1541–1656 K), which corresponds to a conversion of 30% of the O-atoms into CO at the plateau. The work from Grégoire et al. [9] compared the DEC and EMC with another linear LIB electrolyte, dimethyl carbonate (DMC). That work also exhibited a large effect from the methoxy groups present in the DMC molecule. In summary, DMC, DEC, and EMC were studied in 99.75% He/Ar near 1.3 atm at temperatures from 1230 to 2375 K. On the one hand, the ppm levels of CO at the plateau were as follows (from the highest to the lowest at 1660 K, 1.2 atm in [9]): DMC produced a higher CO mole fraction, followed by EMC, and then DEC, with a step decrease in the CO level by roughly 700 ppm with each methoxy group replaced by an ethoxy one. On the other hand, DMC and DEC were more reactive than EMC at near-atmospheric pressure.

A comparison of DEC and EMC is provided in Figure 6 at 1464 ± 5 K, and near 5.5 atm. The higher-pressure conditions explored herein do not change the main characteristics of DEC and EMC pyrolysis observed in Grégoire et al. [9]: (1) the CO concentrations obtained at the end of the test (at roughly 2.4 ms) are still higher for EMC than DEC, and (2) DEC is noticeably more reactive than EMC. Note that the current time histories in Figure 6 were also recorded in a highly diluted mixture, i.e., 99.75% He/Ar, matching the Grégoire et al. [9] dilution parameter, so further discussion on the effect of pressure can be found below. While the CO experimental profile of EMC pyrolysis shows a slower increase, it ultimately crossed with the result from DEC at roughly 1480 µs at this temperature.

The EMC experimental profiles are compared to the detailed chemical kinetics mechanisms from the literature in Figure 7. Six models are available, namely Takahashi et al. [15], Grégoire et al. [9], Feng et al. [17], Yu et al. [16], Khan-Ghauri et al. [36], and Luo et al. [20]. Their numerical predictions are shown for three representative results at (a) high (1602 K), (b) intermediate (1541 K), and (c) low (1469 K) temperatures. As one can see, the high-temperature case (Figure 7a) is well predicted by all models, with results from Takahashi et al. [15] and Feng et al. [17] slightly over-reactive in the first 400 µs. Note that the Grégoire et al. [9] and Khan-Ghauri et al. [36] models produce equivalent CO time histories, as updates in the most recent mechanism (Khan-Ghauri et al. [36]) focused on the electrolytes’ oxidation reactions as well as the unique chemistry of a fire suppressant candidate (bis(2,2,2-trifluoroethyl) carbonate) for LIB fire hazard [37]. Moreover, Grégoire et al. [9] indicated that the major changes were mostly implemented for DMC and DEC; however, sensitivity and rate-of-production analyses demonstrated that the model’s performance for EMC was suffering from the methanol (CH₃OH) sub-mechanisms in our conditions, an issue that was not visible with other validation targets data like ignition delay time or laminar flame speed. The importance of the CH₃OH sub-mechanism at high pressure is investigated in the next section.

Moderate agreement is visible at 1541 K and 5.44 atm in Figure 7b for the prediction of the plateau, but the onset of CO is not well-reproduced by any model. The largest discrepancies in the CO levels were denoted at 500 µs for the Takahashi et al. [15] and Feng et al. [17] models that differ by 51% from the experimental measurement, closely followed by the remaining models, which deviate by a maximum of 23% near 330 µs. Lastly, Figure 7c used a CO profile obtained at a lower temperature, 1469 K and 6.02 atm, and the models have difficulty matching the slow CO growth, as the curvature of all models exhibits a much steeper rise at the beginning of the test time, and therefore an early transition to the plateau. This behavior indicates that the EMC models require further improvements for the lower temperatures at pressures near 6 atm.

3.3. Effect of Pressure

The effect of pressure can be observed using the new CO time-history profiles presented herein and the CO time-history results obtained in similar conditions at near-atmospheric pressure from the Grégoire et al. [9] study. Figure 8 shows two cases where similar temperatures were recorded, namely 1381 ± 5 K (Figure 8a) and 1487 ± 4 K (Figure 8b), with the two pressures of roughly 1.3 and 5.5 atm. One can see that the overall effect of a higher pressure is the augmentation in the reactivity for both electrolytes, leading to larger concentrations of CO in the time frame of our experiments. The effect of an increase in the pressure also leads to an increase in the amount of CO produced at the plateau (see numerical predictions in the Supplementary Materials). Along with the experimental profiles are the numerical predictions from the Luo et al. [20] mechanism, which performs relatively well at intermediate and low temperatures for both electrolytes considered in this work. Unsurprisingly, the reaction mechanism is able to predict the DEC and EMC pyrolysis behavior near 1.3 atm, as most of the kinetic efforts in the literature were conducted for near-atmospheric conditions, even if the comparison with EMC shows a 14% difference in the CO ppm levels between the experimental measurement and modeled profile.

At 5.6 atm, the model is even more over-reactive for EMC, whereas the model is now under-reactive for DEC. On the one hand, the lack of performance regarding DEC at 5.6 atm could be due to pressure-dependent reactions that are not necessarily having a large contribution at 1.3 atm. Moreover, EMC seems to carry over a set of reactions that are neither properly tuned at 1.3 atm nor at 5.6 atm. The increase in the reactivity could also indicate the onset of other reactions at the early stage of EMC thermal decomposition. Overall, the detailed chemical kinetics mechanism from Luo et al. [20] exhibits satisfying performance and is to be used for further analyses. More comparisons (1.3 vs. 5.5 atm) with DEC and EMC pyrolysis CO time-history profiles are available in the Supplementary Materials at different temperatures.

4. Discussion

Understanding the complexity of the pyrolysis of these electrolytes is crucial to allow better prediction and control of the processes at higher pressures, typically generated within the LIBs during thermal runaway. Sensitivity and rate-of-production analyses were carried out for DEC and EMC using the Luo et al. [20] mechanism for two pressures, 1.3 and 5.5 atm, at 1380 and 1490 K, respectively. As can be expected and seen in our former study at 1.3 atm [9], the chemistry involved for the DEC and EMC electrolytes’ pyrolysis is drastically different.

The sensitivity analysis for CO (see Figure 9a) reveals that the most dominant reactions remain the same at higher pressure (5.5 atm) as the ones discussed in Grégoire et al. [9] at 1.3 atm, with the reactions DEC $\rightleftarrows$ CCOC*OOH + C₂H₄ (R1) and the subsequent reaction CCOC*OOH $\to$ C₂H₅OH + CO₂ (R2) giving ethylene (C₂H₄) and ethanol (C₂H₅OH). The reaction (R2) is highly inhibiting the production of CO because it produces CO₂ instead, which prevents an important O-atom conversion into CO.

DEC ⇄ CCOC*OOH + C₂H₄

(R1)

CCOC*OOH → C₂H₅OH + CO₂

(R2)

Nevertheless, the set of reactions for C₂H₅OH at near-atmospheric pressure is primarily led by C₂H₄OH radicals obtained via the reactions C₂H₅OH + H $\rightleftarrows$ SC₂H₄OH + H₂ (R3) and C₂H₅OH + H $\rightleftarrows$ PC₂H₄OH + H₂ (R4)—producing CO with the sequence C₂H₄OH $\to$ CH₃CHO $\to$ CH₃CO $\to$ CO—whereas a higher sensitivity from the reaction DEC $\rightleftarrows$ CCOC*OOCj + CH₃ (R5), written in reverse in the model), at higher pressure is observed. It contrasts starkly with the inhibiting reaction (R2) at 1.3 atm.

C₂H₅OH + H ⇄ SC₂H₄OH + H₂

(R3)

C₂H₅OH + H ⇄ PC₂H₄OH + H₂

(R4)

DEC ⇄ CCOC*OOCj + CH₃

(R5)

As shown in Figure 10, the routes for CO production at higher pressure are first CH₃CO(+M) $\rightleftarrows$ CH₃ + CO(+M) (R6), closely followed by HCO + M $\rightleftarrows$ H + CO + M (R7). At 5.5 atm, the reaction CH₃CHO + H $\rightleftarrows$ CH₃CO + H₂ (R8) has a larger contribution compared to the results at 1.3 atm (see Figure 9b), and direct CO formation from CH₃CO is preferred at 5.5 atm over the reaction (R7) at 1.3 atm.

CH₃CO(+M) ⇄ CH₃ + CO(+M)

(R6)

HCO + M ⇄ H + CO + M

(R7)

CH₃CHO + H ⇄ CH₃CO + H₂

(R8)

The CH₃CHO comes from the intermediate species mentioned above, CCOC*OOCj, from the reaction (R9), such as:

CCOC*OOCj ⇄ CH₃CHO + CH₃OCO

(R9)

The acetaldehyde reaction pathway leading to CO can be described with the reactions (R8)–(R6). Conversely, the reaction CH₂O + H $\rightleftarrows$ HCO + H₂ (R10) contributes noticeably at 1.3 atm, see Figure 9b, from the reaction pathway of C₂H₅OH $\to$ CH₂OH $\to$ CH₂O $\to$ HCO $\to$ CO, following the set of reactions (R1)–(R2)–(R11)–(R12)–(R10)–(R7). While the reaction (R2) is still active at higher pressure, its inhibiting effect is shortened in duration, as visible in Figure 9a.

CH₂O + H ⇄ HCO + H₂

(R10)

C₂H₅OH ⇄ CH₃ + CH₂OH

(R11)

CH₂OH(+M) ⇄ CH₂O + H(+M)

(R12)

The crucial effect of increased pressure lies in the production of CCOC*OOCj via the reaction (R5). When the same sensitivity and rate-of-production analyses are performed with a slightly less performant model—in this case the Khan-Ghauri et al. [36] mechanism—the same core of reactions is involved; however, their sensitive distribution differs considerably. Firstly, the reaction DEC $\rightleftarrows$ CCOC*OOCj + CH₃ (R5), adding CO via CH₃CHO and subsequent reactions (R9)–(R8)–(R6), is remarkably less sensitive in the Khan-Ghauri et al. model. Secondly, the reactions DEC $\rightleftarrows$ CCOC*OOH + C₂H₄ (R1) and CCOC*OOH $\rightleftarrows$ CjOC*OOH + CH₃ (R13) are more sensitive, which induces a competition with CjOC*OOH $\rightleftarrows$ CH₂O + CO + OH (R14) during the early stage of the pyrolysis. Thirdly, and at a later stage, a stronger pressure effect for the SC₂H₄OH sub-mechanism is observed with the reactions SC₂H₄OH(+M) $\rightleftarrows$ CH₃CHO + H(+M) (R15) and SC₂H₄OH(+M) $\rightleftarrows$ CH₃ + CH₂O(+M) (R16). These three steps could explain the slower CO induction and higher CO plateau from this model [36]. The detailed comparison between the Luo et al. [20] and the Khan-Ghauri et al. [36] models can be found in the Supplementary Material.

CCOC*OOH ⇄ CjOC*OOH + CH₃

(R13)

CjOC*OOH ⇄ CH₂O + CO + OH

(R14)

SC₂H₄OH(+M) ⇄ CH₃CHO + H(+M)

(R15)

SC₂H₄OH(+M) ⇄ CH₃ + CH₂O(+M)

(R16)

Figure 11 shows the sensitivity analysis for EMC (CCOC*OOC in the model). At 1.3 atm (based on Grégoire et al. [9]), the EMC key reactions were determined as CCOC*OOC $\rightleftarrows$ COC*OOH + C₂H₄ (R17), followed by COC*OOH $\rightleftarrows$ CH₃OH + CO₂ (R18), and it remains the same at 5.5 atm.

CCOC*OOC ⇄ COC*OOH + C₂H₄

(R17)

COC*OOH → CH₃OH + CO₂

(R18)

Two reaction pathways are available for CH₃OH, i.e., the sequences CH₃OH $\to$ CH₃O $\to$ CH₂O $\to$ HCO $\to$ CO (R19–R20–R10–R7) and CH₃OH $\to$ CH₂OH $\to$ CH₂O $\to$ HCO $\to$ CO (R21-R12-R10-R7), where both routes are strongly enhanced at higher pressure, see Figure 11c.

CH₃OH + H ⇄ CH₃O + H₂

(R19)

CH₃O(+M) ⇄ CH₂O + H(+M)

(R20)

CH₃OH + H ⇄ CH₂OH + H₂

(R21)

As commented for DEC pyrolysis, the EMC decomposition also shows an alternative pathway via CH₃OCO with the reaction CH₃OCO $\rightleftarrows$ CH₃O + CO (R22) promoted by CCOC*OOC(+M) $\rightleftarrows$ CH₃OCO + C₂H₅O(+M) (R23), see Figure 11a,b.

CH₃OCO ⇄ CH₃O + CO

(R22)

CCOC*OOC(+M) ⇄ CH₃OCO + C₂H₅O(+M)

(R23)

Note that the CH₃OCO directly produces CO with the reaction (R22) but is not the main CO producer. Additionally, and as highlighted in Figure 12 using a rate-of-production analysis, the more traditional reaction HCO + M $\rightleftarrows$ H + CO + M (R7) is present and is fed by CH₃OCO via the sequence CH₃OCO $\to$ CH₃O $\to$ CH₂O $\to$ HCO $\to$ CO (R22–R20–R10–R7). Interestingly, the reaction (R17) inhibiting effect is tempered at higher pressure, enhancing the reaction (R23) during the EMC pyrolysis. It can be noted that this is the main difference between the Luo et al. [20] and the Khan-Ghauri et al. [36] models, which could explain the slightly larger discrepancies in the numerical predictions using [36], and this comparison can be found in the Supplementary Materials.

Ultimately, similar reaction pathways for both DEC and EMC persist at higher pressure when compared to the CO time-history results obtained in similar conditions at near-atmospheric pressure in Grégoire et al. [9] and serving as baselines. Nonetheless, the sensitivities are modified significantly, increasing both the CO induction time and the CO ppm levels. While models from the literature exhibit various performance, it seems that the core reactions from the Luo et al. [20] and the Khan-Ghauri et al. [36] mechanisms are identical, implying that no key reaction is missing and that fine tuning (i.e., high-level calculations) is still needed for DEC and EMC pyrolysis conditions, especially at higher pressure, for realistic understanding in the battery cell during thermal runaway.

5. Conclusions

This research investigated how pressure influences the pyrolysis of two commonly used linear electrolytes, diethyl carbonate (DEC) and ethyl methyl carbonate (EMC). As gas forms within the battery cell, pressure increases until the gases are vented, which could potentially result in a fire hazard. The pyrolysis of DEC and EMC was examined in the gas phase, in a 99.75% He/Ar mixture, under high-temperature conditions and pressures close to 5.5 atm. A quantum cascade laser was utilized to measure the time-dependent CO production, generating a unique set of experimental data. CO time-history results obtained under similar conditions at near-atmospheric pressure from Grégoire et al. [9] were used as reference points for comparing DEC and EMC. Numerical predictions based on detailed chemical kinetics models from the literature were performed, and the Luo et al. [20] model exhibited satisfactory performance to be used for further analyses. Highlights in the reaction pathways at high pressure and their impact on CO formation were described and illustrated. In essence, the DEC and EMC sub-mechanisms at 1.3 atm are driven by ethanol (C₂H₅OH) and methanol (CH₃OH), respectively, whereas at higher pressure, the DEC and EMC find alternative routes with CH₃CHO and CH₃OCO, respectively. Note that the former one, CH₃CHO produces dominantly CO via CH₃CHO $\to$ CH₃CO $\to$ CO, instead of the more traditional HCO + M $\rightleftarrows$ H + CO + M (R7). The latter one still relies on the reaction (R7) with the sequence CH₃OCO $\to$ CH₃O $\to$ CH₂O $\to$ HCO $\to$ CO along with a direct CO formation from CH₃OCO $\rightleftarrows$ CH₃O + CO (R22). Finally, more validation targets obtained with laser absorption experiments will further improve the insight on the fire hazard from LIBs.