Lithium-ion battery, also known as "rocking chair battery", is composed of four main components: positive electrode, negative electrode, electrolyte and separator. It mainly relies on the movement of lithium ions between the positive and negative electrodes to work. The working principle of a lithium-ion battery is shown in Figure 1.

When a lithium-ion battery is charged, lithium ions move from the positive electrode to the negative electrode, while electrons flow in through an external circuit. This process is reversed during discharge. The more lithium embedded in the electrode, the more total energy the battery can store and the longer it can last.

Positive side reaction:

Negative side reaction:

Total reaction of whole cell:

The most popular commercial anode material is graphite. The positive electrode is mainly a layered oxide (such as lithium cobalt oxide), a polyanion (such as lithium iron phosphate), or a spinel (like lithium manganese oxide). The electrolyte is generally composed of an organic solvent and a lithium salt. The solvent is usually a mixture of organic carbonates such as ethylene carbonate (EC) or diethyl carbonate (DEC), etc., lithium salts such as lithium hexafluorophosphate (LiPF₆), lithium hexafluoroarsenate (LiAsF₆), lithium perchlorate (LiClO₄) and lithium tetrafluoroborate (LiBF₄), etc. The separator separates the positive and negative electrodes while allowing ions to pass through. The high oxidizability of the positive electrode, the high reducibility of the negative electrode and the low melting point and flammability of the electrolyte make the battery have potential safety hazards such as fire and explosion.

Some typical battery safety incidents are listed in Table 1. Fig. 2 is a picture of a partial battery fire accident. It can be seen that lithium-ion batteries can rupture, catch fire, or explode when exposed to high temperatures or short circuits. Lithium-ion batteries have a low probability of fire and explosion, but these widely reported incidents have worried consumers and forced millions of batteries to be recalled. Therefore, it is necessary to further study the thermal stability of each component of lithium-ion batteries and the heat release mechanism of single batteries, and further understand the nature of battery fire and explosion, so as to alleviate consumers'fear of battery safety and point out the direction for the design of the next generation of safer batteries.

According to the previous introduction of battery safety accidents, the safety problems of lithium-ion batteries mainly come from the thermal runaway of batteries^[6]. As shown in Fig. 3, abuse conditions are the main source of thermal runaway triggering, such as mechanical abuse (needling, impact, extrusion, etc.), electrical abuse (overcharge, overdischarge, short circuit, etc.), thermal abuse (overheating). These three types of trigger forms have certain internal relations. Mechanical triggering leads to short circuit and electrical triggering, while electrical triggering releases heat and causes thermal triggering, which is the core of battery thermal runaway. Under the action of various internal and external factors, the harmful chemical exothermic reaction inside the battery starts. When the accumulated heat is greater than the heat dissipation, the internal temperature of the lithium-ion battery continues to rise, and eventually thermal runaway occurs.

To regulate the battery internal temperature, it is crucial to understand the heat generation and heat transport inside the battery. Even with a robust external thermal management system, the higher heat production inside the battery along with poor heat transport can still result in a noticeable temperature rise and large internal temperature gradient. Therefore, understanding the internal heat generation mechanism of the battery is the primary problem to solve the safe operation of the battery.

When the battery operates normally, the internal heat is generated by electrochemical reaction, phase change and mixing enthalpy, and the latter two are often ignored^[7,8]. The total heat Q generated by a lithium-ion battery consists of two components, Q_rev and Q_irre. Q_rev is the reversible heat of reaction due to the entropic heat and depends on the entropic heat coefficient, the temperature derivative of the equilibrium potential. The Q_irre is irreversible heat, which is due to the overpotential heat produced by ohmic losses inside the cell, interfacial charge transfer and mass transfer losses. The calculation formula is shown in (4-6). Where V is the battery voltage, U is the equilibrium potential, I is the operating current, R is the internal resistance, and t is the operating time.

The above formula is suitable for evaluating the heat generation of a small lithium-ion battery without the heat of mixing, the heat of phase change, the heat of side reaction, and neglecting the Joule heat of the current collector. For the heat generation of large batteries used in pure electric vehicles, Kim et al. Considered that in addition to the heat generation of electrochemical processes (reversible reaction heat and irreversible impedance heat), the Joule heat generated by current collectors should also be included^[9,10]. Kim et al. Used a two-dimensional model to couple the Joule heat generated by the current collector into the electrochemical model in the form of an electrical model according to Ohm's law, and the heat generation equation obtained is as follows:

Where A is the specific surface area of the cell (m^-2),J is the current density (A·m^-2),U is the open circuit voltage (V),V is the working voltage of the cell (V),a_p and a_n are the specific surface (m^-2),i_p and i_n of the positive and negative electrodesAre the linear current densities of the positive and negative electrodes (A·m^-2),R_p and R_n are the resistances of the positive and negative electrodes (Ω), respectively. The first term on the right side of the equation is the heat due to charge transfer at the electrode/electrolyte interface, including an irreversible part representing the energy loss due to the deviation of the cell potential from the open circuit potential caused by electrochemical polarization, and a reversible part representing the heat due to entropy change; The latter two terms come from the Joule heating of the positive and negative electrodes, respectively. Reasonable design of the position and size of the tab can minimize the values of the latter two items.

Determining the heat generation rate of lithium-ion battery is of great significance for the design of battery thermal management system and the prediction of transient temperature rise during dynamic operation. Mao et al. Studied the heat generation characteristics of LiCoO₂/C battery under different conditions, and measured the reversible and irreversible heat generation by calorimetric method and electrochemical method respectively, both of which were consistent, but the former was more feasible^[11]. It is determined by analysis that the reversible heat generation is related to the electrode material and battery capacity, while the irreversible heat generation is related to the resistance and current. Taking the commercial cylindrical 18650 lithium-ion battery with NCM ternary cathode material as an example, Liu et al calculated the heat generation rate of the battery by full matrix orthogonal experimental design method^[12]. Based on the consideration of depth of discharge (DOD), discharge rate and ambient temperature, 176 overpotential tests and 66 entropy coefficient tests were carried out, and it was found that the average heat generation rate was approximately quadratically nonlinear dependent on the discharge rate. In addition, the third-order transient heat generation rate model of lithium-ion battery was established by response surface methodology (RSM) and correlated with each influencing parameter.

When the battery is exposed to abusive conditions, the temperature may exceed the normal operating range, and the active materials can decompose or react with each other, eventually leading to thermal runaway. The electrochemical reaction process in lithium-ion batteries at high temperature is very complex. Taking a lithium-ion battery with a positive electrode made of lithium cobalt oxide as an example, the thermal runaway process can be summarized as shown in Figure 4^[13]. As shown in Fig. 4, with the increase of temperature, exothermic side reactions such as decomposition of SEI film, reaction between anode material and electrolyte, reaction between cathode material and electrolyte, decomposition of electrolyte, and reaction between anode and binder occurred in the battery. Many of these occur in parallel.

Feng et al. Conducted adiabatic thermal runaway tests on a large number of battery samples, disassembled the battery, and conducted differential scanning calorimetry (DSC) tests on the combination of battery components^[14]. The test results were jointly analyzed to analyze the common thermal runaway behavior of the battery. The thermal runaway process is divided into three stages according to the breakdown of the separator and the redox reaction between the positive electrode and the negative electrode (Fig. 5). In the first stage, the diaphragm is intact, and the "negative electrode + electrolyte" system and the "positive electrode + electrolyte" system react separately. The heat generated by the "positive electrode + electrolyte" system at this stage is low and can be ignored, and the heat generated by the "negative electrode + electrolyte" system accounts for about 17% of the total heat of the thermal runaway process. In the second stage, the diaphragm collapses, and a part of the solvent still exists in the electrolyte (the boiling point of EC is 248 ℃), while the other part evaporates. Internal short circuit occurs, which releases the electric energy stored in the battery, but the heat released by internal short circuit only accounts for 9% of the total heat. The third stage: the redox reaction between the positive electrode and the negative electrode starts and releases a large amount of heat, accounting for 74% of the total heat. It should be noted that the accelerating rate calorimeter (ARC) can detect the heat release behavior of battery thermal runaway under adiabatic conditions and provide temperature and pressure data, but it lacks the heat production of battery thermal runaway product gas and battery material combustion, so the three-stage division does not include the heat production of combustion.

The contributions of Internal short circuit (ISC) and exothermic chemical reactions in the thermal runaway process are difficult to separate. Under the condition of mechanical abuse, the occurrence of ISC will release a large amount of Joule heat, which can directly trigger the runaway of heat. The mechanical abuse (needling, indentation, etc.) of ISC mainly refers to the invasion of foreign bodies or mechanical deformation of the battery under the action of external forces, resulting in the formation of electrical connection between the positive and negative materials.

Puncture test to simulate internal short circuit is a commonly used method to evaluate the safety performance of batteries. Wang et al. Of the University of Science and Technology of China studied the internal short circuit phenomenon of lithium-ion batteries caused by penetration through needling experiments, numerical simulation and thermal analysis of battery diaphragms.Needling was found to determine the short-circuit current as well as the battery heat dissipation, and in the case of thermal runaway, the larger diameter nail provided a larger short-circuit current, resulting in a larger total heat production rate and a shorter time required to trigger thermal runaway^[61]. In the case of non-thermal runaway, the larger the nail diameter, the larger the shortening area, and the stronger the heat dissipation capacity. The cell voltage after insertion decays exponentially with time, and the fluctuation and rebound of the voltage during insertion are attributed to the partial interruption of the internal short circuit path. Xu et al. Studied the thermal runaway mechanism of lithium-ion soft-pack batteries through a series of needling experiments, and selected four parameters, including state of charge, needling speed, needling position and nail diameter, to study their effects on thermal runaway^[62]. The results show that the higher the organic carbon content of the negative electrode, the easier the thermal runaway of the battery during the needling process, and the higher the peak temperature. The needle position determines the battery short circuit position; The needle speed has little effect on the thermal runaway of the battery. The larger the nail diameter, the more serious the cell damage, and the larger the initial heating rate and peak temperature. In order to carry out detailed field diagnosis, Huang et al. Of the University of Alabama in the United States proposed a small, slow, in-situ sensing pinprick test method, which can be applied to 3 Ah soft-pack batteries to separate the process of ISC and thermal runaway.The details of ISC can be provided at the level of individual pole piece and current collector. Multiple in-situ temperature peaks are observed within 100 s before thermal runaway. These initial peaks exceed the safety limit, but the temperature drops rapidly after each peak without causing thermal runaway immediately^[63]. Further study shows that the initial temperature peak occurs when the spike tip reaches the aluminum current collector, forming a low-resistance ISC between the negative electrode and the aluminum current collector. The rapid drop in temperature after each peak is attributed to the sudden drop in ISC current, which is mainly due to the cracking of the aluminum foil and the increase in contact resistance.

Because metal pins conduct both heat and electricity, needling is not an effective way to study material response and failure mechanisms. Zhu et al., Oak Ridge National Laboratory, USA, conducted mechanical indentation tests on commercial lithium-ion batteries with different capacities and SOCs, and systematically studied the external characteristics and internal structural changes of batteries when ISC occurred by in-situ and ex-situ methods. X-ray computed tomography (XCT) results showed that.ISC is the combined result of shear banding or other strain localization modes in the electrode assembly, shear offset in the electrode particle coating, and concomitant ductile fracture in the current collector,It is proposed that the irregular strain localization pattern, the mismatch of the mechanical properties of different layers, and the geometric characteristics of the indenter eventually lead to the tearing/puncturing of the battery separator at different locations^[64]. Lee et al. Of Konkuk University, Korea, proposed a mechanical-electrochemical-thermal two-way nonlinear coupling analysis method to analyze the internal short circuit caused by quasi-static indentation and predict the voltage drop and temperature rise^[65]. This method uses the material nonlinearity of battery components to calculate the internal short circuit caused by mechanical deformation and positive and negative electrode contact, taking into account the detail layer. The heat generated by the internal short circuit is calculated by the electrochemical model composed of Randles circuits. The heat source in the electrochemical model is coupled to the thermal model to calculate the temperature variation with time. The voltage drop and temperature rise predicted by the model are in good agreement with the experimental results of 3. 2 Ah soft-pack battery with three different diameters.

Overcharge and overdischarge are the most common battery system failures, which seriously affect the safety and service life of lithium-ion batteries. Generally, charging system failure, unreasonable design of battery management system (BMS), inconsistency of single unit in battery system and loose connection will lead to electrical abuse of battery system.

Overcharge refers to the excessive continuous charging of the battery above the operating voltage or by applying a voltage higher than the specified charging voltage, which is one of the most common faults of the battery system. If there is no appropriate response, it may lead to catastrophic consequences such as fire and explosion. When the voltage continues to rise, the NCM cathode material will be destroyed and decomposed under high voltage conditions (> 4.5 V). As the voltage continues to rise (4.9 ~ 5 V), the traditional electrolyte (1 M LiPF₆/EC∶DEC∶DMC=1∶1∶1) will be oxidized to produce alkanes such as CH₄; Due to the limited capacity of the graphite negative electrode, too many lithium ions can not be fully embedded in the negative electrode, and the lithium evolution reaction begins. At the same time, excessive lithium removal will also lead to the structural collapse of the positive electrode, resulting in a slight drop in voltage and an increase in the internal resistance of the battery. As the overcharge continues, the accumulation of Joule heat is more serious. When the temperature is higher than 60 ℃, the excessive temperature will aggravate various side reactions inside the battery (such as electrolyte oxidation, reaction between delithiated positive electrode and electrolyte, reaction between electrolyte and lithium, etc.), and produce various gases (CO, CO₂, O₂, H₂, and C₂H₄, etc.); The exothermic side reaction will further aggravate the temperature rise of the battery, leading to the decomposition of the SEI film. With the continuous accumulation of heat, when the battery temperature reaches the melting temperature of the separator, a large area of internal short circuit will occur, which will lead to the triggering of thermal runaway^[66⇓~68].

Scholars at home and abroad have carried out a large number of theoretical and experimental studies on the overcharge failure characteristics of lithium-ion batteries. Ren et al. Of Tsinghua University evaluated the overcharge performance of commercial soft-pack batteries with Li_xNi_1/3Co_1/3Mn_1/3O₂-Li_yMn₂O₄ composite cathode and graphite anode under different test conditions, taking into account the factors such as charge current, confinement plate and heat dissipation, characterized the electrode materials under different overcharge conditions, and identified the side reactions inside the battery^[69]. Electrolyte decomposition, transition metal dissolution and phase transformation occur in the overcharged positive electrode, but there is no obvious exothermic behavior before thermal runaway occurs, while lithium precipitation on the negative electrode accelerates the thermal runaway process caused by overcharge. Liu et al. from the University of Science and Technology of China used incremental capacity method and electrochemical impedance spectroscopy to qualitatively and quantitatively study the aging behavior and mechanism of lithium-ion batteries under slight overcharge cycles, and found that slight overcharge would lead to the loss of active materials in batteries and accelerate the aging of batteries^[70]. The thermal stability of the aged battery was studied by ARC. The results show that the stability of the aged battery becomes worse. Lithium evolution from the negative electrode caused by overcharge plays a key role in the change of thermal stability. Togasaki et al. of Waseda University in Japan studied the cycle life of the tandem lithium-ion battery module under normal or overcharge conditions, and analyzed its decay behavior by electrochemical impedance spectroscopy (EIS) and differential pressure analysis (DVA). It was found that the state of charge of the battery was greater than 105% during the cycle, and only the third-order capacity decay was observed in the module under overcharge conditions^[71]. The capacity of the module of a battery without overcharge or an overcharged battery with SOC < 103% decays with the square root of the number of cycles. EIS and DVA analysis confirmed that the deterioration of overcharged batteries with SOC ≥ 105% in the module was more serious than that of other batteries, especially at the positive electrode. The study reveals the potential risks of using series batteries with a wide range of SOCs and provides important insights for safe operation without balanced circuits.

Overdischarge is another common electrical abuse condition. Overdischarge usually does not cause a thermal runaway accident of the battery, but it can cause irreversible capacity loss of the battery^[72]. Generally, the overdischarge process of lithium-ion battery can be divided into three stages: with the continuous discharge of lithium-ion battery, the two processes of delithiation of negative electrode and lithiation of positive electrode proceed at the same time, and the voltage of battery decreases continuously^[73]; When the voltage is too low, the dissolution reaction of the copper current collector begins, so the negative electrode enters the plateau phase of the electrochemical reaction, and the dissolved copper ions can enter the electrolyte, pass through the separator, and deposit on the positive electrode; At the same time, excessive delithiation of the negative electrode will cause the SEI film to decompose and produce gas, and if charged again, a new SEI film will be formed. With the increase of copper ion deposition, internal short circuit is gradually formed, which may lead to serious thermal runaway. It can be seen that the dissolution of copper current collector is the most important side reaction under overdischarge abuse conditions. In addition, during the over-discharge process, the Li⁺ in the lithiated anode will continuously move to the positive electrode, which eventually leads to the formation of holes in the porous anode and the saturation of the positive electrode material, resulting in the structural collapse of the electrode material^[74].

Guo et al. Of Tsinghua University studied the overdischarge process of large-capacity lithium-ion batteries by discharging them to -100% SOC^[73]. A significant voltage plateau was observed at approximately -12% SOC, and an internal short circuit was detected after the cell was overdischarged as it passed through the plateau. The SEM and XRD results show that the internal short circuit caused by overdischarge is caused by the deposition of Cu on the electrode, indicating that there may be Cu dissolution near -12% SOC on the voltage plateau. Ouyang et al. Of the University of Science and Technology of China studied the decay behavior of lithium-ion batteries during over-discharge cycles at different cut-off voltages and the effect of high temperature environment on their decay through experiments^[75]. In the process of overdischarge, the battery showed severe electrothermal behavior, the voltage and current decreased sharply, and the surface temperature increased sharply. After overdischarge, the discharge capacity, energy density, internal resistance and other parameters are improved. In the process of over-discharge cycle, the rated capacity is poor and the capacity degradation is serious, which is further reflected in the changes of battery surface temperature, charge-discharge voltage and internal resistance. In addition, the electrothermal parameters of the battery, such as temperature rise and internal resistance, increase exponentially with the deepening of overdischarge.

Since the commercialization of lithium-ion batteries, many researchers have combined numerical simulation techniques with experimental methods to study their heat generation characteristics. The thermal model widely used at present was established by Bernardi in 1985. Based on the basic principle of energy conservation, the electrode process was studied and the heat equation was derived. The model equation is as follows:^[7]

Where I is the battery current and V is the battery volume; E_OC is the cell equilibrium potential, U is the cell operating voltage, and T is the cell operating temperature. Based on Bernardi model, A1-Hallaj, Wu and Jeon established one-dimensional, two-dimensional and three-dimensional models respectively to calculate the temperature distribution of specific types of lithium-ion batteries, and the simulation results are in good agreement with the experimental results^[76][77][78].

The Bernardi equation is used to calculate the internal heat production of the battery, which has the advantages of saving time and high efficiency, but it ignores the detailed electrochemical process and assumes that the heat production is uniform, which is not accurate in fact, especially for the single battery. However, for the battery management system, the time required to consider the detailed electrochemical process will exceed the actual demand, and the Bernardi equation can reduce the necessary calculation time and get reasonable results. Kim et al. Improved the method, defined the heat generation inhomogeneity, and established a 2D model to successfully simulate the thermal behavior of the battery^[79].

The electrochemical-thermal coupling model is another improvement of the heat generation model. It relies on the porous electrode model, combined with the energy conservation equation of Newman and Pals, to simulate and predict the detailed electrochemical processes and thermal behavior inside the battery^{[80,81][82,83]}.

Lithium-ion battery will produce a large amount of heat in the process of thermal runaway, which will accelerate the rise of battery temperature. Although the battery is made of materials with different thermal stability, its heat release is very concentrated, so the battery can be regarded as a whole and the temperature can be balanced. Most chemical reactions are affected by the reaction temperature and the concentration of reactants, so is the thermal runaway of the battery. When the temperature rises, the thermal runaway reaction will be more intense, which can be described by the Arrhenius formula. Arrhenius formula is the basis for the study of temperature-dependent chemical reactions and reaction kinetics, and is widely used in the study of thermal runaway of batteries^[93].

The basic expression of the Arrhenius formula is as follows, which describes the relationship between the chemical reaction concentration C, the reaction temperature T, as well as the reaction rate \kappa.

Where A is the pre-exponential factor (also called frequency factor), R⁰ is the molar gas constant (or called ideal gas constant), and E_a is the apparent activation energy. In order to describe the reaction kinetics more accurately, it is necessary to accurately obtain the Kinetic triplet: f (C), E_a, and A. A common method to obtain kinetic triplets is to fit these parameters based on DSC and TGA tests. At present, the commonly used methods for kinetic analysis of DSC data are Ozawa method and Kissinger method.

Hatchard et al. First established a lumped thermal runaway model based on the material side reactions in the LCO/Gr battery, and successfully predicted the thermal runaway onset temperature under the hot box test. The material side reactions considered included the decomposition heat of the SEI film, the reaction heat between graphite lithium intercalation and the electrolyte caused by the decomposition of the SEI film, and the reaction heat between LCO and the electrolyte^[94]. In order to further understand the thermal runaway behavior of the battery, Kim et al. Extended it to three dimensions according to Hatchard's modeling method, and the three-dimensional model realized the visualization of the non-uniform temperature distribution inside the battery and the abuse reaction transfer.Kim's mechanism model includes the decomposition heat of SEI film, the reaction heat of graphite intercalation lithium and electrolyte caused by the decomposition of SEI film, the reaction heat of LCO and electrolyte, and the decomposition heat of electrolyte, which has been used up to now. The model equation is as follows:^[95]

Where R_i(s^-1) is the reaction parameter of each side reaction, and A_i(s^-1), E_a,i(J·mol^-1), and m_i are the pre-exponential factor, activation energy, and reaction order corresponding to each side reaction, respectively; c_sei is the dimensionless number of the metastable species in the SEI film; c_neg is the dimensionless number of lithium intercalation in the negative electrode; c_e is the dimensionless concentration of electrolyte; The \alpha is the conversion rate.

In recent years, researchers have continuously optimized Kim's thermal runaway mechanism model, adding the melting heat of the diaphragm, the decomposition heat of PVDF and the heat released by internal short circuit on the basis of the above four side reactions. Although the Kim mechanism model is widely used, it does not consider the internal thermal runaway reaction mechanism of different batteries. Ren et al. Of Tsinghua University proposed a new scheme to establish a reliable battery thermal runaway model based on the dynamic analysis of battery components, and the research process is shown in Figure 12^[96]. Firstly, according to the ARC test values of some batteries and the DSC data of each battery component combination, the heat production of the battery component combination and the thermal runaway characteristic value of the whole battery are jointly analyzed, and the main exothermic reaction is selected as the heat source input of the model.Including the decomposition of SEI film, the reaction of cathode and electrolyte, the reaction of cathode and binder, the oxygen release of anode and cathode reaction, the oxygen release of anode and binder reaction, and the complete decomposition reaction of anode; Then, the component combinations corresponding to the main exothermic reactions were tested by four DSC tests with variable heating rates, and the kinetic parameters were fitted according to the results. The Kissinger method was applied to determine the pre-exponential factor A_x and activation energy {E_{\mathrm{a}}}_{,x} of some exothermic reactions in advance without considering the reaction mechanism function. In DSC experiments at variable heating rates, the variation of peak temperature with heating rate was fitted to the Kissinger equation as follows:

Where β_i is the heating rate; The {T_{p,}}_i is the peak temperature; Experiments with u corresponding to different heating rates. The activation energy {E_{\mathrm{a}}}_{,x} was obtained from the slope fitting, and the pre-exponential factor A_x was obtained from the intercept on the line by plotting the dot plot with the abscissa as 1/T_p and the ordinate as ln(β_i/ {T_p}^2).

Nonlinear fitting methods such as genetic algorithm were used to identify the kinetic parameters [A_x,E_a,x,n_x,ΔH_x] by minimizing the root mean square error to obtain the optimal combination, which was verified with DSC heat flow curve data.

Finally, the chemical reaction heat generation model of battery thermal runaway is established based on the thermal runaway mechanism through the Arrhenius formula, and the model results are in good agreement with the adiabatic thermal runaway experimental results and hot box experimental results of 24 Ah lithium-ion battery.

Wang et al. Also used this scheme to determine the main exothermic reactions and reaction sequence in the thermal runaway process of NCM811/SiC battery, and extracted the detailed thermodynamic parameters of each exothermic reaction from the experimental results through material thermal analysis^[97]. Based on the interaction sequence and material dynamics, a battery thermal runaway model was established to accurately predict the thermal runaway behavior of large-size NCM811/SiC battery. Finally, the thermal interaction process of the material was quantitatively determined by the model, and the results show that the thermal interaction between SiC-electrolyte, NCM811-electrolyte and NCM811-SiC can lead to the increase of the maximum temperature by 318, 222 and 174 ℃, respectively.

Due to the variety of lithium-ion battery material systems, the complexity of the structure and the difficulty of obtaining the parameters, the current battery thermal runaway prediction model is not universal. At present, the prediction of battery behavior is based on experimental data and simulation results, which are fitted into a functional relationship, close to the actual results, but rely on experiments, and do not consider the chemical changes in the battery. Future work on thermal runaway prediction should start with the side reactions when thermal runaway occurs, and predict the thermal behavior of batteries from the chemical reaction level.

In order to improve the thermal properties of cathode materials, a lot of work has been done, mainly by element doping and surface coating^[98].

The O₂ produced by the side reaction between intercalation cathode and electrolyte can lead to microstructural defects, such as the formation of pores in the battery, which has a serious impact on the electrochemical performance and safety of the battery. Element doping can form a stable crystal structure to effectively improve the thermal properties of layered oxide materials^[99]. Researchers have partially replaced Ni or Mn with cationic metals such as Co, Mn, and Mg in LiNiO₂ or Li_1.05Mn_1.95O₄ to improve the thermal properties^[100]. Zhou et Al., Dalhousie University, Canada, partially replaced Co with Ni and Al in the LiNi_1/3Mn_1/3Co_1/3O₂ to form LiNi_0.4Mn_0.33Co_0.13Al_0.13O₂,Al doping, which improved the thermal properties of the material^[101]. Professor Cui Yi of Stanford University reported that doping Ni, Mn and other alloy elements in LiCoO₂ can significantly increase the initial temperature of decomposition and avoid side reactions at high temperature^[99].

The second way to improve the thermal performance of the cathode material is to introduce an interfacial protective layer on its outer surface, which mainly protects the cathode surface from direct contact with the electrolyte, thereby preventing side reactions, phase transitions, enhancing structural stability, and reducing the disorder of cations in the crystal sites^[5]. Therefore, the addition of the protective layer can reduce the generation of side reaction heat. The coating material of the positive electrode can be either a chemical material or an inert thermal material. Chemically inert materials mainly include inorganic MgO, Mg₂TiO₄, ZnO, ZrO₂,ZrF_x 、ZrP₂O₇ 、Li₂ZrO₃ 、SiO₂ 、SnO₂ 、TiO₂ 、TiP₂O₇ 、NaAlO₂ 、UAl₂O₃ UNAlPO₄AlF₃, U FeF₃, and organic film (polydimethyldiallylammonium chloride). Fluorine-containing protective layer materials have good performance due to their inertness, which reduces the generation of heat. For example, the thermal runaway onset temperature of AlF₃-NCM is delayed by 20 ° C^[102]. Similarly, solid oxide materials tend to be better in protecting the positive electrode. The coating of the SiO₂ inhibits the side reaction between the electrode and the electrolyte and improves the cycle performance of the LiNi_0.6Co_0.2Mn_0.2O₂ at high temperature; It also shows obvious HF scavenging ability, thus improving the capacity retention rate during the cycle. Direct contact of the electrolyte with the highly unstable oxidized positive electrode is prevented, improving thermal stability^[103]. Phosphate coating material is a promising coating material because of its strong covalent bond, which can also improve the thermal stability of the material. For example, NCA811 coated with Ni₃(PO₄)₂ was cycled at 55 ° C. The results show that the chemical attack of HF on the coated cathode is reduced, while the surface of the uncoated pristine cathode is jagged^[104].

Thermosensitive coating materials are currently promising cathode protection materials, in which the protective layer materials have a positive temperature coefficient, which can stop electrochemical reactions and side reactions, and close the battery circuit at high temperatures through polymerization. Conjugated polymers poly (3-octylthiophene) (P3OT) and poly (3-decylthiophene) (P3DT) are commonly used as thermal inert materials. It is worth noting that the better thermal performance of these thermosensitive materials has only been reported in LCO. The P3OT used by Ji et Al. Of Wuhan University intercalates between Al and LiCoO₂ layers to form a sandwich-like structure Al/P3OT/LiCoO₂(LCO-PTC), which shows strong positive temperature coefficient behavior and can shut down the electrochemical reaction between 90 and 100 ℃ before thermal runaway^[105]. The battery using the LCO-PTC positive electrode only showed a slow decrease in the open-circuit voltage in the hot box test at 150 ° C, demonstrating high thermal stability and safety. Xia et al. Suggest that the electrochemical reaction can be stopped at a temperature of about 110 ° C for the LiCoO₂ with a P3DT protective layer^[106]. The P3DT modified separator has an automatic overcharge protection mechanism and can be used in LiFePO₄ batteries.

Electrode zoning is also considered a promising technique to protect electrodes from momentary internal short circuits. In this technology, the battery pole piece is designed in the form of a grid-like slit, and after the corresponding grid area is impacted, it breaks off before the diaphragm breaks down, thus producing electrical isolation. The exciting innovation is that the battery with grid-like slit electrodes can still work properly before mechanical shock. Each grid region of the zoned cell is completely isolated from the rest, but shares the separator and electrolyte. After the impact, the voltage of the shedding part drops, while the remaining part is basically unaffected^[107]. These strategies open up new ways to suppress thermal runaway, but the resulting side reactions need to be further studied.

The SEI film plays a vital role in the performance of the anode. Its thermal decomposition deteriorates the safety of the negative electrode and the battery system as a whole. Therefore, a recent goal is to develop artificial SEI membranes to improve the mechanical and thermal properties of SEI membranes. Mild oxidation, metal deposition, and polymerization (or coating) are commonly used techniques to improve the SEI film. Ding et al. Of Pacific Northwest National Laboratory found that the graphite anode coated with AlF₃ had higher initial discharge capacity, better cycle life, higher capacity retention and excellent rate capability compared with the uncoated graphite anode^[108]. This is due to the formation of a more stable and conductive SEI film on the AlF₃ coated graphite particles, which improves the performance of the graphite anode. Li₄Ti₅O₁₂(LTO) is a promising coating material for artificial SEI films. Such as LTO-coated medium carbon microsphere (LTO-MCMB), LTO-coated graphite (LTO- Gr), LTO- coated carbon microsphere (LTO-CMB) composite, LTO coated carbon nanotube (LTO-CNTs), composite of LTO and multilayer carbon nanotube (CNTs) co-doped with nitrogen (N) and boron (B) (N-B-C-LTO) and LTO/carbon nanotube/graphene (LTO-CNT-G)^{[109][110][111][112][113][114]}. In conclusion, the SEI film plays an important role in the electrochemical performance of the anode. Therefore, improving the mechanical and thermal properties of the artificial SEI film can improve the thermal safety performance of the negative electrode and the battery as a whole^[98].

Separator is an important factor affecting the performance and safety of lithium-ion batteries. Conventional diaphragm materials include fibers, gels, or polymers, with polymers being the most commonly used diaphragm materials. The unique characteristics of polymers used in lithium-ion batteries include: ultra-thin thickness (8 ~ 25 μm), which can provide lower internal resistance, higher current, power and energy density; Small pore size (< 1 μm) can block the penetration of electrode active materials and conductive additives; Sufficient porosity (40% ~ 60%) can store enough liquid electrolyte to provide sufficient ionic conductivity; Has excellent mechanical and chemical stability^[115].

In lithium-ion batteries, the separator prevents direct electrical connection between electrodes by providing mechanical separation and high electrical resistance. However, there is a small leakage current, called self-discharge. Self-discharge varies with battery chemistry, age, cycle time, and storage temperature. On the other hand, the separator, due to its porous structure, can retain enough electrolyte and provide sufficient ionic conductivity to be a medium for Li⁺ transfer between electrodes. When the cell temperature rises near the melting point of the separator, the pores of the separator tend to close, a process called separator pore closure. The temperature at which septal closure is triggered depends on the melting point of the septum^[116]. During the pore closure of the separator, the ionic conductivity and electrochemical reaction are delayed, and the cell temperature does not drop sharply. Therefore, the separator must maintain its mechanical integrity, otherwise it will cause internal short circuits, generate huge thermal energy, and eventually lead to thermal runaway of the battery. Therefore, reducing the contraction of the diaphragm and delaying the collapse at high temperature is the goal of improving the stability of the diaphragm.

Three-layer polymer composite membranes (PE layer sandwiched between two PP layers, PP/PE/PP) have been commercialized, and this design extends the time between pore closure and contraction of the membrane. At temperatures around 135 ° C, the PE partially melts and the septum closes the pores, while the PP maintains mechanical integrity to avoid internal short circuits. However, the temperature interval of PP/PE/PP is smaller, that is, the melting temperature is similar to that of PP^[3]. Alternative materials with high thermal stability, such as ceramics, polyethylene phthalate (PET), polyimide (PI), non-woven fabric membrane, polymer electrolyte that can be used as separator, are still under research, but there are problems of high cost and low production efficiency^{[117][118][119][120][121]}. Therefore, many researchers have focused on the development of high thermal stability separators. Long et al. Of Sichuan University rationally designed a high temperature and fire-resistant nano-CaCO₃-based composite film, which has superior thermal stability (almost no shrinkage at 300 ℃), excellent non-flammability and extremely low heat release (equivalent to 11% of PP)^[122]. In addition, the alkaline nano CaCO₃ can neutralize the hydrofluoric acid inevitably existing in the LiPF₆-based electrolyte and ensure the long-term stability of the interface. Min et al. From Shenzhen University proposed a method to prepare a new separator by introducing a covalent organic framework (COF) into poly (aryl ether benzimidazole) (OPBI), and compared with polypropylene separator, OPBI @ COF separator has the characteristics of high electrolyte absorption rate (428%), high ionic conductivity (1.214 mS·cm^-1), good thermal stability, and good flame retardancy^[123]. The discharge specific capacity of the LiFePO₄/Li battery assembled with the OPBI @ COF20 separator is 98.98% after 200 cycles at 148.3 mAh·g^-1,0.5 C, and the OPBI @ COF separator can inhibit the growth of lithium dendrite and ensure the safety of the battery after a long time of operation.

Thermal protection materials such as paraffin microspheres can be used to protect the negative electrode and separator, creating a protective polymer layer between the negative electrode and separator and permanently shutting down the battery when the battery temperature reaches a critical value^[124]. The separator with thermally triggered flame-retardant behavior can enclose the flame retardant inside the polymer separator to avoid direct decomposition into the electrolyte, which adversely affects the battery performance^[125]. When the temperature of the lithium-ion battery rises, the polymer separator will melt and release the flame retardant, which will eventually prevent the combustion of the flammable electrolyte. Fig. 13 (a) shows a new type of electrospun core-shell microfiber separator with flame retardant properties manufactured by Professor Cui Yi of Stanford University^[126]. The ultrafine fiber has a core-shell structure, the core is triphenyl phosphate (TPP), the flame retardant of organophosphorus, and the shell is polyvinylidene fluoride-hexafluoropropylene (PVDF-HFP). Encapsulating TPP inside the PVDF-HFP protective polymer shell can prevent TPP from being directly exposed to the electrolyte and largely slow down its dissolution, inhibiting the negative impact of TPP on the electrochemical performance of the battery. In addition, if thermal runaway occurs, the PVDF-HFP polymer shell will melt with the increase of temperature, and the encapsulated TPP flame retardant will be released into the electrolyte, thus effectively inhibiting the combustion of highly flammable electrolyte. Subsequently, the members of the group proposed a new electrochemically stable separator by coating the surface of a commercial polyolefin separator with an electrolyte-insoluble flame retardant additive^[127]. Two materials were selected to extinguish the flame by the mechanism of trapping free radicals, combined with dense Sb₂Br₃ smoke, eliminating oxygen. The new composite separator has a two-layer structure: a dual component composed of an insoluble flame retardant and a commercial polyolefin as a backbone, as shown in Figure 13 B.

At present, the research on high safety battery separators mainly focuses on: adding inorganic/organic materials on the surface of the separator to improve its thermal stability and wettability; The chemical functional group of the polymer is optimized, and the mechanical strength and the chemical stability are improved; Inhibiting lithium dendrite; Development of diaphragms with thermally triggered flame retardant behavior, etc. Continuous research and collaborative improvement are needed to ensure the safety of batteries.

As the blood of lithium-ion batteries, electrolyte almost participates in the chain reaction of thermal runaway, which plays a decisive role in the safety performance of batteries. At present, although the combination of organic solvents and lithium salts in the electrolyte achieves the best balance in terms of electrochemical performance, cost and toxicity, it is not safe enough.

At present, the main ways to improve the safety of electrolyte from lithium salts and solvents include: using stable lithium salts and flame retardant additives to reduce the risk of combustion; The overcharge additive is used to inhibit the continuous increase of the overcharge voltage; Use non-flammable electrolyte solvents or develop aqueous electrolytes; Synthetic polymers and solid electrolytes solve the problems of leakage, internal short circuits, and flammability^[13].

The most common way to prevent thermal runaway is to add flame retardant components to the solvent or to completely discard the flammable solvent. Considering the complexity of the whole electrolyte system, the replacement of solvent may lead to the failure of complete electrochemical performance. Preliminary studies have focused on the use of a small amount of flame retardant additives. The flame retardant additives of the electrolyte studied so far mainly prevent the propagation of combustion by trapping the active free radicals and acids produced by the combustion reaction. According to the flame retardant mechanism and element types, the flame retardant additives studied are mainly divided into four categories: phosphorus-containing flame retardant additives, fluorine-containing flame retardant additives, ionic liquid additives and composite flame retardant additives^[13]. The different flame retardant properties and electrochemical effects of these different types of flame retardant additives are mainly due to the different flame retardant free radicals and chemical structures. Mao et al. Of Shanghai Space Power Research Institute used Tris (2,2,2-trifluoroethyl) phosphite (TTFP), a phosphate additive, to construct a non-combustible electrolyte. The optimized electrolyte can improve the performance of the battery, and under the condition of equivalent electrochemical performance, the battery has thermal runaway at higher temperature and better thermal stability^[128]. On the other hand, the self-discharge performance of the battery with the optimized electrolyte is low. Wang et al. Of the University of Science and Technology of China conducted a comparative study on the effects of trivalent phosphorus flame retardants and pentavalent phosphorus flame retardants on the safety and electrochemical performance of lithium-ion batteries^[129]. The results show that the pentavalent phosphorus electrolyte has a wider electrochemical window and higher withstand voltage than the trivalent phosphorus electrolyte, indicating that the pentavalent phosphorus flame retardant is more suitable as an efficient additive for the cathode material system. At the same time, the trivalent phosphorus flame retardant can be used as a solid electrolyte interface layer forming agent, which can decompose it even under low load, and promote the formation of a stable solid electrolyte interface layer on the surface of the graphite anode material. Halogen-containing flame retardants used in lithium-ion batteries are mainly organic fluorides. Xia et al., Ningbo Research Institute, Chinese Academy of Sciences, proposed a non-flammable electrolyte with 1,1,1,3,3,3-hexafluoroisopropyl methyl ether (HFPM) as a co-solvent, which can maintain 82% of the capacity after 200 cycles at a high cut-off voltage of 4.9 V^[130]. Compared with phosphate flame retardants, fluorine flame retardant additives are better in maintaining electrochemical performance. Ionic liquids generally refer to liquid salts that consist entirely of anions and cations at room temperature. Therefore, ionic liquid electrolytes are expected to replace traditional organic electrolytes and improve the safety of lithium-ion batteries. Trifluoromethylsulfonimide (TFSI) has been widely studied due to its good electrochemical performance and thermal stability.

Highly concentrated electrolytes and partially highly concentrated electrolytes, which generally exhibit good electrochemical performance and thermal stability, are also considered to have potential in improving the safety of lithium-ion batteries due to their low flammability. Many basic studies have shown that the unique solution structure of highly concentrated electrolytes, in which all solvent molecules and even anions participate in the solvation sheath, makes them have different properties compared with traditional electrolytes of the same composition. Although the high thermal stability of high concentrated electrolyte has been widely proved by ignition test and thermogravimetric analysis (TGA), the safety of batteries using high concentrated electrolyte has not been evaluated. Hou et al. Of Tsinghua University selected two high concentrated electrolytes, LiFSI/DMC (molar ratio 1: 1.9) and LiFSI/TMP (molar ratio 1: 1.9).The safety performance of NCM811/Gr and NCM523/Gr batteries was evaluated, and the failure mechanism of materials and monomers was systematically studied. As shown in Fig. 14, it was found that the high concentration electrolyte based on LiN(SO₂F)₂ could not solve the intrinsic safety problem of lithium-ion batteries, and the huge heat of battery triggering heat runaway was generated by the reaction between lithium graphite and LiN(SO₂F)₂^{[131⇓~133][134]}. Zhang et al of Pacific Northwest National Laboratory reported that the local high concentration electrolyte with non-solvated diluent can maintain the similar Li⁺ solvation structure with high concentration electrolyte, which has become the most popular research topic at present, but it has only been tested at the ignition and material levels^[135⇓~137]. Wu et al. Of Tsinghua University first reported the safety performance of NCM811/Gr-SiO soft-pack battery with non-combustible local high concentration electrolyte (1 M LiFSI/FEC: TEP: BTFE), which can effectively stabilize the interface of NMC811 positive electrode and SiO negative electrode under high operating voltage.Compared with the reference electrolyte, the battery with non-combustible local high concentration electrolyte has higher T₁, higher 4.4℃,T₂, lower 47.3℃,T₃ by 71. 8 ℃, longer time from self-heat generation to thermal runaway triggering by about 8 H, and superior high-pressure cycling stability^[138].

The addition of appropriate additives to the electrolyte can significantly improve the thermal stability of the battery, but the leakage of the electrolyte will still cause potential safety hazards. Resolving battery failures, such as leaks and internal shorts, requires further development of solid-state electrolytes. Hoang et al. Of Sungkywan University in Korea proposed an optimized design of a high-voltage stable solid-state lithium battery system by combining a solid-state electrolyte composed of PVA-g-PCA and HTpca nanoconductors with a Al₂O₃ coating LCO active material^[139]. The PVA-g-PCA-80IL-5 HTpca solid state electrolyte has good ionic conductivity at room temperature with a (4.78·10^-3S·cm^-1),Li⁺ transfer number of 0.72 and a wide electrochemical window of up to 5.17 V vs Li⁺/Li. Cells cycled 500 times against a metallic Li negative electrode using a PVA-g-PCA-60IL-5HTpca prepared positive electrode maintained long-term stability and low interfacial resistance. As a conductive binder, PVA-g-PCA-60IL-5HTpca ensures the dense filling and uniform distribution of LCO particles, as well as the excellent adhesion strength (> 24 MPa) with the current collector. The solid-state lithium battery assembled with the LCO active material coated with Al₂O₃ has a specific capacity of 188.7 mAh·g^-1 and can be cycled more than 200 times at 4.5 V and 60 ° C. Chen et al., Institute of Physics, Chinese Academy of Sciences, combined with EIS and X-CT analysis, found that lithium metal can penetrate into the defect sites of LATP bulk phase at high temperature^[140]. The high reactivity of lithium and LATP at the defect site leads to higher interfacial reactivity and earlier thermal runaway. The thermal runaway can be significantly delayed by adding LiPO₂F₂ to modify the defect sites (atomic structural defects, cracks, voids, etc.) of LATP particles and inhibit the Li/SE interfacial reaction. The key to the development of solid electrolyte is to give full play to the advantages of high safety, improve technology and reduce costs.

The role of current collectors in batteries is to provide electron conduction pathways for the positive and negative electrodes, and metals such as aluminum and copper are commonly used, which are electrochemically stable within the potential window, thus preventing degradation. However, the joining of the two most conductive materials within the cell causes the most severe internal short circuit and is one of the conditions for thermal runaway to occur. Therefore, removing or isolating the current collector during an internal short circuit can significantly reduce the risk of thermal runaway triggering and propagation.

Pham et al., University of London, reported a metal-coated polymer current collector^[141]. In the pin-prick test, the cell with the aluminum-coated polymer current collector was 100% successful in preventing thermal runaway, keeping the cell voltage > 4 V, while the comparison cell eventually developed thermal runaway. Professor Cui Yi's team developed a lightweight polyimide-based current collector with a thickness of about 9 μm to replace the existing metal current collector^[142]. The novel current collector has a sandwich structure, in which the middle layer is an organic supporting membrane embedded with a flame retardant, and the outer layer is a metal layer with a thickness of about 500 nm. The metal layer with the thickness can achieve the same conductivity as the metal foil current collector in the battery, and the polymer substrate has good thermal stability and mechanical strength. In the case of thermal runaway, based on the different temperature coefficient between the metal layer and the organic matter, the flame retardant is released, which can effectively prevent the battery from continuing to burn.

To sum up, the future research and design of lithium-ion batteries will mainly focus on the development of high thermal stability of positive electrode, high potential of negative electrode, high temperature resistance of separator, low flammability of electrolyte, solid state of electrolyte and multi-functional current collector.