Uncovering hydrogen anion triggered thermal runaway mechanism of a high-energy LiNi0.5Co0.2Mn0.3O2/graphite pouch cell


 The continuous energy density increase of lithium ion batteries (LIBs) inevitably accompanies with the rising of safety concerns. Here, the thermal characteristics of a high-energy 5 Ah LiNi0.5Co0.2Mn0.3O2/graphite pouch cell using a thermally stable dual-salt electrolyte were analyzed. Heat determination during cycling at two boundary scenarios of adiabatic and isothermal environment clearly states the necessity of designing an efficient and smart battery thermal management system. More importantly, it is innovatively proposed that the hydrogen anion (H−) induced H2 releasing at anode side and H2 migration to cathode side is the rooted thermal runaway trigger of the pouch cell, while the phase transformation of lithiated graphite anode and the O2-releasing by delithiated cathode are just accelerating factors for thermal runaway. These findings will shed promising lights on thermal runaway route map depiction and thermal runaway prevention, as well as formulation of electrolyte for high energy safer LIBs.


Introduction
This decade has witnessed the increasing challenges coming from petrochemical energy crisis and environmental pollution. Hence, the demand for developing advanced and renewable energy technology with clean processes and high e ciency is becoming more and more necessary. Great efforts towards promoting alternative renewable energy resources (such as solar, wind, photovoltaic and tidal energy, etc.) have been devoted. 1 The intermittent nature of these renewable energy resources facilitates the booming of varied energy storage systems. Due to their advantages like high energy density, low maintenance and will contribute to the initiation of thermal runaway of LIBs. 26 It is noted here that, battery thermal runaway will occur at any state of charge (SOC) in practical cases. But in most of previous reports, the thermal runaway mechanisms are always deciphered at high SOC. Until now, due to the complexity of exothermic chain reactions inside LIBs and the limitation of the existing testing approaches for thermal safety study, it is still di cult to obtain a clear and accurate thermal runaway route map depicting the rooted interactions among cathode, anode, electrolyte and separator.
Obviously, during the hazardous thermal runaway (smoke, re and even explosion) of LIBs, electrolyte is almost involved in every exothermic chain reaction. 4,[27][28] As the "blood" of the LIBs, the organic electrolyte has a crucial in uence on the electrochemical and inherent thermal safety of LIBs. Lithium hexa uorophosphate (LiPF 6 ) with well-balanced properties is always adopted as the main conducting lithium salt for the widely commercialized carbonate-based electrolytes. However, the thermally unstable and moisture sensitive LiPF 6 is susceptible to generate undesired reactive species such as HF and POF 3 , inducing the destruction of electrode/electrolyte interface layers and the transition metals dissolution from cathode. [29][30][31] Hence, tremendous efforts have been devoted to design and synthesize alternative lithium salts. 29,32 Wherein, thermally stable and highly conductive lithium imides of lithium bis(tri uoromethanesulfonyl) imide (LiTFSI) and lithium bis( uorosulfonyl) imide (LiFSI) have aroused great interests. But, both LiTFSI and LiFSI easily cause corrosion of Al current collector at high voltages exceeding 4 V. [33][34][35][36] Formulating high cost concentrated electrolyte will effectively prohibit LiFSI-induced Al corrosion, but, the LIB still undergoes thermal runaway due to the strong heat-releasing reaction between LiFSI salt and lithiated graphite. 23 Another economic way to suppress Al corrosion is to formulate blended-salt electrolytes by mixing LiTFSI or LiFSI with lithium di uoro(oxalate) borate (LiDFOB) or lithium bis(oxalato)borate (LiBOB). 9,28,36−43 Recently, blended-salt electrolytes showing synergistic effects have achieved great progress in the burgeoning eld of next-generation lithium batteries. 28 Nevertheless, to the best of our knowledge, the electrochemical and thermal safety evaluation of blended-salt electrolytes in large format and large capacity LIBs are seldom reported.
Here, we conduct electrochemical and thermal safety study of a pouch-type 5 Ah LiNi 0.5 Co 0.2 Mn 0.3 O 2 /graphite (NCM523/G) battery with a dual-salt electrolyte, which is formulated by dissolving LiTFSI and LiDFOB lithium salts in carbonate solvents of ethylene carbonate (EC), propylene carbonate (PC), and ethyl methyl carbonate (EMC). In speci c, the electrochemical performances of 5 Ah NCM523/G pouch cell are investigated over a wide temperature range (-40℃ ~ 60℃). To state the importance of designing an e cient and smart battery thermal management system, heat generation of 5 Ah NCM523/G pouch cell during charge-discharge processes is determined at both adiabatic (ARC) and isothermal (IMC) conditions. More importantly, varied advanced characterization techniques (such as temperature-resolved XRD, ARC, on-line titration gas analysis system) are used to elucidate the thermal compatibility of battery materials disassembled from pouch cell with different SOC (100% and 0%).
Finally, we innovatively propose that the hydrogen anion (H − ) induced H 2 releasing at anode side and H 2 migration to cathode side is the rooted thermal runaway trigger of this NCM523/G pouch cell, while the phase transformation of lithiated graphite anode and the O 2 -releasing by delithiated NCM523 cathode are just accelerating factors for thermal runaway. This study provides a deeper insight for understanding of the inherent mechanism of thermal runaway of LIBs, and lights the way to advanced design philosophy of next generation safer LIBs.

Results And Discussion
Electrochemical performances over a wide temperature range Undoubtedly, LIBs operating over a wide temperature range presents great importance in practical applications such as those in electric vehicles, space and military missions. For wide temperature LIBs, the most challenging work is to get a compromise between subzero temperature performance and high temperature performance. The most direct and e cient strategy is formulating wide temperature electrolyte by dissolving thermally stable lithium salts in low melting point and high boiling point solvents with low viscosity. 7 EMC, T m = -53℃) and high boiling point (EC, T b =243℃; PC, T b = 242℃) are selected for formulating the wide temperature range electrolyte. In the 1st formation cycle at room temperature (RT), 5 Ah NCM523/G pouch cell with the dual-salt electrolyte demonstrated high initial Coulombic e ciency of 85.8% (5.14/5.99 Ah) than that (47.6%, 2.18/4.58 Ah) of the pouch cell with 1 M LiPF 6 EC/PC/EMC (1:1:3, by volume) (Fig. 1a). Moreover, severe gas swelling of pouch cell with 1 M LiPF 6 EC/PC/EMC is observed (Left inset in Fig. 1a), due to the PC-induced formation of unstable solid-electrolyte-interphase (SEI) layer on graphite anode. [44][45] As a sharp contrast, pouch cell with the dual-salt electrolyte shows no obvious swelling (Right inset in Fig. 1a), bene ting from the formation of stable SEI layer by LiDFOB salt. [46][47] Therefore, 5 Ah NCM523/G pouch cell with the dual-salt electrolyte is employed for the following electrochemical and thermal safety study.
After the vacuum degassing procedure of formation process, the determined gravimetric energy density of 5 Ah NCM523/G pouch cell with dual-salt electrolyte is highly up to 208.8 Wh kg − 1 . At RT at 0.5 C rate for 400 cycles, the pouch cell presents a high average Coulombic e ciency of 99.88% and delivers a discharge capacity retention of 92.72%, 84.10%, 74.90% and 68.77%, at the 100th, 200th, 300th, and 400th cycle, respectively (Fig. 1b-c). When the temperature is increased up to 60 ℃, the discharge capacity retention of the pouch cell at 0.2 C rate is 82.23%, 79.02%, 76.56% and 74.48% at the 100th, 200th, 300th and 400th cycle, respectively, with an average Coulombic e ciency reaching to 99.92% (Fig. 1d-e). In subsequent, subzero temperature performance is revealed by discharging the roomtemperature fully charged 5 Ah NCM523/G pouch cell, at varied subzero temperatures of -10 °C, -20 °C, -30 °C, and − 40 °C, respectively (Fig. 1f). The corresponding discharge capacity retention is 85.79%, 75.76%, 64.46%, and 40.81%. In general, the high-energy 5 Ah NCM523/G pouch cell using the formulated dual-salt electrolyte is very competent for wide temperature range applications. Moreover, it is noted that, to the best of our knowledge, 29 this is the rst case to evaluate the application of dual-salt electrolytes in LIBs over a wide temperature range, which is signi cant for the commercialization process of dual-salt electrolytes.
Heat generation during charge-discharge operation Thermal anomalies inside the cell or pack in the absence of proper thermal management can instigate accelerated cell capacity degradation and even hazardous thermal runaway. Therefore, the heat generation accompanied with battery charge-discharge operation is receiving more and more attention.
Generally, heat generated from battery cycling can be divided into reversible heat and irreversible heat. Irreversible heat refers to the ohmic heat from the polarization or overpotential of the cell, while reversible heat is determined by the measurement of cell entropic coe cient depending on the intrinsic nature of the electrode materials (relating with the atom arrangement in the crystal lattice). 48 Deciphering the heat generation law of a single cell is essential to design and optimize the battery thermal management system, which ensures batteries in a pack or module running in an ideal temperature range. Here, two boundary scenarios of adiabatic and isothermal environment, representing the worst and best case for the heat management, respectively, are considered for heat determination of the pouch cell with dual-salt electrolyte.
ARC (an adiabatic calorimeter) is a pivotal integrated technology to study the thermal safety of LIBs at multilevel, and research progress on ARC is reviewed by us very recently. 22 In general, ARC will simulate an accurate adiabatic condition by keeping the cavity temperature consistent with the sample temperature, preventing the self-generated heat loss of sample. Therefore, ARC is the worse-case scenario. Here, the 5 Ah NCM523/G pouch cell with the dual-salt electrolyte is placed in the adiabatic cavity of ARC (BTC500, HEL, Figure S1) and connected with a charge/discharge apparatus. Obviously, the surface temperature of pouch cell increases during both charge and discharge process ( Fig. 2a and Figure S2a-c). For example, at 0.5 C rate, the overall heat generation (19.5℃, 1.9 kJ) during the charge process is much higher than that (6.3℃, 0.6 kJ) of the discharge process. At varied C-rates, the selfheating rate curves during charge and discharge process are symmetrical ( Figure S2d-f), evidencing that the heat generation mainly consisted of the irreversible joule heat and the reversible electrochemical reaction heat. [49][50] Obviously, the reversible electrochemical reaction heat dominates the total heat at low rates and irreversible joule heat dominates the total heat at high rates. These results clearly tell us that, at the worst case of adiabatic condition, the 5 Ah NCM523/G pouch cell at higher rates will easily get thermal runaway in few cycles, revealing the importance of battery thermal management. As for the battery thermal management, it is necessary to determine the heat generation of a single cell at a constant temperature. Here, as a best-case scenario, IMC (isoBTC, HEL) is used to test the heat generation of the 5 Ah NCM523/G pouch cell at the isothermal condition ( Figure S3). At 0.5 C rate at 30℃, the overall heat generation (1.2 kJ) during the charge process is much higher than that (0.4 kJ) of the discharge process (Fig. 2b). Clearly, the released energy during 1 charge-discharge cycle at isothermal condition (1.6 kJ) is lower than that at the adiabatic condition (2.5 kJ). Interestingly, at 0.5 C rate, more heat is generated at both 10℃ and 50℃ when compared to 30℃, suggesting the importance of selecting work temperature for battery ( Figure S4). The reversible and irreversible heat are determined based on the ohmic resistance and total heat generation. 51 The internal resistance of the pouch cell is calculated by the hybrid pulse power characterization method ( Figure S5). On the whole, the reversible heat dominates the total heat generation at 0.5 C rate, and the endothermic and exothermic reactions are distinguishable and transformable during the charge or discharge process (Fig. 2c). Compared to fresh pouch cell, long-term (400 cycles) cycled pouch cell demonstrates much higher heat release power at higher rates, especially at 1 C rate (Fig. 2d). In summary, the design of an e cient and smart battery thermal management system must comprehensively consider the effects of work temperature, state of charge (SOC), charge-discharge current rate, and charge-discharge protocol on heat generation, and IMC testing will give the answers. Thermal runaway feature and mechanism As a high-energy storage reservoir, LIBs easily get thermal runaway when operated under mechanical, electrical, and thermal abuse conditions. To develop an e cient battery safety risk controlling strategy, it is necessary to obtain some critical parameters, such as onset temperature for self-heating (Tonset), selfheating rate (SHR), thermal runaway temperature (Ttr), and maximum temperature (Tmax), etc. Here, 5 Ah NCM523/G pouch cell using dual-salt electrolyte is placed in the cavity of ARC (BTC500, HEL, Fig. 3a and Figure S6), and the typical heat-wait-search (HWS) mode is used to study thermal runaway features of the pouch cell. Under the HWS mode of ARC, the built-in camera in cavity captures that smoke and ame is rapidly spurted out of the pouch cell (100% SOC, after the formation process) (Fig. 3b). Obviously, the T onset , T tr (SHR over 1℃ min − 1 as the criteria) and T max is determined to be 91℃, 171℃, and 516℃, respectively (Fig. 3c). After the thermal runaway process, the aluminum plastic lm is severely damaged (inset in Fig. 3c). It is noted here, that most previous battery thermal runway investigations focus on the fully charged cell (100% SOC) because of its violent thermal runway hazardous, while the fact that fully discharged cell (0% SOC) also undergoes thermal runway is neglected. Here, to explore the rooted mechanism for the triggering of thermal runway behavior, 5 Ah NCM523/G pouch cell without dual-salt electrolyte and 5 Ah NCM523/G pouch cell with dual-salt electrolyte but without formation process are fabricated and tested by ARC under same conditions, and interestingly, both pouch cells do not present severe exothermic reactions related to the thermal runway of pouch cell below 250℃ (Fig. 3d). However, when the 5 Ah NCM523/G pouch cell with dual-salt electrolyte is cycled for one formation cycle (0% SOC), the thermal runaway occurs under the same testing condition in ARC (T onset = 141℃, T tr = 199℃, T max = 280℃). These results clearly tell us that the formed interfacial layer components between the electrolyte and electrodes play a crucial role in triggering the thermal runaway process of pouch cell. Therefore, we con rm that pouch cell after formation process will go thermal runaway anyway regardless of the SOC, and pouch cell with higher SOC demonstrates faster and more severe thermal runaway process ( Figure  S7).
To decipher the rooted mechanism for triggering the thermal runaway of NCM523/G pouch cell, ARC (BTC130, HEL) equipped with a small bomb chamber is used to study the thermal compatibility of battery components ( Fig. 4a and Figure S8). In the Ar-lled glove box, the anode and cathode are carefully and separately dissembled from fully charged (100% SOC) and fully discharged (0% SOC) 5 Ah NCM523/G pouch cell. The as-formulated fresh dual-salt electrolyte presents a high T onset of 218℃, suggesting its high thermal stability (Fig. 4b). The fully delithiated cathode/electrolyte (100% SOC) and lithiated cathode/electrolyte (0% SOC) demonstrate T onset of 134℃ (Fig. 4c) and 171℃ (Fig. 4d), respectively.
And the fully lithiated anode/electrolyte (100% SOC) and delithiated anode/electrolyte (0% SOC) exhibit the T onset of 95℃ (Fig. 4e) and 128℃ (Fig. 4f), respectively. These results reveal that, at different SOCs, cathode/electrolyte and anode/electrolyte will inevitably go thermal runaway and anode/electrolyte is easier to get thermal runaway than cathode/electrolyte. In addition, higher thermal reactivity is presented at higher SOCs. Therefore, it is concluded here that high SOC is just the accelerating factor of battery thermal runaway, but not the root cause for triggering it. After thermal runaway of the dual-salt electrolyte, anode/electrolyte, and cathode/electrolyte, the remained incondensable gas species in the sealed bomb chamber is collected and analyzed by the mass spectrometer (MS), respectively (Insets in Fig. 4b-e). Wherein, the remained gas species for thermal runaway of sole as-formulated dual-salt electrolyte is mainly consisted of CO 2 (51.8%) and CH 4 (24.7%) (Insets in Fig. 4b). For fully delithiated NCM523 cathode/electrolyte (100% SOC), O 2 (87.9%) dominates the collected gas species after thermal runaway (Insets in Fig. 4c). Revealed by temperature-resolved XRD equipped with an in-situ heating module, when the temperature exceeds 200℃, the fully delithiated NCM523 cathode will release O 2 by phase transformation of layered structure ((003) R ) to spinel structure ((111) S ) and disordered spinel structure ((108) R and (110) R ) to the NiO-like rock-salt structure ((440) S ) (Fig. 4g). As an interest comparison, lithiated NCM523 cathode (0% SOC) reacting with electrolyte produces main gas species of CO 2 (48.7%), CH 4 (26.8%) and O 2 (13.4%). The dramatic decrease of O 2 percentage is possibly attributed to the wellpreserved crystal structure of lithiated NCM523 cathode (0% SOC) (Fig. 4h). To further verify this, XRD patterns of the powders collected after thermal runaway of 0 and 100% SOC 5 Ah NCM523/G pouch cell are tested. It is presented that the fully delithiated NCM523 cathode (100% SOC) undergoes phase transformation and crystal structure collapse while the main layered structure of lithiated NCM523 cathode (0% SOC) is preserved ( Figure S9). These results indicate that O 2 releasing accompanying with phase transformation of delithiated NCM523 cathode at high temperature around 200℃ may aggravate the burning or explosion, but is not the rooted mechanism for triggering thermal runaway. Obviously, the rooted factor for triggering the thermal runaway still depend on anode side. After thermal runaway of fully lithiated anode/electrolyte (100% SOC), the dominated gas species is H 2 (68.2%), while CO 2 (39%), CH 4 (26.5%) and H 2 (18.8%) are the main gases determined for thermal runaway of unlithiated anode/electrolyte (0% SOC). Except the evolution of CO 2 and CH 4 from thermal decomposition of electrolyte, H 2 gas constitutes the most dominating gas species in the thermal runaway of anode/electrolyte at 0% and 100% SOC. This implies that, clarifying the origin of H 2 evolution is of great signi cance to understand the rooted triggering factor for thermal runaway. Temperature-resolved XRD equipped with an in-situ heating module is also utilized to analyze the bulk phase change of graphite anode at different SOC. For lithiated graphite anode (100% SOC), the exothermic phase transformation of LiC 6 (001) to LiC 12 (002), and LiC 12 (002) to C (002) occurs at ca. 100℃ and ca. 200℃, respectively ( Fig. 4i). But, even if there is no such exothermic phase transformations for delithiated graphite anode (0% SOC) (Fig. 4i), the delithiated graphite anode/electrolyte (0% SOC) still exhibits thermal runaway and the self-heating happens earlier than lithiated NCM523 cathode/electrolyte (0% SOC). Hence, the LiC 6 -LiC 12 -graphite phase transformation is also excluded to be the rooted factor for triggering thermal runaway. In subsequent, it is not di cult to infer that the thermal runaway of pouch cell is induced by the exothermic reactions related to the broken of the formed SEI layer on graphite anode. This is also evidenced by the fact that, cycled graphite anode (delithiated state, 0% SOC) reacting with electrolyte shows exothermic peak at ca. 80 ~ 170℃, which is not appeared for the uncycled pristine graphite anode and electrolyte, during the DSC test ( Figure S10). Conventionally, SEI layer on graphite anode are typically determined to be consisted of inorganic species (such as LiF, Li 2 CO 3 , Li 2 O, Li 2 C 2 , Li 2 C 2 O 4 , LiOH, etc.) and organic species (such as ROCO 2 Li, (CH 2 OCO 2 Li) 2 , ROLi, CH 3 Li, etc.). [52][53][54][55][56][57] Although the broken of SEI layer is always identi ed as the initial stage during the thermal runaway process, but the exothermic reactions related to the thermal decomposition of listed inorganic and organic species are not presented in detail to date. Moreover, the underlying cause for large amounts of H 2 evolution at graphite anode side is still remained unclear. Therefore, some unknown species must be still unidenti ed in the SEI layer of graphite anode. In some reports, it is suggested that the graphite anode surface is enriched of hydrogen, 26, 58-59 but no further detailed evidence is provided. Herein, by employing a delicately-designed deuterium-oxide (D 2 O) titration device connecting with an on-line gas analysis mass spectrometry (MS) system (Fig. 5a), we unprecedentedly identify the existence of H − containing species on the surface of the graphite anode. Fully-lithiated graphite (100% SOC) and fully-delithiated graphite (0% SOC) is titrated by deuterium-oxide (D 2 O) with the following guideline reactions: Li x C 6 + xD 2 O → xLiOD + C 6 + x/2 D 2 ↑, (x ≤ 1); H − + D 2 O →OD − + HD↑. Surprisingly, HD (m/z = 3) signal is observed for both fully-lithiated graphite (100% SOC) and fully-delithiated graphite (0% SOC). In addition, the mole of HD and D 2 (m/z = 4) at fully-lithiated graphite (100% SOC) is 0.14 µmol mg − 1 and 4.26 µmol mg − 1 , respectively (Fig. 5b), which decrease to 0.014 µmol mg − 1 and 0.029 µmol mg − 1 , respectively, when the graphite anode is fully-delithiated (0% SOC) (Fig. 5c). It is, the rst time, discovered that H − containing species, most probably in the form of LiH, do exist in the SEI layer of graphite anode, and the H − containing species exhibit highly electrochemical reversibility at graphite anode surface during cycling. In another delicately-designed experiment, fullylithiated graphite (100% SOC) is heated at 90℃, in the titration vessel of abovementioned on-line gas analysis mass spectrometry (MS) system. H 2 signal (m/z = 2) is detected after heating to 90℃, con rming that the broken of SEI layer is accompanied by evolution of H 2 ( Figure S11).
Obviously, the released amount of H 2 (Fig. 4e-f) is highly correlated with the determined amount of H − containing species. Highly lithiated graphite (high SOC) contains more H − containing species and releases more H 2 when heating. Moreover, it is previously proposed that the recombination of the active hydrogen to form H 2 will lead to huge generation of heat, which is even larger than that of H 2 burning in oxygen, 26 implying that the H − -induced H 2 evolution accompanies with large amounts of heat releasing.
Furthermore, the presence of H − is also con rmed in graphite anode disassembled from NCM523/G and NCM811/G pouch cells (using conventional LiPF 6 based carbonate electrolyte), regardless of SOC, indicating this phenomenon is universal applicable for varied NCM/G cells ( Figure S12). These amazing results indicate that the H − induced H 2 releasing at the graphite anode side may be the rooted thermal runaway trigger of NCM523/G pouch cell, while the phase transformation of lithiated graphite anode and the O 2 -releasing by delithiated NCM cathode are just accelerating factors for thermal runaway.
The aforementioned released gases and temperature pro les during thermal runaway of anode/electrolyte and cathode/electrolyte are separately determined and analyzed, not considering of anode/cathode crosstalk. However, in practical cases, despite anode and cathode is physically separated by the separator, the porosity of separator always allow the crosstalk of byproducts. 10,60 Then the question comes: during the triggering process of thermal runway, is there any crosstalk effect when the released gas species migrate through porous separators? To answer this question, a self-made two bomb chamber testing system is delicately fabricated, where the anode and cathode is placed separately, but the released gas species can ow freely by the connected pipeline, and two bomb chambers are heated by one same heating wire in the quasi-adiabatic cavity of ARC ( Figure S13, and insets in Fig. 6a-b). Connected with the bomb with fully delithiated NCM523 cathode/electrolyte, the fully-lithiated graphite/electrolyte reaction delivers a same T onset of 95℃ and a lower T tr (from 306℃ to 247℃) ( Fig. 6a), compared to the one bomb test of the fully-lithiated graphite/electrolyte mentioned above (Fig. 4e). This clearly tell us that the triggering of anode/electrolyte thermal runaway at low temperatures is not affected by gas species generated by cathode/electrolyte reactions. But at high temperatures, the released gas species (especially O 2 ) will accelerate the thermal runaway of anode/electrolyte. Another two-bomb testing reveals that the gas species (especially H 2 ) produced by anode/electrolyte decrease the thermal stability of cathode/electrolyte (Fig. 6b). In speci c, T onset drops from 136℃ to 116℃, and a sharp temperature rise from 159℃ to 285℃ is appeared in the HWS curve. Furthermore, gas species collected from the two-bomb system (100% SOC electrodes) are mainly consisted of CO 2 (28.5%), CH 4 (32.7%), H 2 (15.2%), and O 2 (12%), suggesting that H 2 from anode and O 2 from cathode side is consumed, and subsequently, evidencing the occurrence of the crosstalk of the released gas species during thermal runaway ( Figure S14). Moreover, H 2 is preliminarily calculated to have higher a nity to NCM ( Figure S15), suggesting that the released H 2 has high tendency to react with the NCM in elevated temperatures. In summary, it is concluded here that the H − -induced H 2 releasing at anode side and H 2 migration to cathode side is the rooted thermal runaway trigger of NCM523/G pouch cell, while the phase transformation of lithiated graphite anode and the O 2 -releasing by delithiated NCM523 cathode are just accelerating factors for thermal runaway. Based on all the analysis and experiments, a modi ed and upgraded thermal runaway route map for 100% SOC NCM523/graphite pouch cell is depicted here (Fig. 6c): 1) under abuse (mechanical, electrical, or thermal) conditions, when the battery temperature increases to ca. 100℃, mild exothermic reactions related to phase transformation of LiC 6 to LiC 12 and SEI layer destruction happens; The SEI layer broken is accompanied by H − -induced H 2 releasing and corresponding large amounts of heat releasing; 2) Parts of released H 2 will diffuse to cathode side, possibly reacting with the fully delithiated NCM to release heat; The reactions in step 1 and step 2 will raise the temperature to ca. 200℃, at which the polyole n separators have been melted and partial short circuit between the cathode and anode poles will continue to push up the temperature; 3) when the temperature is pushed up to the range of 200-250℃, three severe exothermic reactions happens (electrolyte decomposition; phase transformation of LiC 12 to graphite; and O 2 -releasing by delithiated NCM523 cathode) and the released large amounts of gases (O 2 , H 2 , CH 4 , CO, C 2 H 4 , etc.) lead to the nal severe thermal runaway (smoke, re, and even explosion).

Conclusion
In this paper, a LiTFSI-LiDFOB based dual-salt electrolyte has been successfully and unprecedentedly demonstrated to possess high compatibility with a high-energy (208.8 Wh kg − 1 ) 5 Ah NCM523/G pouch cell. This pouch cell delivers excellent electrochemical performances over a wide temperature range (-40 60℃). In subsequent, the thermal safety characteristics of this high-energy pouch cell is investigated. Heat determination under adiabatic condition states the necessity of designing battery thermal management system, while heat determination under isothermal condition reveals that an e cient and smart battery thermal management system must comprehensively consider the effects of work temperature, state of charge (SOC), charge-discharge current rate, and charge-discharge protocol on heat generation. More importantly, by varied advanced characterization techniques, it is innovatively proposed Electrochemical measurements over wide temperature range The 5 Ah NCM/G pouch cells were charged and discharged using a LAND battery testing system (Wuhan LAND electronics Co., Ltd. China). All pouch cells were charged and discharged at 0.1 C at room temperature for the one formation cycle, then vacuum degassed to remove gas pocket. For room temperature cycling performance at 0.5 C, all pouch cells were charged to 4.2 V, followed by a constant potential step at 4.2 V for 10 min, then discharged to 3.0 V. For 60℃ cycling performance at 0.2 C, all pouch cells were charged to 4.2 V, followed by a constant potential step at 4.2 V for 10 min, then discharged to 3.0 V. For low temperature discharge performance, all pouch cells were charged to 4.2 V at 0.2 C, followed by a constant potential step at 4.2 V for 10 min, then discharged to 3.0 V at 0.1 C at varied subzero temperatures of -10 °C, -20 °C, -30 °C, and − 40 °C.
Heat generation at adiabatic condition For heat generation determination at adiabatic condition, the 5 Ah NCM/G pouch cells were put inside the ARC cavity (BTC500, HEL, England) and connected with a charge-discharge device (Wuhan LAND electronics Co., Ltd. China) (See Figure S1). The adiabatic test mode of ARC is used, and three different current rate of 0.25 C, 0.5 C, and 1 C were used to investigate the temperature increase and self-heating rate of the pouch cell during the charge and discharge process. After each charge or discharge process, the cavity temperature of ARC was cooled back to the initial temperature of 30℃.
Heat generation at isothermal condition Isothermal heat generation during charge and discharge process was determined by an isothermal battery calorimeter (IBC, iso-BTC, HEL, England). The tested pouch cell was enclosed by graphite lm with high heat conductivity. Heating sheets were xed on graphite lm ( xed on the side close to pouch cell), and power sensor was xed on pouch cell (See Figure S3). The pouch cell was tested at 0.5 C at varied temperatures of 10 °C, 30 °C, and 50 °C, respectively, and the corresponding total speci c heat generation rate (q) was determined.
The total speci c heat generation rate (q) included the reversible heat generation rate (q r ) and irreversible heat generation (q i ). q was determined by the IBC testing, and q i is calculated by I 2 R, where I is the working current, and R is the total internal resistance. The reversible heat generation rate (q r ) was obtained by q r = qq i . The total internal resistance R of the pouch cell was determined by the hybrid pulse power characterization using the battery test system. The pouch cell was adjusted to a speci c SOC (10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, and 100%), and then the following protocol was used: 1 C rate pulse for 10 s discharge, followed with a 40 s rest, then a 0.75 C pulse was used to charge for 10 s. Each SOC adjustment was followed by a one-hour rest period. Finally, the total internal resistance R during charge and discharge was calculated as mentioned previously. 51 Study of thermal runaway feature and mechanism Page 13/20 Thermal runaway study of 5 Ah NCM/G pouch cell was conducted using an ARC BTC500 (HEL, England), which was equipped with a digital camera (see Figure S6). Here, typical Heat-Wait-Seek (HWS) mode was adopted for testing: one heating step was 5 °C; detection limit was 0.03 °C min − 1 ; the temperature was raised from 40 to 250 °C. The thermal runaway criteria was 1℃ min − 1 .
Thermal compatibility study of battery materials was investigated by a small standard ARC (BTC130, HEL, England, see Fig. 4a and Figure S8). The anode and cathode are carefully dissembled from fully charged (100% SOC) and fully discharged (0% SOC) 5 Ah NCM523/G pouch cells, in an argon-lled glovebox. The electrode materials were carefully removed from the current collector with a surgical blade. 0.5 mL electrolyte and 1 g electrode material were transferred into the small bomb chamber made of Hastelloy alloy. Here, typical HWS mode was adopted for testing: one heating step was 5 °C; detection limit was 0.03 °C min − 1 ; the temperature was raised from 40 to 250 °C. The thermal runaway criteria was 1℃ min − 1 . When the small bomb chamber was cooled down after thermal runaway, the released gases was collected to be analyzed by a mass spectrum (MS, HPR-20, Hiden Analytical Ltd.).
The thermal stability of battery materials was also tested in a Simultaneous Thermal Analyzer (STA, Netzsch): heating rate was 5℃ min − 1 ; the temperature was raised from 25 to 500 °C. Temperatureresolved X-ray diffraction XRD (with Cu Kα radiation, λ = 1.5406 Å) was tested on a Ultima IV of Rigaku: the sample was sealed in an Ar-lled container; the angle ranged from 10°to 80°with a scan speed of 20°m in − 1 ; the temperature was raised from 30 °C to 400 °C with a heating rate of 1 °C min − 1 .
Graphite anode with 0% SOC and 100% SOC was well-sealed in a container of the titration unit, respectively. After connecting the sample-containing titration unit to the on-line gas analysis system, the Ar carrier gas was switched to go through the titration unit, then 1 ml D 2 O was injected and the gas species (mainly D 2 and HD) was recorded by the (MS HPR-20, Hiden Analytical Ltd.). In another test, the unit containing the 100% SOC graphite anode was heated to 90℃, and the released gas species were detected (See Figure S11a).
To study the crosstalk effects of released gas species, a self-made two bomb chamber testing system is delicately fabricated (Figure S13), where the anode and cathode is placed separately, but the released gas species can ow freely by the connected pipeline, and two bomb chambers are heated by one same heating wire in the quasi-adiabatic cavity of ARC.    (a) Schematic illustration of delicately-designed on-line D2O titration gas analysis MS system. The HD and D2 evolution rate curve after D2O titration on samples of (b) fully-delithiated graphite (0% SOC) and (c) fully-lithiated graphite (100% SOC) . (mg-1, divided by the sample mass for titration). The graphite anodes are carefully and separately dissembled from fully charged (100% SOC) and fully discharged (0% SOC) 5 Ah NCM523/G pouch cells.