A Review on the Thermal Hazards of the Lithium-Ion Battery and the Corresponding Countermeasures

As one of the most promising new energy sources, the lithium-ion battery (LIB) and its associated safety concerns have attracted great research interest. Herein, a comprehensive review on the thermal hazards of LIBs and the corresponding countermeasures is provided. In general, the thermal hazards of the LIB can be caused or aggravated by several factors including physical, electrical and thermal factors, manufacturing defect and even battery aging. Due to the activity and combustibility of traditional battery components, they usually possess a relatively high thermal hazard and a series of side reactions between electrodes and electrolytes may occur under abusive conditions, which would further lead to the thermal failure of LIBs. Besides, the thermal hazards generally manifest as the thermal runaway behaviors such as high-temperature, ejection, combustion, explosion and toxic gases for a single battery, and it can even evolve to thermal failure propagation within a battery pack. To decrease these hazards, some countermeasures are reviewed including the application of safety devices, fire-retardant additives, battery management systems, hazard warnings and firefighting should a hazard occur.


Introduction
In light of the steadily increasing energy demands and the consensus regarding the reduction of pollution, humans have paid great attention to the development of new energy such as solar energy, wind energy, tidal energy, lithium-ion battery (LIB) and fuel cell [1][2][3][4][5]. Advantages including high energy density, less pollution, stable performance and long-life cycle compared to many alternatives have made LIBs the dominant power sources for electrical applications [6,7]. Especially, with the approaching era of electric vehicles (EVs), it is foreseeable that the use of LIB will be more common in the future.
However, behind the boom, there exist challenges that must be faced. As a result of the high energy density of LIBs, they are sensitive to abusive conditions such as high temperature, crashing, overcharge, over-discharge and short-circuit, etc. [8][9][10]. Moreover, the typical components of a battery, e.g., plastic packing, separator and electrolyte are combustible. Thereby, the accidents induced by LIBs are frequently reported and can occur in a range of applications from mobile telephones, to EVs and even airplanes [11][12][13][14]. Especially with the increasing energy density, e.g., the popularity of LiNi 0.8 Mn 0.1 Co 0.1 (NMC811), the problems will be exacerbated [15,16]. The abusive conditions described above will destroy the original stable structure of battery, which triggers chain reactions inside battery and eventually leads to thermal runaway. Among, thermal runaway of a battery typically manifests as high temperature, gas ejection, violent combustion and even explosion, which will lead to catastrophic results.
Regarding the thermal hazards of the LIB, much research has been directed toward enhancing the inherent safety of battery and improving the thermal management to prevent thermal failure. For the former, it mainly focuses on enhancing the fire-retardant performance of battery components including the cathode, anode, electrolyte, additives and separator [17][18][19][20][21][22][23][24][25][26][27]. As for the latter, it includes the inclusion of battery safety devices, thermal management of LIBs during usage, thermal hazards warning and firefighting should a hazard occur. The thermal management of the LIB can usually be categorized into several types including air, liquid and phase change material (PCM) based systems [28][29][30][31][32][33]. Prior to the occurrence of a thermal hazard, LIBs are prone to experiencing evident increases in temperature and may also release representative gases such as CO, H 2 and SO 2 . Therefore, battery safety devices such as vents can be used to terminate the deterioration of failure, meanwhile a temperature sensor or gas sensor can also be applied to warn to the hazard [34][35][36][37]. Finally, if the thermal hazards are accompanied by severe combustion, it is necessary to fight the fire to impede its progression. Until now, researchers have paid great attention to firefighting methods and extinguishers so that the flame or combustion of the LIB can be extinguished effectively [38][39][40][41].
Although much work has been done on the thermal hazards of the LIB, comprehensive summaries on the thermal hazards of batteries involving battery components, a single battery and a battery pack are scarce. Besides, most researchers focus on the thermal management of battery during normal usage, countermeasures that suppress the thermal hazard require further study. The current work provides a comprehensive review on the thermal hazards of battery, and related thermal hazard prevention techniques. The Section 1 partially summarizes the safety-related LIB incidents that have occurred in recent years. The Section 2 is regarding the common causes of battery thermal hazard and the respective mechanism. The Section 3 demonstrates the thermal hazards of the LIB, involving typical battery components, a single battery and a battery pack. It is worthy to be noted that the influence of low-pressure environment and cathode chemistry on the thermal hazard is also involved. The Section 4 is concerning how to prevent the thermal hazard of a battery and improve its safety which can be divided into inherent safety methods and extra countermeasures. Table 1 lists several representative incidents of LIB failure in recent years which can be divided into three main types including mobile telephone, EV and airplane [13,[42][43][44]. With regard to safety-related battery incidents in mobile telephones, it is apparent that nearly all mainstream mobile telephones have experienced such incidents, e.g., Apple, Samsung and Huawei. Take the Samsung Note 7 as an example, it was released on 3 August 2016, and then its first reported explosion was on August 24, 2016. In the end, quantities of incidents forced Samsung to recall all Note 7 devices around the world on 2 September 2016. This greatly damaged consumer confidence in Samsung, and ultimately resulted in a sharp decrease in its market share and a loss of 17 billion dollars. The root reason of Note 7 incidents was attributed to the battery fault, i.e., Samsung adopted an excessively thin separator to increase the energy density of battery that substantially increased the possibility of the battery short-circuiting. Similar faults also occurred in the other failed batteries. For EVs, most safety-related battery incidents appeared under the conditions of crashing, charging, discharging and self-ignition, which subsequently led to the short-circuit of batteries. Although the failure rate of EV is approximately 1/10,000, which is much lower than that of traditional vehicles (7.6/10,000 [45]), this issue still significantly hinders the development of EVs. In the case of incidents on airplanes, these usually resulted from the failure of passengers' electronic equipment where the battery caught fire, filled the cabin with smoke and led to catastrophic results. These accidents forced civil aviation bureaus around the world to take strict administration regarding portable electronics. Research revealed that battery failures in airplanes

Electrical Factor
External short circuit, overcharge and over-discharge are common conditions of electrical failure. Where, the external short circuit of battery occurs once electrodes with a voltage difference are connected by conductors. It usually results from the deformation of battery, water immersion, conductor aging, improper usage and long-time charging, etc. During an external short circuit, the battery is in a state of fast discharging and the discharging current may be much larger than that of normal condition. Hereafter, the battery undergoes a violent temperature rise which may lead to serious consequence.
As the open circuit voltage of battery is charged above the cut-off voltage, overcharge occurs. The failure of battery management system is the ordinary reason of overcharge so that the charging of battery will proceed continuously. As a result, the internal pressure of battery increases, the deformation of battery and leakage of electrolytes occur, and the performance of battery also significantly decreases. Besides, severe heat and gas generation can also be seen during overcharge process. Compared to normal charging process, the heat generation behavior during overcharge will be much greater as a result of the side reactions inside battery and the increased internal resistance. In addition, excessive loss of lithium ions at the cathode during overcharging will lead to structural collapse of the cathode and subsequent oxygen release. The released oxygen accelerates the decomposition of electrolyte, and thereafter, gases are generated. Consequently, the thermal hazard associated with an overcharged battery is greatly heightened compared to that of a normal battery due to the excessive energy stored in the overcharged battery.
Similarly, when the open circuit voltage of battery is discharged below the cut-off voltage, over-discharge occurs and the failure of battery management system is also a typical reason for over-discharge. Over-discharge causes an excessive loss of lithium ions on the anode which will destroy the stable structure of the anode and cause irreversible damage. Meanwhile, gases such as CO and CO 2 can also be generated, resulting in battery swelling. Furthermore, over-discharge will cause the dissolution of copper collector. The dissolved copper migrates and deposits onto the surface of the anode, which will pierce the separator, causing a short circuit to form. Consistent with the other abusive conditions, substantial heat release also occurs upon over-discharge of the LIB.

Thermal Factor
Besides the overheating caused by physical or electrical factors, thermal failure can also be triggered by external high temperature and overheating. Thermal factor results in a fierce rise of battery temperature, the melting of separator, the decomposition of electrodes/electrolytes and numerous side reactions etc., and finally leads to battery thermal runaway. In other words, thermal abuse is the root cause of battery thermal runaway. It should be noted that both physical and electrical factors also ultimately cause thermal runaway. Further, it can be found that almost all the abusive conditions are accompanied with internal short circuit, i.e., an internal short circuit is the most common feature of thermal runaway. It occurs when the cathode and the anode contact with each other due to the failure of battery separator. Once the internal short circuit is triggered, the electrochemical energy stored inside the battery spontaneously releases, generating large amounts of heat.

Manufacturing Defect and Aging
Apart from the external factors, battery internal defect due to poor manufacturing, such as a low-quality separator, material contaminant, and improperly arranged constituents, can also cause battery failure and result in thermally hazardous conditions [66]. As stated in Section 2, the main cause of incidents in the Samsung Note 7 was a battery fault, in other words, Samsung adopted an excessively thin separator to increase the energy density of battery which raised the possibility of battery short circuit [67]. Mohanty et al. [68] undertook a systematic investigation on several plausible cathode defects such as agglomeration, pinholes, metal particle contamination and non-uniformity. The findings showed that cathode agglomeration aggravated the cycle efficiency and resulted in faster capacity fading. Electrode pinholes showed substantially lower discharge capacities compared to the baseline of electrodes. Metal particle contaminants had an extremely negative effect on performance, and the electrodes with severe non-uniform coatings showed poor cycle life. In summary, cathode material defects decrease battery performance and increase the risk of thermal hazard. In addition, a low-quality separator will reduce the efficiency of Li + passing through the separator and result in serious Li plating, which will further penetrate the separator and induce the internal short circuit [69]. The improper arrangement of constituents is also harmful to the operation of battery, deteriorates the heat generation and heat release, therefore decreases battery safety.
Finally, due to deterioration associated with battery aging, its thermal hazards will increase correspondingly. The aged battery will lose quantities of lithium and active materials [70,71]. Partially inactivated Li + will transfer to Li metal and deposit onto the surface of electrodes and even the separator to form Li plating. Upon the progression of battery aging, the degree of Li plating will increase gradually and subsequently, dendrite is generated. The dendrite may penetrate the separator and form a bridge between the electrodes, which will cause micro-shorting inside the battery and eventually lead to failure of the battery. At the same time, the thickness of the solid electrolyte interface (SEI) layer will gradually increase with the aging of the battery, which will lead to increases in the battery resistance, heat generation and thermal hazards.
To better understand the thermal hazards of LIBs, some aspects regarding typical battery components, a single battery and a battery pack will be reviewed. In addition, the influence of certain parameters such as low pressure and cathode chemistry are also considered.

Thermal Hazards of Typical Battery Components
In general, a LIB is mainly composed of electrodes, electrolytes and a separator. The electrodes can further be divided into the cathode and the anode. The thermal hazard of the LIB usually results from the destructive reactions of battery components such as the decomposition of electrode/electrolyte, the reaction between electrodes, the reaction between electrode and electrolyte, etc. In order to gain a comprehensive understanding on the thermal hazards of the LIB, the hazards of typical battery components will be discussed following.
PF 5 can also react with carbonate solution to form methyl fluoride (CH 3 F) and other substances.
In addition, it is generally regarded that LiBF 4 is superior to LiPF 6 with respect to factors such as high-temperature performance, reaction characteristics with water and chemical stability [77,78]. However, the anion of LiBF 4 is very small and it can be easily combined with lithium ions in solution, resulting in its weak ion transport ability. To overcome this weakness, researchers chelated lithium salt with oxygen to synthesize new large anionic lithium salts, such as lithium diethylborate (LiBOB) and lithium fluoroethylborate (LiDFOB) [79,80].
After a long period of development, some researchers believe that liquid electrolytes have inherent safety disadvantages. No matter how good the electrolyte additives and solutions are, they cannot prevent the formation of lithium dendrites. Therefore, researchers have started to develop solid-phase electrolytes [81][82][83][84].

Thermal Hazards of the Electrodes
Nowadays, lithium nickel manganese cobalt oxide (NMC) ternary positive electrode material is one of the most widely used cathode materials, with high capacity, low cost and relatively good safety, which has been vigorously promoted in the current power battery industry [85]. The researchers found that with the higher content of Ni element, the cost would be significantly reduced, and the specific capacity would be higher [86]. However, the battery cycle life and the stability of electrode and electrolyte would be lower. Therefore, it is important to study the effect of element content in the NMC ternary material. Ma et al. compared the thermal runaway behavior of various NMC cathode materials with electrolytes at high temperature by accelerating calorimeter (ARC), while the results showed that under different cut-off voltages, using a traditional carbonate solution as the electrolyte, with the increase of Ni content, the initial temperature corresponding to the reactions between electrode and electrolyte was reduced, i.e., the fire risk of the LIB increased [16]. The material used in previous research was a common NMC cathode, made from a hydroxide precursor system, with a powder particle size of about 10 µm. The surface was not coated or modified to improve the performance. On the basis of traditional polycrystalline NMC materials, new single-crystal NMC materials have been developed, which are expected to improve battery stability at elevated temperatures [87].
For positive electrode materials, the amount of lithium embedded will directly affect the amount of heat produced. The more lithium that is embedded, the less heat the whole LIB system will generate. If there is only the electrode material without electrolyte, almost no heat will be released [88] and the reactivity between positive materials and different electrolytes will also be different. Wang et al. analyzed the heat production from the reaction between cobalt acid lithium batteries and different electrolytes using a C80 micro-calorimeter, and the results showed that when compared with chain solvent, the annular solvent was more stable. Among the various chain carbonate solutions, DMC showed the highest reactivity, while diethyl carbonate (DEC) presented the lowest [89]. According to the above research, the heat production mechanism of positive electrode with electrolyte can be summarized as follows: The stability of electrode material will decrease and the temperature will increase, resulting in the decomposition of the positive electrode material and oxygen release during the charging process. The oxygen will react with the electrolyte and negative electrode leading to substantial heat release and gas generation, thus greatly increasing the risk of thermal failure.
For negative electrode materials, most of the commercial anode material is carbon such as graphite. On the anode surface, a layer of SEI would be formed during the first charge and discharge cycle. The SEI layer will not affect the lithium ion transport and can also prevent direct contact occurring between the anode and the electrolyte, which reduces the cycle performance. While the existence of the SEI will to some extent influence the capacity of LIBs, and its thermal stability under high temperature is poor. Barnett et al. investigated the graphite anode with organic electrolyte containing LiPF 6 using ARC and differential scanning calorimetry (DSC), and the results indicated that the initial exothermal temperature of the system was about 80 • C, while the starting temperature of rapid heat production was about 150 • C, which was regarded as the thermal runaway temperature of the LIB. At the same time, other scholars have studied the pyrolysis of the SEI. With the increase of temperature, the initial structure of the SEI would be destroyed, thus losing its protective function, leading to the reaction between electrode and electrolyte and the release of heat [90][91][92].

Thermal Hazards of the Separator
At present, the LIB separator is commonly made of polyethylene (PE), polypropylene (PP) or their composite materials. However, because of their low thermal deformation temperature (PE: within 85 • C, PP: within 100 • C) [93], the polyolefin separators will suffer severe thermal contraction when ambient temperature is relatively high, thus it is not suitable for the usage under high temperature environment. In order to meet the requirement of the complex working environment of the LIB, researchers have developed a variety of new composite separator materials and optimized their performance from the aspects such as preparation process improvement [94][95][96]. Among them, a new type of separator material obtained by the composite of polyaryl ether and thermoplastic resin has attracted extensive attention due to its superior ionic conductivity and high safety [97,98]. Zhong et al. developed a series of poly aryl ether materials, including poly aryl ether ketone (PPEK), poly aryl ether sulfone (PPES), poly (phthalazinone ether sulfone ketone) (PPESK) and so on. Their glass transition temperatures were over 260 • C, and the 5% thermal weightlessness temperatures were close to 500 • C. Upon analysis using an electrostatic spinning process, the poly aryl ether materials with polyvinylidene fluoride (PVDF) as a composite separator of a LIB showed excellent electrochemical performance and thermal stability [99,100].
Although the high-temperature resistant separator can significantly improve the safety condition of the LIB, it cannot completely inhibit the occurrence of thermal hazard. According to the previous research, although the mass ratio of the separator is approximately 3%, it can release up to 33% of heat in the combustion process, see Figure 1, indicating that the combustion characteristics of separator directly affect the fire risk of the LIB [76]. Therefore, the research and development of a new separator material with high thermal safety is essential.

Thermal Hazards of a Single Battery and Battery Pack
Due to the high energy density of the LIB and the inherent hazards of battery components described above, it is common for the LIB to experience thermal hazards especially under abusive conditions. For a single battery, the thermal hazards are generally exhibited as high-temperature, ejection, combustion, explosion and toxic gases during thermal runaway. As for a battery pack, thermal failure propagation within the pack can also be observed.
Although the high-temperature resistant separator can significantly improve the safety condition of the LIB, it cannot completely inhibit the occurrence of thermal hazard. According to the previous research, although the mass ratio of the separator is approximately 3%, it can release up to 33% of heat in the combustion process, see Figure 1, indicating that the combustion characteristics of separator directly affect the fire risk of the LIB [76]. Therefore, the research and development of a new separator material with high thermal safety is essential. Figure 1. The combustion heat distribution of lithium-ion battery: the overall combustion heat on the left was determined using experimental results, while the combustion heat of battery components on the right was obtained by the thermodynamic calculations. Adopted from Ref. [76].

Thermal Hazards of a Single Battery and Battery Pack
Due to the high energy density of the LIB and the inherent hazards of battery components described above, it is common for the LIB to experience thermal hazards especially under abusive conditions. For a single battery, the thermal hazards are generally exhibited as high-temperature, ejection, combustion, explosion and toxic gases during thermal runaway. As for a battery pack, thermal failure propagation within the pack can also be observed.

Thermal Hazards of a Single Battery
With the proceeding of chain reactions inside battery during thermal runaway such as the decomposition of SEI layer, the decomposition of electrode and electrolyte, the reactions between electrode and electrolyte, and the combustion of electrolyte [101,102], substantial heat will be

Thermal Hazards of a Single Battery
With the proceeding of chain reactions inside battery during thermal runaway such as the decomposition of SEI layer, the decomposition of electrode and electrolyte, the reactions between electrode and electrolyte, and the combustion of electrolyte [101,102], substantial heat will be generated which leads to the sharp increase of battery temperature. Figure 2 displays the typical curves of battery surface temperature and mass during the thermal runaway process, where the thermal runaway is triggered with a 2.5 kW electronic heater. As observed, the surface temperature of battery increases stably under the effect of the heater. The SEI layer decomposes at about 69 • C [90], which enables the reaction between electrolyte and anode to take palace. As the temperature builds up, the intercalated lithium in the anode can also react with electrolyte and release heat [102]. At around 130 • C, the polymer separator will melt [103,104], resulting in the internal short circuit between cathode and anode, therefore a number of reactions take place that contribute to the accumulation of heat and pressure. At approximately 150 • C, the safety vent of battery opens to decrease the pressure, hereafter gases releasing appears. As the continuous rising of battery temperature, the released combustible gases will be ignited, and then thermal runaway occurs. It can be observed that the surface temperature has a sharp increase to the peak value after thermal runaway, which ranges from 400-700 • C and it is related to battery chemistry, state of charge (SOC) and capacity etc. [105][106][107].
Once the safety vent cracks, quantities of gases will be ejected, which is accompanied by a clear sound [108,109]. The ejected gases are usually high-temperature, toxic and combustible, which will be ignited soon and exhibited as the first combustion. Furthermore, after thermal runaway appears, the second combustion will be generated. It is much more violent than the first one and is usually accompanied with the ejection of flame. The experimental phenomena during thermal runaway can be seen in Figure 3. According to the result of Fu et al., the axial flame temperature could reach as high as 800 • C [110]. Especially for the condition of overcharge, high-capacity and closed/semi-closed space, tremendous amounts of energy cannot be released effectively which may even result in the explosion of battery. Ouyang et al. researched the thermal runaway behavior of an overcharged 18,650 battery, and they found that the thermal runaway behavior of overcharged battery was much more violent than the normal battery. As depicted in Figure 4, where the safety vent of battery was blown away, the jellyroll was brought out and exposed to air [111]. With the help of an adiabatic calorimeter, vent sizing package 2 (VSP2), Jhu et al. found that the charged battery was prone to experiencing a thermal explosion compared to the uncharged battery, and the heat of reaction was calculated to be 26.2 kJ [112]. accumulation of heat and pressure. At approximately 150 °C, the safety vent of battery opens to decrease the pressure, hereafter gases releasing appears. As the continuous rising of battery temperature, the released combustible gases will be ignited, and then thermal runaway occurs. It can be observed that the surface temperature has a sharp increase to the peak value after thermal runaway, which ranges from 400-700 °C and it is related to battery chemistry, state of charge (SOC) and capacity etc. [105][106][107]. Once the safety vent cracks, quantities of gases will be ejected, which is accompanied by a clear sound [108,109]. The ejected gases are usually high-temperature, toxic and combustible, which will be ignited soon and exhibited as the first combustion. Furthermore, after thermal runaway appears, the second combustion will be generated. It is much more violent than the first one and is usually accompanied with the ejection of flame. The experimental phenomena during thermal runaway can be seen in Figure 3. According to the result of Fu et al., the axial flame temperature could reach as high as 800 °C [110]. Especially for the condition of overcharge, high-capacity and closed/semi-closed space, tremendous amounts of energy cannot be released effectively which may even result in the explosion of battery. Ouyang et al. researched the thermal runaway behavior of an overcharged 18,650 battery, and they found that the thermal runaway behavior of overcharged battery was much more violent than the normal battery. As depicted in Figure 4, where the safety vent of battery was blown away, the jellyroll was brought out and exposed to air [111]. With the help of an adiabatic calorimeter, vent sizing package 2 (VSP2), Jhu et al. found that the charged battery was prone to experiencing a thermal explosion compared to the uncharged battery, and the heat of reaction was calculated to be 26.2 kJ [112].   Commonly, quantities of toxic gases can be generated during the thermal runaway process such as HF, carbon monoxide (CO), nitric oxide (NO), sulfur dioxide (SO2), hydrogen chloride (HCl) and hydrogen (H2) [76,[113][114][115][116]. Due to the limited thermal stabilization of LiPF6, it may decompose at elevated temperatures and it can be described as [117,118]: 6 5 LiPF LiF PF → + On the other hand, the reaction between PVDF binder and lithiated graphite will occur in acidic medium and is as follows [119]:   Commonly, quantities of toxic gases can be generated during the thermal runaway process such as HF, carbon monoxide (CO), nitric oxide (NO), sulfur dioxide (SO2), hydrogen chloride (HCl) and hydrogen (H2) [76,[113][114][115][116]. Due to the limited thermal stabilization of LiPF6, it may decompose at elevated temperatures and it can be described as [117,118]: 6 5 LiPF LiF PF → + On the other hand, the reaction between PVDF binder and lithiated graphite will occur in acidic Commonly, quantities of toxic gases can be generated during the thermal runaway process such as HF, carbon monoxide (CO), nitric oxide (NO), sulfur dioxide (SO 2 ), hydrogen chloride (HCl) and hydrogen (H 2 ) [76,[113][114][115][116]. Due to the limited thermal stabilization of LiPF 6 , it may decompose at elevated temperatures and it can be described as [117,118]: On the other hand, the reaction between PVDF binder and lithiated graphite will occur in acidic medium and is as follows [119]: By means of fire calorimetry, Ribière et al. [76] carried out combustion experiments on pouch batteries and results revealed that an increasing amount of total HF emissions would be generated for the decreasing SOC value of battery. Besides, Ouyang et al. found that the generation of HF increased linearly with the increasing battery quantity and the burning of about 100 batteries (18,650) would release an amount of HF causing the immediate death of humans, 1.5 g [120,121].
In addition, resulting from the incomplete combustion of battery electrolyte as described in Equation (6), there will be some CO generated during thermal runaway [114]: It is well-known that CO can bind with hemoglobin causing oxygen poisoning. By varying battery SOC and ambient pressure, Chen et al. [121] declared that with the increasing battery SOC, the amount of CO 2 decreased, while the amount of CO increased. The decreasing ratio of CO 2 /CO revealed that the combustion efficiency of battery would be lower for the high SOC battery. Besides, a low combustion efficiency could also be obtained under low-pressure conditions, and the battery with higher capacity would generate much more CO during thermal runaway [114].
Based on the research of Ribière et al. [76], NO may be produced as a reaction product of nitrogen (originating from air or fuel-bound nitrogen) and oxygen from air within the flame (thermal route of NO x production), which can be further expressed as: NO will damage the peogaster of human and destroy the ozonosphere. Furthermore, it can be oxidized as corrosive nitrogen dioxide (NO 2 ). With the increase of battery SOC, the amount of NO first ascends and then descends.
Additionally, sulfur-based compounds as additives are known to be used in electrolytes for their property in facilitating SEI formation [122,123]. Whereas, they will undergo degradation to form SO 2 at high temperatures. Similarly, SO 2 is harmful to the peogaster of human and it can also lead to acid rain. Moreover, the amount of SO 2 will ascend with the increasing battery SOC.
The generation of HCl originates from the combustion of polymers inside battery containing a binder, separator and package [124]. It is found that the amount of HCl is irrelevant to battery SOC.
Finally, one possible source of H 2 is the reaction between metal Li and the separator. Common separator materials are PVDF and carboxymethyl cellulose (CMC). As can be seen from the reaction detailed below, PVDF may react with metal Li and release H 2 under high-temperature conditions [124]: A similar reaction of CMC and metal Li may take place to release H 2 which is as follows: H 2 is highly reactive, and it may lead to violent combustion or explosion under external high-temperature.

Thermal Hazards of a Battery Pack
Battery packs are generally consisted of single batteries, in which the quantity of batteries depends on the application ranging from several to thousands. For example, one charge-pal usually contains 2-4 batteries (18,650), while there are more than 7000 batteries for the pack in a Tesla Model S. Consequently, battery pack possesses the thermal hazards of a single battery as described in Section 4.2.1. Whereas, different to a single battery, a battery pack possesses the hazard of thermal failure propagation, that is, the thermal failure of one or several batteries will propagate to the neighboring ones, resulting in catastrophic consequences. The thermal hazard will be heightened during the propagation; hence it is essential to pay attention to the issue of failure propagation within a battery pack [105,[125][126][127][128][129][130][131][132][133].
Ouyang et al. carried out a set of failure propagation experiments in 18,650 packs with a size of 3 × 3 [128]. As depicted in Figure 5, the thermal failure of the battery in the lower right corner of the figure was induced by external heating, and then thermal failure propagated with a domino effect within the pack. Typically, the thermal failure of battery pack could be divided into several phases, i.e., the failure of the batteries in the former phase affected the batteries in the next phase and then caused the hierarchical propagation. Besides, they also researched the influence of several key factors such as the SOC, battery gap, failure location and pack size on the failure propagation. Results showed that higher SOC would worsen the propagation behavior of pack, in which the propagation speed grew linearly with the increasing pack SOC. It can be attributed to the fact that battery SOC affects the internal lithium ions distribution which will further influence the chain-reactions during thermal runaway. By increasing the battery gap, it was found that the risk of failure propagation within pack could be greatly reduced and the propagation speed exhibited a square index decline relationship with the battery gap, which was the result of radiation decay with squared distance. Upon further increasing the battery gap, the propagation behavior would eventually be interrupted. If the battery close to the center of pack underwent thermal failure, the failure propagation was revealed to be aggravated significantly since the failure of the central battery would affect the surrounding batteries with the largest scope and lead to the thermal failure of the whole pack in a relatively short time. Moreover, it was exhibited that with the increase of pack size, it would take more time for the failure propagation to reach the outermost phase, i.e., the thermal failure propagation would be delayed. In short, the thermal failure propagation within a battery pack is severe and the corresponding hazard will be greatly enhanced during the propagation, which may be influenced by several key parameters of the LIB including the SOC, battery gap, package and so on. To decrease the thermal hazard of battery pack, some methods such as improving the separator property, enhancing the heat dissipation, adding an insulation board and decreasing the SOC may be considered.  Due to the application scenarios, it is common to see that batteries are generally arranged in various shapes or modes within battery packs. Ouyang et al. [125] compared the thermal failure propagation of battery packs with various shapes, where the triangle pack, rectangle pack, parallelogram pack, line pack, hexagon pack, and square pack were included. The results of failure propagation speed are summarized in Figure 6, where the triangular pack and the linear pack are much safer as their propagation speeds are smaller than the others. After combining with the space utilization, it was revealed that the triangular pack might be the best choice of battery module due to its lower propagation hazard and higher space utilization. Figure 5. Schematic of thermal failure propagation. Figure 6. The propagation speeds of packs with various shapes where the numbering 1, 2, 3, 4, 5, and 6 represents the triangle pack, rectangle pack, parallelogram pack, line pack, hexagon pack and square pack respectively. Additionally, resulting from the disadvantages of traditional PCM such as low conductivity and diffusivity [134,135], it was found that the failure propagation behavior of a pack wrapped with PCM would be severer. The low conductivity and diffusivity of PCM caused a great deal of heat to be accumulated inside pack, and the close stacking of PCM reduced the heat release, therefore the propagation of thermal failure was accelerated. Besides, traditional PCM was flammable and it would be ignited by external heating which further aggravated the thermal failure propagation. In short, it is demanding to avoid the pack wrapped with traditional PCM being exposed to external heating or high temperature, especially for the flammable PCM. On the other hand, more work should be done to improve the thermal conductivity, diffusivity and fire resistance of PCM.
Apart from the factors above, it was determined that tab configuration also had a huge influence on the failure propagation behavior of the pack as a result of the heat transfer from the tab. According to the work of Lopez et al. [126], it was revealed that a branched style of tabbing as shown in Figure 7 improved the voltage retention as well as the safety of the pack over a serpentine style of tabbing as the shorted trigger battery was electrically better isolated from the rest of the pack when the tabs were branched. Lamb et al. [127] examined the failure propagation behavior of packs consisting of cylindrical and pouch batteries respectively, where the thermal failure was induced in a single battery. They observed that cylindrical batteries were less prone to propagation compared to the pouch batteries owing to the limited contact between neighboring batteries.
On the other hand, numerical simulation methods were also applied to investigate the failure propagation behaviors of battery packs [136][137][138][139][140][141][142][143][144]. Feng et al. [136] established a 3D thermal runaway propagation model for a large format LIB pack based on the energy balance equation. They proposed that the thermal failure propagation could be postponed by increasing the thermal runaway triggering temperature, which was generally reflected as the collapse temperature of battery separator. Furthermore, it was revealed that reducing the total electrical energy released during thermal runaway and enhancing the heat dissipation level were beneficial to suppress the thermal failure propagation. Coman et al. [137,138] presented a numerical model to analyze the heat propagation in a custom-made battery pack in which venting of the electrolyte and jellyroll contents were considered. Their results demonstrated that the heat dissipation due to the ejection of electrolyte and jellyroll contents contributed significantly to the failure propagation. In addition, they declared that the combination of small insulating layers wrapped around the batteries and a conductive heat sink were beneficial to the design of a safer battery pack which could mitigate the thermal failure propagation.

Thermal Hazards of the Lithium-Ion Battery under a Low-Pressure Environment
In particular, an alarming increase in the number of incidents on airlines caused by malfunctioning LIBs has raised more concerns among safety and aviation experts. The Federal Aviation Administration (FAA) has reported 46 incidents involving LIBs on aircraft in 2017, which is up from 31 in 2016 [145]. It is known that the low-pressure environment not only exists on the plane, but also in some high altitude areas. Moreover, it is important to note that there are many risks associated with both transportation and utilization in low-pressure environments. Fire behaviors under low pressure is different from that under normal conditions. Current research mainly focuses on the influence of low-pressure environment on the combustion characteristics of conventional liquid and solid fuels. Experimental results showed that pressure was the main factor in affecting the ignition [146][147][148]. More studies concerning the kinetics and mechanisms of fires at low pressures were also conducted to reveal the pressure effects on the burning process [149][150][151][152][153][154][155][156][157][158][159][160][161][162]. However, the research on the combustion characteristics of LIBs under a low-pressure environment is currently insufficient. Chen et al. took the pioneering studies on the fire behaviors of LIBs under low-pressure environment. The experiments were conducted in a sea-level city, Hefei (100.8 kPa/24 m), and a high-altitude city, Lhasa (64.3 kPa/3650 m) to offer guidance to facilitate the safe handling of battery under normal and high-altitude conditions. The results showed that the ejection behavior of LIBs in Lhasa appeared earlier than that in Hefei as shown in Figure 8, due to the lower pressure required to rupture, meanwhile the mass loss, heat release rate and combustion heat of the fire reduced with decreasing pressure [121]. They also studied the fire behaviors of primary lithium batteries at two different pressures. The mass loss, heat release rate, combustion heat, combustion efficiency and heat flux in Lhasa were smaller than those in Hefei which mean that the primary batteries at lower pressure were less dangerous than those at higher pressure [163]. Fu et al. studied the ignition and combustion characteristics of LIBs under low atmospheric pressure using a low-pressure tank from 30 to 101 kPa. Results indicated that the low atmospheric pressure could largely extend the ignition and weakened the combustion intensity of LIBs [110]. In summary, the study of the combustion characteristics of LIBs under low-pressure conditions has scarcely been examined, while it is of great importance to the safety management of LIBs in air transport and more work on this issue are necessary to be conducted in the future. In short, the thermal failure propagation within a battery pack is severe and the corresponding hazard will be greatly enhanced during the propagation, which may be influenced by several key parameters of the LIB including the SOC, battery gap, package and so on. To decrease the thermal hazard of battery pack, some methods such as improving the separator property, enhancing the heat dissipation, adding an insulation board and decreasing the SOC may be considered.

Thermal Hazards of the Lithium-Ion Battery under a Low-Pressure Environment
In particular, an alarming increase in the number of incidents on airlines caused by malfunctioning LIBs has raised more concerns among safety and aviation experts. The Federal Aviation Administration (FAA) has reported 46 incidents involving LIBs on aircraft in 2017, which is up from 31 in 2016 [145]. It is known that the low-pressure environment not only exists on the plane, but also in some high altitude areas. Moreover, it is important to note that there are many risks associated with both transportation and utilization in low-pressure environments. Fire behaviors under low pressure is different from that under normal conditions. Current research mainly focuses on the influence of low-pressure environment on the combustion characteristics of conventional liquid and solid fuels. Experimental results showed that pressure was the main factor in affecting the ignition [146][147][148]. More studies concerning the kinetics and mechanisms of fires at low pressures were also conducted to reveal the pressure effects on the burning process [149][150][151][152][153][154][155][156][157][158][159][160][161][162]. However, the research on the combustion characteristics of LIBs under a low-pressure environment is currently insufficient. Chen et al. took the pioneering studies on the fire behaviors of LIBs under low-pressure environment. The experiments were conducted in a sea-level city, Hefei (100.8 kPa/24 m), and a high-altitude city, Lhasa (64.3 kPa/3650 m) to offer guidance to facilitate the safe handling of battery under normal and high-altitude conditions. The results showed that the ejection behavior of LIBs in Lhasa appeared earlier than that in Hefei as shown in Figure 8, due to the lower pressure required to rupture, meanwhile the mass loss, heat release rate and combustion heat of the fire reduced with decreasing pressure [121]. They also studied the fire behaviors of primary lithium batteries at two different pressures. The mass loss, heat release rate, combustion heat, combustion efficiency and heat flux in Lhasa were smaller than those in Hefei which mean that the primary batteries at lower pressure were less dangerous than those at higher pressure [163]. Fu et al. studied the ignition and combustion characteristics of LIBs under low atmospheric pressure using a low-pressure tank from 30 to 101 kPa. Results indicated that the low atmospheric pressure could largely extend the ignition and weakened the combustion intensity of LIBs [110]. In summary, the study of the combustion characteristics of LIBs under low-pressure conditions has scarcely been examined, while it is of great importance to the safety management of LIBs in air transport and more work on this issue are necessary to be conducted in the future.

Thermal Hazards of Typical Commercial Lithium-Ion Batteries
Lithium cobaltate oxide (LCO) cathode material was first put forward by the team of Prof. Goodenough in 1980, and it was also the first time that Li was introduced into the positive electrode in batteries, which promoted the application and development of a negative electrode without Li and, at the same time, improved the charging voltage of batteries. Most importantly, the lack of Li metal anode greatly improved the safety of batteries [164]. Furthermore, Prof. Goodenough's team proposed the lithium manganate oxide (LMO) cathode material in 1983. Although its theoretical capacity is only about half of that of LCO, its reversible specific capacity is excellent [165]. With the emergence of LCO and LMO, many lamellar and spinel LIB positive electrode materials based on transition metal oxides were developed in the following decade. Subsequently, in 1997, Goodenough's team reported the application of lithium iron phosphate (LFP) positive electrode materials, and, in further research, solved a series of problems of LFP-positive electrode materials by carbon coating modification, so that LFP could be applied commercially on a large scale [166]. Therefore, the commercial cathode material structural system has been built, and the following positive electrode materials are improved on the basis of these three materials, while the NMC and lithium nickel cobalt aluminate oxide (NCA) materials improve the performance of the battery by regulating the content of the three elements in LCO.

Lithium cobaltate oxide
At present, LCO cathode material still occupies a major part in the 3C field, while the safety is relatively poor, so it is gradually being replaced by other positive electrode materials [167,168]. The active surface of the LCO cathode material can catalyze the decomposition of the carbonate electrolyte then release a large amount of heat, thus causing the thermal runaway of battery [169,170]. Zhang et al. studied the overcharge capacity degradation mechanism of LCO/graphite battery and found that the dissolution of current collector led to changes in the structure of the SEI

Thermal Hazards of Typical Commercial Lithium-Ion Batteries
Lithium cobaltate oxide (LCO) cathode material was first put forward by the team of Prof. Goodenough in 1980, and it was also the first time that Li was introduced into the positive electrode in batteries, which promoted the application and development of a negative electrode without Li and, at the same time, improved the charging voltage of batteries. Most importantly, the lack of Li metal anode greatly improved the safety of batteries [164]. Furthermore, Prof. Goodenough's team proposed the lithium manganate oxide (LMO) cathode material in 1983. Although its theoretical capacity is only about half of that of LCO, its reversible specific capacity is excellent [165]. With the emergence of LCO and LMO, many lamellar and spinel LIB positive electrode materials based on transition metal oxides were developed in the following decade. Subsequently, in 1997, Goodenough's team reported the application of lithium iron phosphate (LFP) positive electrode materials, and, in further research, solved a series of problems of LFP-positive electrode materials by carbon coating modification, so that LFP could be applied commercially on a large scale [166]. Therefore, the commercial cathode material structural system has been built, and the following positive electrode materials are improved on the basis of these three materials, while the NMC and lithium nickel cobalt aluminate oxide (NCA) materials improve the performance of the battery by regulating the content of the three elements in LCO.

Lithium cobaltate oxide
At present, LCO cathode material still occupies a major part in the 3C field, while the safety is relatively poor, so it is gradually being replaced by other positive electrode materials [167,168]. The active surface of the LCO cathode material can catalyze the decomposition of the carbonate electrolyte then release a large amount of heat, thus causing the thermal runaway of battery [169,170]. Zhang et al. studied the overcharge capacity degradation mechanism of LCO/graphite battery and found that the dissolution of current collector led to changes in the structure of the SEI layer and electrodes, thereby reducing the battery capacity and increasing the risk of thermal runaway [171]. In view of the safety problem associated with LCO, researchers have proposed a variety of solutions. Ji et al. [172] recommended a phosphazene additive N 3 P 3 (OPh)F 5 (PFPN), which could be oxidized to form a protective cathode electrolyte interface (CEI) before the electrolyte, greatly improving the safety of battery. Deng et al. [173] coated and doped LCO-positive electrode materials by adding Al and Ti elements, which greatly improved the thermal stability, increased the thermal runaway critical temperature and reduced the heat release. With the increasing demand for high-voltage LCO batteries, further studies are needed for the modification of LCO cathode materials, safe electrolytes and additives at high voltage.

Lithium iron phosphate
LFP is the positive electrode with the best thermal stability among the current commercial LIB cathode materials. Yamada et al. compared the thermal stability of various electrode materials at high temperature, and LFP showed lower oxygen and heat release [174]. Jiang et al. used an ARC to analyze the thermal stability of LCO, NMC and LFP with LiBOB and LiPF 6 electrolytes, and the thermal stability of LFP was found to be the best in both systems [175]. The high thermal stability of LFP is rooted in its strong structural stability during charging and discharging [176]. However, due to the poor conductivity of LFP, the migration rate of lithium ions is very low, therefore surface carbon coating, and Cu 2+ or Mg 2+ plasma doping are required for modification [177,178]. Furthermore, the modification of LFP is still a research focus. How to increase the electrochemical performance of LFP while maintaining the high level of safety is a key concern of researchers.

Lithium nickel manganese cobalt oxide
NMC ternary material is developed on the basis of LCO positive electrode materials by adjusting the content of Ni, Mn and Co elements. Co element plays an important role in the electronic conductance of battery, while Ni element can improve the capacity. Mn element is responsible for stabilizing the structure of positive electrode material and preventing the capacity decline of battery [179,180]. With the rapid development of EVs, ternary materials have attracted extensive attention due to their high capacity, while ternary materials with a high Ni content have gradually replaced low-Ni materials due to the lower price. However, with the increase of Ni content, the safety of batteries is greatly compromised [181]. The content of Ni in NMC materials increases gradually, from NMC111 to NMC532 and NMC622, but this still cannot meet the current demand for power battery capacity and low cost. Therefore, NMC811 materials have received extensive attention in recent years, but the extremely high reactivity between positive electrode and electrolyte also brings great safety risks [16]. Ma et al. studied the effect of vinylene carbonate (VC) and fluoroethylene carbonate (FEC) additives on the reactivity of lithium intercalation carbon (Li x C) and electrolyte, and the results showed that the addition of FEC could reduce the self-heating rate between graphite and electrolyte [182]. Three component additives, such as VC211 and PES211, could improve the cycling performance and the chemical stability of electrode/electrolyte, and the effect was much better than that of single or binary additives [183]. According to the latest research results, the thermal runaway of NMC batteries is mainly caused by chemical crosstalk between the electrodes, the oxygen released by cathode diffuses to the anode, at which point, an oxidizing reaction occurs and emits a large amount of heat, causing the thermal runaway of the NMC batteries. This is different from the traditional view that the main reason is the breakage of the diaphragm material, which leads to the internal short circuit then causes thermal runaway. Thus, it is possible to fundamentally solve the safety problem of high Ni ternary material [184].

Lithium nickel cobalt aluminate oxide
Similar to NMC material, NCA ternary material was also developed on the basis of LCO positive electrode materials by adjusting the content of Ni, Co and Al elements. Within, Ni and Co elements are applied to improve the capacity and the electronic conductance, respectively. The main purpose of Al addition is to stabilize the lattice structure and avoid structural collapse during charging/discharging. So far, NCA materials are generally regarded as one of the most appealing materials due to their comparable operating voltage, energy density and ideal properties for practical applications [185]. Especially for the LiNi 0.80 Co 0.15 Al 0.05 O 2 cathode, which has achieved ubiquitous commercial success. The remaining problem for NCA material is the rapid capacity fading during cycling, which is due to the mixing of Li/Ni cations [186]. Because the radii of Ni 2+ and Li + are similar, it is easy for Ni 2+ to enter the lithium layer, therefore results in capacity loss and thermal hazard. At present, several studies have been reported on the strategies to improve the stability of NCA materials [187][188][189]. These studies are mainly based on various cationic doping in Ni-rich cathode materials to cease the migration of Ni 2+ towards the Li + site. Moreover, doping in the Li + intercalation site is also a very important strategy to reduce structural disorientation. Among, magnesium (Mg) and sodium (Na) are the most well-studied dopants that can be embedded into the Li + site.

Management and Countermeasures
In view of the non-negligible thermal hazards of LIBs involving components, a single battery and a battery pack, management or countermeasures regarding these thermal hazards must be conducted. In the present study, the following is discussed: (1) the use of safety devices within battery; (2) the application of fire retardant (FR) additives; (3) the thermal management of battery; (4) provision of a warning once hazard occurs and (5) the firefighting after thermal hazard forms. The former three countermeasures aim to enhance the inherent safety of LIBs and decrease the possibility of hazard, and the others are applied to constrain the deterioration of thermal hazard, thereby reducing the damage.

Safety Devices within Battery
Due to the inherent hazards accompanied with LIB, manufacturers have developed several safety devices within the battery such as a safety vent, current interrupt device (CID) and positive temperature coefficient (PTC) to prevent the appearance of excessive current, temperature or pressure. Therefore, safety device can be regarded as a protective measure owed by the battery itself.

Safety Vent
During the operation of the LIB, especially under abusive conditions, it is common to see that gases will be released inside battery. Thereby, the safety vent was proposed in response to the continuous increase of pressure inside battery, allowing the escape of excessive gases. Generally, a safety vent is composed of a gasket containing a puncture film and a spike. The spike is located at the top of the battery, while the gasket lies below it. Once the pressure inside battery builds up, the gasket squeezes upward and deforms, enabling the puncture film punctured by the spike, therefore the dangerous rupture of battery casing can be terminated [190][191][192][193]. Moreover, safety vents can be designed to operate at pre-set internal battery temperatures to meet the requirement of practices.

Current Interrupt Device
Another common safety device is called the CID, and it can be activated by high temperature. Under the conditions such as overheating, overcharging and short-circuiting, an increase of temperature can be seen inside the battery causing an increase in the vapor pressure of the electrolyte solvent; therefore, the solder joints welded to the aluminum sheet and the pressure relief sheet will fall off and the pressure relief sheet will turn over leading to the broken circuit inside battery [193]. Compared to the mechanism of safety vent releasing the internal pressure, the CID is used to interrupt battery current, which subsequently prevents the occurrence of hazard.

Positive Temperature Coefficient
Besides, the PTC based on materials are also widely used among LIBs, whose resistance increases dramatically in response to the rapid rise of temperature [194][195][196][197][198][199]. For example, if a large current flows across the PTC element, its temperature increases abruptly, while the PTC works. A concomitant and abnormally high resistance of the PTC element prevents current flow which limits the further deterioration of heat generation. Once the large current is removed, the battery and PTC element cool and the resistance of the latter drops, allowing the resumption of charge/discharge. Overall, the primary purpose of PTC devices is to protect batteries against abusive electrical conditions such as external short circuit and overcharge etc.

Fire Retardant
As known, the combustibility of conventional battery components, e.g., electrolyte and separator bring an inherent hazard to LIB which may induce the occurrence of fire or combustion. Thereby, an increasing number of researchers endeavor to investigate how to change flammable components into non-flammable ones or how to develop new substance to add into battery to reduce the combustibility. Herein, a summary is detailed on the studies on FR additives of electrolyte and separator in LIB.

Fire-retardant Electrolyte
The electrolyte of a commercial LIB is a mixture of organic carbonates such as solvents and lithium salt, in which linear carbonates such as DMC, EMC or diethyl carbonate (DEC) and cyclic carbonates such as EC and propylene carbonate (PC) may be included. These organic carbonates are highly flammable which causes great concerns regarding their safety in LIB [200,201]. To solve this problem, the most straightforward way is to add FR components into solvents or thoroughly replace the flammable solvents [202].
However, some conflicts still exist between the property of FR additive and the electrochemical performance of LIB. In order to ensure the electrochemical performance of battery as much as possible, additives are necessary to meet the following requirements as depicted in Figure 9: (1) good chemical stability, no chemical reaction with battery components; (2) electrochemical inertia, no adverse electrochemical reactions within the normal operating voltage range of LIB; (3) suitable physical properties including conductivity, viscosity, boiling point, density, solubility etc.; (4) low toxicity, good machinability and appropriate cost [202].
Normally, the FR mechanism can be divided into two types: gas physical mechanism and chemical free-radical capture mechanism [203,204]. The former is based on oxygen isolation via the formation of a FR vapor and absorbing heat simultaneously, while the latter one is based on the additives capturing active free radicals and acids produced by combustion reaction to stop combustion development. Based on the elements contained and the working mechanism, currently studied FR additives for electrolytes can be divided into four main categories: phosphorus FR additives, fluoride FR additives, ionic liquid FR additives and composite FR additives, as illustrated in Figure 10.
possible, additives are necessary to meet the following requirements as depicted in Figure 9: (1) good chemical stability, no chemical reaction with battery components; (2) electrochemical inertia, no adverse electrochemical reactions within the normal operating voltage range of LIB; (3) suitable physical properties including conductivity, viscosity, boiling point, density, solubility etc.; (4) low toxicity, good machinability and appropriate cost [202]. Normally, the FR mechanism can be divided into two types: gas physical mechanism and chemical free-radical capture mechanism [203,204]. The former is based on oxygen isolation via the formation of a FR vapor and absorbing heat simultaneously, while the latter one is based on the additives capturing active free radicals and acids produced by combustion reaction to stop combustion development. Based on the elements contained and the working mechanism, currently   Figure 10. (1) Phosphorus Additives Considering their advantages such as environmental persistence, ability to bioaccumulation and low toxicity, organophosphorus compounds are one of the most popular candidates for FR additives in LIBs [205]. They will decompose at high temperature to produce hydrogen phosphate (H3PO4), which will continue to transform into meta phosphorous acid (HPO2), PO· and PO2·. HPO2 will promote the carbonation reaction of carbonate, PO· and PO2· will capture H· and O· and; therefore suppress the combustion of electrolytes. Until now, Xiang et al. [206] investigated the electrolyte with three kinds of dimethyl methylphosphonate (DMMP) additives respectively which could significantly suppress the flammability while only causing a small decrease in the electrochemical performance of battery, see Figure 11(a). Hu et al. [207] synthesized a new phisohinamidate bis(N,N-diethyl) (2-methoxyethoxy) methylphosphonamidate (DEMEMPA) as the N-P collaborative FR additive in electrolyte and found that 10 vol.% of DEMEMPA to 0.9 M LiPF6/EC/DMC could significantly inhibit the burning of electrolytes and provide a wide electrochemical window, see Figure 11(b). Besides, Murmann et al. [203] performed studies on several phosphates with different fluorination concentrations to determine the influence of (1) Phosphorus Additives Considering their advantages such as environmental persistence, ability to bioaccumulation and low toxicity, organophosphorus compounds are one of the most popular candidates for FR additives in LIBs [205]. They will decompose at high temperature to produce hydrogen phosphate (H 3 PO 4 ), which will continue to transform into meta phosphorous acid (HPO 2 ), PO· and PO 2 ·. HPO 2 will promote the carbonation reaction of carbonate, PO· and PO 2 · will capture H· and O· and; therefore suppress the combustion of electrolytes. Until now, Xiang et al. [206] investigated the electrolyte with three kinds of dimethyl methylphosphonate (DMMP) additives respectively which could significantly suppress the flammability while only causing a small decrease in the electrochemical performance of battery, see Figure 11a. Hu et al. [207] synthesized a new phisohinamidate bis(N,N-diethyl) (2-methoxyethoxy) methylphosphonamidate (DEMEMPA) as the N-P collaborative FR additive in electrolyte and found that 10 vol.% of DEMEMPA to 0.9 M LiPF 6 /EC/DMC could significantly inhibit the burning of electrolytes and provide a wide electrochemical window, see Figure 11b. Besides, Murmann et al. [203] performed studies on several phosphates with different fluorination concentrations to determine the influence of fluorination degree within organophosphate on the flammability and electrochemical performance. Their results depicted that the partially fluorinated phosphate offered the best cycling and FR results.
(H3PO4), which will continue to transform into meta phosphorous acid (HPO2), PO· and PO2·. HPO2 will promote the carbonation reaction of carbonate, PO· and PO2· will capture H· and O· and; therefore suppress the combustion of electrolytes. Until now, Xiang et al. [206] investigated the electrolyte with three kinds of dimethyl methylphosphonate (DMMP) additives respectively which could significantly suppress the flammability while only causing a small decrease in the electrochemical performance of battery, see Figure 11(a). Hu et al. [207] synthesized a new phisohinamidate bis(N,N-diethyl) (2-methoxyethoxy) methylphosphonamidate (DEMEMPA) as the N-P collaborative FR additive in electrolyte and found that 10 vol.% of DEMEMPA to 0.9 M LiPF6/EC/DMC could significantly inhibit the burning of electrolytes and provide a wide electrochemical window, see Figure 11(b). Besides, Murmann et al. [203] performed studies on several phosphates with different fluorination concentrations to determine the influence of fluorination degree within organophosphate on the flammability and electrochemical performance. Their results depicted that the partially fluorinated phosphate offered the best cycling and FR results. (2) Fluoride Additives Flurorinated substituents in organic molecules are known to suppress the flammability of molecules and reduce the boiling temperature and viscosity. By replacing hydrogen atoms with fluorine atoms, H· radicals can be reduced or eliminated to yield a non-flammable electrolyte. Ding et al. [208] introduced tris(2,2,2-trifluoroethyl) phosphate (TFP) as the FR additive into a LiPF6-based electrolytes containing EC/PC/EMC, and their results showed that the addition of TFP caused a continuous decrease in the dielectric constant, and a continuous increase in the glass transition and boiling temperatures. Therefore, the flammability of electrolytes was substantially reduced. Zeng et al. [209] synthesized a novel fluorinated alkyl phosphonate bis(2, 2, 2-trifluoroethyl) methylphosphonate (TFMP) as the FR additive to reduce the flammability of electrolyte, see Figure  12. Their experiments demonstrated that the addition of 20 vol.% TFMP could yield an electrolyte that was hardly flammable while having the least impact on the electrochemical performance. (

3) Ionic Liquid
An ionic liquid usually refers to a liquid salt at room temperature solely composed of anions and cations. It possesses certain characteristics such as it is non-volatile, nonflammable, pollution free and has a wide electrochemical window. Hence, the electrolyte with ionic liquid added is hoped to replace the traditional organic electrolyte to improve the safety of LIBs [210][211][212][213][214]. Rectenwald et al. [215] reported the successful design of a new class of lithium salts, phosphoryl-rich FR ions (2) Fluoride Additives Flurorinated substituents in organic molecules are known to suppress the flammability of molecules and reduce the boiling temperature and viscosity. By replacing hydrogen atoms with fluorine atoms, H· radicals can be reduced or eliminated to yield a non-flammable electrolyte. Ding et al. [208] introduced tris(2,2,2-trifluoroethyl) phosphate (TFP) as the FR additive into a LiPF 6 -based electrolytes containing EC/PC/EMC, and their results showed that the addition of TFP caused a continuous decrease in the dielectric constant, and a continuous increase in the glass transition and boiling temperatures. Therefore, the flammability of electrolytes was substantially reduced. Zeng et al. [209] synthesized a novel fluorinated alkyl phosphonate bis(2, 2, 2-trifluoroethyl) methylphosphonate (TFMP) as the FR additive to reduce the flammability of electrolyte, see Figure 12. Their experiments demonstrated that the addition of 20 vol.% TFMP could yield an electrolyte that was hardly flammable while having the least impact on the electrochemical performance. (2) Fluoride Additives Flurorinated substituents in organic molecules are known to suppress the flammability of molecules and reduce the boiling temperature and viscosity. By replacing hydrogen atoms with fluorine atoms, H· radicals can be reduced or eliminated to yield a non-flammable electrolyte. Ding et al. [208] introduced tris(2,2,2-trifluoroethyl) phosphate (TFP) as the FR additive into a LiPF6-based electrolytes containing EC/PC/EMC, and their results showed that the addition of TFP caused a continuous decrease in the dielectric constant, and a continuous increase in the glass transition and boiling temperatures. Therefore, the flammability of electrolytes was substantially reduced. Zeng et al. [209] synthesized a novel fluorinated alkyl phosphonate bis(2, 2, 2-trifluoroethyl) methylphosphonate (TFMP) as the FR additive to reduce the flammability of electrolyte, see Figure  12. Their experiments demonstrated that the addition of 20 vol.% TFMP could yield an electrolyte that was hardly flammable while having the least impact on the electrochemical performance. Figure 12. The synthetic route of bis(2, 2, 2-trifluoroethyl) methylphosphonate (TFMP). Adopted from Ref. [209].

(3) Ionic Liquid
An ionic liquid usually refers to a liquid salt at room temperature solely composed of anions and cations. It possesses certain characteristics such as it is non-volatile, nonflammable, pollution free and has a wide electrochemical window. Hence, the electrolyte with ionic liquid added is hoped to replace the traditional organic electrolyte to improve the safety of LIBs [210][211][212][213][214]. Rectenwald et al. [215] reported the successful design of a new class of lithium salts, phosphoryl-rich FR ions (FRIONs) for safer LIB, see Figure 13. The thermogravimetric analysis of these lithium salts showed An ionic liquid usually refers to a liquid salt at room temperature solely composed of anions and cations. It possesses certain characteristics such as it is non-volatile, nonflammable, pollution free and has a wide electrochemical window. Hence, the electrolyte with ionic liquid added is hoped to replace the traditional organic electrolyte to improve the safety of LIBs [210][211][212][213][214]. Rectenwald et al. [215] reported the successful design of a new class of lithium salts, phosphoryl-rich FR ions (FRIONs) for safer LIB, see Figure 13. The thermogravimetric analysis of these lithium salts showed that they were thermally stable up to around 200 • C. Moreover, the pyrolysis combustion flow calorimetry revealed that these salts would produce high char yields upon combustion which indicated the excellent FR property. Besides, the promising ionic liquid pentafluorocyclotriphosphazene (FPPN) is known to exhibit both an excellent FR performance and electrochemical properties. Dagger et al. [216] investigated a standard electrolyte mixed with 5 wt.% FPPN towards abuse tolerance in a 5 Ah battery and their results showed that FPPN significantly reduced the self-heating rate of the battery in the temperature range from 80 to 110 • C. Further, they conducted nail penetration and external short circuit experiments but found no significant difference between standard and FPPN-containing batteries.

(4) Composite Additives
Use of a single flame retardant generally requires adding a large amount. In addition, the solubility of the additive and the electrolyte compatibility are also relatively limited. Thus, it is necessary to use a composite approach by comparing the advantages and disadvantages of various types of retardants to configure more suitable additives for LIB. Wu et al. [217] synthesized a phosphazene compound triethoxyphosphazen-N-phosphoryldiethylester (PNP) by a facile method and characterized it as a FR electrolyte additive for LIB, see Figure 14. They found that the self-extinguishing time (SET) value of 10% PNP in the electrolyte was decreased by 40% compared to the base electrolyte, implying a strong inhabitation to the flammability. Additionally, Li et al. [218] synthesized a multi-functional FR electrolyte additive pentafluoro cyclotriphosphazene (PFPN). The electrolyte with 5 vol.% of PFPN was found to be a non-flammable electrolyte, moreover it could reduce the charge transfer resistance of battery resulting in a decreased electrode polarization and

(4) Composite Additives
Use of a single flame retardant generally requires adding a large amount. In addition, the solubility of the additive and the electrolyte compatibility are also relatively limited. Thus, it is necessary to use a composite approach by comparing the advantages and disadvantages of various types of retardants to configure more suitable additives for LIB. Wu et al. [217] synthesized a phosphazene compound triethoxyphosphazen-N-phosphoryldiethylester (PNP) by a facile method and characterized it as a FR electrolyte additive for LIB, see Figure 14. They found that the self-extinguishing time (SET) value of 10% PNP in the electrolyte was decreased by 40% compared to the base electrolyte, implying a strong inhabitation to the flammability. Additionally, Li et al. [218] synthesized a multi-functional FR electrolyte additive pentafluoro cyclotriphosphazene (PFPN). The electrolyte with 5 vol.% of PFPN was found to be a non-flammable electrolyte, moreover it could reduce the charge transfer resistance of battery resulting in a decreased electrode polarization and enhanced electrochemical performance at low temperature. Finally, several representatives belonging to the four kinds of additives above are listed in Figure 15.
solubility of the additive and the electrolyte compatibility are also relatively limited. Thus, it is necessary to use a composite approach by comparing the advantages and disadvantages of various types of retardants to configure more suitable additives for LIB. Wu et al. [217] synthesized a phosphazene compound triethoxyphosphazen-N-phosphoryldiethylester (PNP) by a facile method and characterized it as a FR electrolyte additive for LIB, see Figure 14. They found that the self-extinguishing time (SET) value of 10% PNP in the electrolyte was decreased by 40% compared to the base electrolyte, implying a strong inhabitation to the flammability. Additionally, Li et al. [218] synthesized a multi-functional FR electrolyte additive pentafluoro cyclotriphosphazene (PFPN). The electrolyte with 5 vol.% of PFPN was found to be a non-flammable electrolyte, moreover it could reduce the charge transfer resistance of battery resulting in a decreased electrode polarization and enhanced electrochemical performance at low temperature. Finally, several representatives belonging to the four kinds of additives above are listed in Figure 15.  Commonly, the choice of FR electrolyte additive requires a comprehensive consideration including conductivity, solubility, retardant efficiency and cost etc. Table 2 summarizes the advantages/disadvantages of the typical FR additives stated above. It can be observed that the key challenges of these four kinds of FR additives which remain to be solved are: (1) the phosphorus FR needs to prevent reductive decomposition and cointercalation; (2) a better retardant efficiency and a lower cost are necessary for fluoride FR; (3) the ionic liquid FR shall be enhanced by reducing its cost while enhancing the electrochemical compatibility; (4) the composite FR must simplify their synthesis process and thus they can be obtained at a lower cost and in a shorter time. In brief, the development of FR additive aims to achieve a balance between the electrochemical performance and the retardant effect. Going forward, we will obtain knowledge of a more advanced substance and then make good use of synergistic effect to get a win-win situation.  Commonly, the choice of FR electrolyte additive requires a comprehensive consideration including conductivity, solubility, retardant efficiency and cost etc. Table 2 summarizes the advantages/disadvantages of the typical FR additives stated above. It can be observed that the key challenges of these four kinds of FR additives which remain to be solved are: (1) the phosphorus FR needs to prevent reductive decomposition and cointercalation; (2) a better retardant efficiency and a lower cost are necessary for fluoride FR; (3) the ionic liquid FR shall be enhanced by reducing its cost while enhancing the electrochemical compatibility; (4) the composite FR must simplify their synthesis process and thus they can be obtained at a lower cost and in a shorter time. In brief, the development of FR additive aims to achieve a balance between the electrochemical performance and the retardant effect. Going forward, we will obtain knowledge of a more advanced substance and then make good use of synergistic effect to get a win-win situation. Within LIB, a separator is considered to be a critical component in securing battery safety [219][220][221][222]. It acts as the separation between the cathode and the anode while providing a pathway for lithium ion migration. Once the separator fails, an internal short circuit between electrodes may occur and lead to the thermal failure of LIBs [223][224][225]. As shown in Figure 16, a safe separator has strict demands on the following six factors: (1) chemical/mechanical stability; (2) thickness; (3) thermal stability; (4) wettability; (5) barrier property and (6)  Within LIB, a separator is considered to be a critical component in securing battery safety [219][220][221][222]. It acts as the separation between the cathode and the anode while providing a pathway for lithium ion migration. Once the separator fails, an internal short circuit between electrodes may occur and lead to the thermal failure of LIBs [223][224][225]. As shown in Figure 16, a safe separator has strict demands on the following six factors: (1) chemical/mechanical stability; (2) thickness; (3) thermal stability; (4) wettability; (5) barrier property and (6) porous property. As stated before, a commercialized separator typically consists of PE or PP, and has persistent drawbacks of a low melting temperature and poor wettability for liquid electrolytes which make it difficult to widely adopt in large-scale LIB. To overcome these obstacles and enhance its thermal safety, a composite separator may be a feasible solution. Yeon et al. [226] synthesized several FR composites (Al2O3-sputtered PE separator, Al(OH)2-composite, Mg(OH)3-composite) and verified in their experiments that these composites exhibited superior FR properties compared to traditional separator as they could release water at high temperatures. Meanwhile, bromine is another choice as a composite FR separator. Lee et al. [227,228] developed and investigated a brominated poly(2,6-dimethyl-1,4-phenylene oxide) (BPPO) coated separator to improve the FR ability and thermal ability. Besides, Zhang et al. [229] developed a sustainable, heat-resistant and cellulose-based composite FR separator for high-performance LIB, namely FCCN, i.e., flame-retardant and thermal resistant cellulose-based composite nonwoven separator which showed superior heat tolerance and good mechanical strength. As stated before, a commercialized separator typically consists of PE or PP, and has persistent drawbacks of a low melting temperature and poor wettability for liquid electrolytes which make it difficult to widely adopt in large-scale LIB. To overcome these obstacles and enhance its thermal safety, a composite separator may be a feasible solution. Yeon et al. [226] synthesized several FR composites (Al 2 O 3 -sputtered PE separator, Al(OH) 2 -composite, Mg(OH) 3 -composite) and verified in their experiments that these composites exhibited superior FR properties compared to traditional separator as they could release water at high temperatures. Meanwhile, bromine is another choice as a composite FR separator. Lee et al. [227,228] developed and investigated a brominated poly(2,6-dimethyl-1,4-phenylene oxide) (BPPO) coated separator to improve the FR ability and thermal ability. Besides, Zhang et al. [229] developed a sustainable, heat-resistant and cellulose-based composite FR separator for high-performance LIB, namely FCCN, i.e., flame-retardant and thermal resistant cellulose-based composite nonwoven separator which showed superior heat tolerance and good mechanical strength.

Battery Management System
In practical applications, LIBs are required to operate at different conditions where a complicated temperature environment such as high temperature, low temperature and a largely varying temperature etc. may be involved. As stated by many scholars, temperature greatly affects battery from both safety and performance points of view [230][231][232][233][234]. Hence, it is of great importance to configure an efficient battery management system (BMS) that allows for control of battery temperature and prevents the occurrence of thermal hazard.
BMS applied in EV is usually comprised of sensors, actuators and controllers, and it is used to achieve three main aims: (1) to protect battery from being damaged; (2) to make battery operate within the proper voltage and temperature interval; (3) to maintain the performance of battery to meet the requirement of vehicles [235]. Traditionally, BMS can be briefly divided into two categories: internal thermal management (ITM) and external thermal management (ETM), which is further displayed in Figure 17. internal thermal management (ITM) and external thermal management (ETM), which is further displayed in Figure 17. Figure 17. Summary of battery thermal management systems including the external management system (air cooling system, liquid cooling system, PCM cooling system) and the internal management system (alternating current, component optimization).
ITM, also known as intrinsic safety management, refers to those thermal management methods applied at the battery level in which the thermal management strategies exist in the core region of the battery monomer. One common approach is via optimizing battery components to be safer so that less heat is generated and better uniformly can be achieved. Choi et al. [236] proposed a cooling method to remove the heat generated within the lead-acid battery by means of controlling electrolyte circulation rate. Their results showed that such an approach could provide a uniform temperature field and the temperature could be controlled to a desirable level. As shown in Figure  18a, Mohammadian et al. [237] introduced a particular type of internal cooling method, where liquid electrolyte served as a coolant and flowed through the micro-channels dispersed in electrodes. Besides, Bandhauer et al. [238] developed a novel system that utilized an internal evaporator with micro-channels incorporated in a thick current collector and then applied to side cooling, see Figure  18b. Figure 17. Summary of battery thermal management systems including the external management system (air cooling system, liquid cooling system, PCM cooling system) and the internal management system (alternating current, component optimization).
ITM, also known as intrinsic safety management, refers to those thermal management methods applied at the battery level in which the thermal management strategies exist in the core region of the battery monomer. One common approach is via optimizing battery components to be safer so that less heat is generated and better uniformly can be achieved. Choi et al. [236] proposed a cooling method to remove the heat generated within the lead-acid battery by means of controlling electrolyte circulation rate. Their results showed that such an approach could provide a uniform temperature field and the temperature could be controlled to a desirable level. As shown in Figure 18a, Mohammadian et al. [237] introduced a particular type of internal cooling method, where liquid electrolyte served as a coolant and flowed through the micro-channels dispersed in electrodes. Besides, Bandhauer et al. [238] developed a novel system that utilized an internal evaporator with micro-channels incorporated in a thick current collector and then applied to side cooling, see Figure 18b.  Figure 18. Internal thermal management: (a) internal cooling by micro-channel in electrode, adopted from Ref. [237]; (b) internal cooling by micro-channel evaporator in current collector, adopted from Ref. [238].
Another way to achieve ITM is via the application of alternating current (AC) that can efficiently heat battery without a substantial change of SOC. Stuart et al. [239] first proposed using AC to warm battery up rather than using an external heater. For this method, the heating rate increased with the increasing signal amplitude, however the effect of signal frequency on the heating performance was not taken into account. To address this problem, Ruan et al. [240] proposed an effective strategy to analyze the optimal frequency. Their results demonstrated that constant frequency was more promising than variable one for engineering realization and the optimal frequency could be evaluated according to the intermediate temperature.
Nevertheless, in practical applications, the ITM method is difficult to operate and some disadvantages such as low cooling efficiency and temperature gradient etc. also limit its usage [240]. Comparatively, another BMS, i.e., ETM has received a more universal application. Figure 18. Internal thermal management: (a) internal cooling by micro-channel in electrode, adopted from Ref. [237]; (b) internal cooling by micro-channel evaporator in current collector, adopted from Ref. [238].
Another way to achieve ITM is via the application of alternating current (AC) that can efficiently heat battery without a substantial change of SOC. Stuart et al. [239] first proposed using AC to warm battery up rather than using an external heater. For this method, the heating rate increased with the increasing signal amplitude, however the effect of signal frequency on the heating performance was not taken into account. To address this problem, Ruan et al. [240] proposed an effective strategy to analyze the optimal frequency. Their results demonstrated that constant frequency was more promising than variable one for engineering realization and the optimal frequency could be evaluated according to the intermediate temperature.
Nevertheless, in practical applications, the ITM method is difficult to operate and some disadvantages such as low cooling efficiency and temperature gradient etc. also limit its usage [240]. Comparatively, another BMS, i.e., ETM has received a more universal application.
Corresponding to the ITM, ETM refers to the thermal management methods that improve heat dissipation from external environment. In order to promote heat dissipation, it is important to choose a suitable medium. On the basis of the medium, ETM can be further divided into three kinds: air cooling system (ACS), liquid cooling system (LCS) and PCM cooling system (PCM-CS).
There are two types of air flow patterns in ACS: natural convection (NC) and forced convection (FC). For the ACS, air flow rate is an important factor in the promotion of the heat transfer coefficient, thus improving the efficiency of heat dissipation. Additionally, certain factors such as battery arrangement, flow path number and width also influence the performance. Wang et al. [241] investigated the thermal behaviors of an air-cooled LIB. Their results showed that temperature non-uniformity caused by heat accumulation and concentration was mitigated by reciprocating air-flow with optimized reversing strategy and forward air-flow. Liu et al. [242] conducted a set of optimization methods in reciprocating air flow of LIB thermal management. They found that the reciprocating flow could achieve good dissipation, reduce the temperature difference, improve the temperature uniformity and effectively reduce the maximum temperature of battery. As displayed in Figure 19, Zhang et al. [243] proposed a novel battery pack with a wedge-shaped runner, optimizing the position of inlet/outlet, the width of the wedge-shaped flow path, the inclination angle of the battery and the clearance among batteries. Their results showed that the temperature distribution was exacerbated with the increase of the inlet temperature and improved with the increasing inlet velocity. However, despite the fact that researchers have never stopped optimizing the structure of ACS, it is still hard for ACS to meet the demands of EV at present. Nelson et al. [244] argued that using an air-based system it was difficult to cool the battery to less than 52 • C when the battery temperature was above 66 • C.
Appl. Sci. 2019, 9, x 26 of 46 Corresponding to the ITM, ETM refers to the thermal management methods that improve heat dissipation from external environment. In order to promote heat dissipation, it is important to choose a suitable medium. On the basis of the medium, ETM can be further divided into three kinds: air cooling system (ACS), liquid cooling system (LCS) and PCM cooling system (PCM-CS).
There are two types of air flow patterns in ACS: natural convection (NC) and forced convection (FC). For the ACS, air flow rate is an important factor in the promotion of the heat transfer coefficient, thus improving the efficiency of heat dissipation. Additionally, certain factors such as battery arrangement, flow path number and width also influence the performance. Wang et al. [241] investigated the thermal behaviors of an air-cooled LIB. Their results showed that temperature non-uniformity caused by heat accumulation and concentration was mitigated by reciprocating air-flow with optimized reversing strategy and forward air-flow. Liu et al. [242] conducted a set of optimization methods in reciprocating air flow of LIB thermal management. They found that the reciprocating flow could achieve good dissipation, reduce the temperature difference, improve the temperature uniformity and effectively reduce the maximum temperature of battery. As displayed in Figure 19, Zhang et al. [243] proposed a novel battery pack with a wedge-shaped runner, optimizing the position of inlet/outlet, the width of the wedge-shaped flow path, the inclination angle of the battery and the clearance among batteries. Their results showed that the temperature distribution was exacerbated with the increase of the inlet temperature and improved with the increasing inlet velocity. However, despite the fact that researchers have never stopped optimizing the structure of ACS, it is still hard for ACS to meet the demands of EV at present. Nelson et al. [244] argued that using an air-based system it was difficult to cool the battery to less than 52 °C when the battery temperature was above 66 °C. Figure 19. A representative air-cooling system. Adopted from Ref. [243].
Compared to the ACS, a liquid-based cooling system possesses better performance due to its higher heat transfer coefficient. Normally, the LCS can be divided into two kinds according to whether a phase change occurs during the process, namely whether water changes into vapor which usually takes place in the heat pipe. Apparently, the phase change may be a driving force in heat absorption. Shang et al. [245] designed a LCS for LIB by varying the contact surface, and the results showed that increasing inlet mass flow could effectively limit the maximum temperature, however, it could not significantly improve the temperature uniformity. As shown in Figure 20, Zhao et al. [246] developed two cooling approaches via channeled liquid flow to minimize the thermal non-uniformity in a battery pack and found that the two LCSs could reduce the non-uniformity of battery under 5C discharging operation to values lower than 2.2 and 0.7 °C, respectively. In addition, Figure 19. A representative air-cooling system. Adopted from Ref. [243].
Compared to the ACS, a liquid-based cooling system possesses better performance due to its higher heat transfer coefficient. Normally, the LCS can be divided into two kinds according to whether a phase change occurs during the process, namely whether water changes into vapor which usually takes place in the heat pipe. Apparently, the phase change may be a driving force in heat absorption. Shang et al. [245] designed a LCS for LIB by varying the contact surface, and the results showed that increasing inlet mass flow could effectively limit the maximum temperature, however, it could not significantly improve the temperature uniformity. As shown in Figure 20, Zhao et al. [246] developed two cooling approaches via channeled liquid flow to minimize the thermal non-uniformity in a battery pack and found that the two LCSs could reduce the non-uniformity of battery under 5C discharging operation to values lower than 2.2 and 0.7 • C, respectively. In addition, Wang et al. [247] developed a new LCS based on thermal silica plates, which could greatly improve the cooling capacity and therefore decrease the maximum temperature difference within a battery.
Nevertheless, the fluidity of liquid is a huge barrier for electronic system which requires advanced and complicated design before application, therefore resulting in a high cost. Recently, Ren et al. [248] presented a novel method of employing water evaporation, as displayed in Figure 21. The thin sodium film (SA-1 film) with a water content of 99 wt.% was prepared and their results revealed that under the condition of a cycle rate larger than 1C, the temperature rise rate of battery could be reduced by half. Meanwhile, the liquid barrier was successfully solved by the proposed SA-1 film such that there was not a large effect on the electronic system. Wang et al. [247] developed a new LCS based on thermal silica plates, which could greatly improve the cooling capacity and therefore decrease the maximum temperature difference within a battery. Figure 20. The procedure of a representative liquid cooling method. Adopted from Ref. [246].
Nevertheless, the fluidity of liquid is a huge barrier for electronic system which requires advanced and complicated design before application, therefore resulting in a high cost. Recently, Ren et al. [248] presented a novel method of employing water evaporation, as displayed in Figure 21. The thin sodium film (SA-1 film) with a water content of 99 wt.% was prepared and their results revealed that under the condition of a cycle rate larger than 1C, the temperature rise rate of battery could be reduced by half. Meanwhile, the liquid barrier was successfully solved by the proposed SA-1 film such that there was not a large effect on the electronic system. Wang et al. [247] developed a new LCS based on thermal silica plates, which could greatly improve the cooling capacity and therefore decrease the maximum temperature difference within a battery. Figure 20. The procedure of a representative liquid cooling method. Adopted from Ref. [246].
Nevertheless, the fluidity of liquid is a huge barrier for electronic system which requires advanced and complicated design before application, therefore resulting in a high cost. Recently, Ren et al. [248] presented a novel method of employing water evaporation, as displayed in Figure 21. The thin sodium film (SA-1 film) with a water content of 99 wt.% was prepared and their results revealed that under the condition of a cycle rate larger than 1C, the temperature rise rate of battery could be reduced by half. Meanwhile, the liquid barrier was successfully solved by the proposed SA-1 film such that there was not a large effect on the electronic system. PCM is a type of material that is able to absorb thermal energy and then release it by a reverse process; hence, it is considered as a novel medium in BMS [249]. Commonly, it consists of three categories: the organic, the inorganic and the eutectic. The organic PCM contains carbon element, in which paraffin is the most common compound. For organic PCM, it possesses the advantage of no needing a container as their volume variation is relatively low, besides the chemical stability and PCM is a type of material that is able to absorb thermal energy and then release it by a reverse process; hence, it is considered as a novel medium in BMS [249]. Commonly, it consists of three categories: the organic, the inorganic and the eutectic. The organic PCM contains carbon element, in which paraffin is the most common compound. For organic PCM, it possesses the advantage of no needing a container as their volume variation is relatively low, besides the chemical stability and non-corrosiveness [250]. The inorganic PCMs usually contain one or more metallic atoms. They have a good thermal conductivity and latent heat. Additionally, they are nonflammable and nontoxic. Nevertheless, some of them are corrosive and the reversibility of phase change is difficult to maintain in the long term. Moreover, the metals contained in inorganic PCM result in a high density which greatly limits its practical application. The eutectic PCM is a specific mixture of the previous PCMs, which permits the obtainment of a PCM with properties different to pure materials.
For most PCMs, their latent heat is large, while their thermal conductivity is relatively poor. Hence, considerable studies have been carried out to enhance the thermal conductivity of existing PCMs as displayed in Table 3, and some other materials with high thermal conductivity such as expanded graphite (EG), foamed metal and nanomaterials are also introduced to synthesize composite phase change material (CPCM). Wang et al. [251] developed an improved BMS based on the CPCM with aluminum boxes. Their results showed that the CPCM could significantly reduce the average temperature and improve temperature uniformity of a battery pack. Huang et al. [252] investigated the cooling effectiveness of CPCM on the thermal response of a battery under overcharge condition and found that the battery cooled down remarkably as compared to the natural convection condition when CPCM was attached to the pouch battery. Ling et al. [253] compared two CPCMs, a highly thermal conductive 60 wt.% RT44HC/EG composite and a 60 wt.% RT44HC/fumed silica composite with a lower thermal conductivity. Their results indicated that the low thermal conductivity of the CPCM would induce an even higher temperature difference over the battery pack in the absence of PCM. In summary, each type of BMS has its own merits and drawbacks, herein a comparison is conducted among these different types of BMS, shown in Table 4. Most scholars agree on the fact that air cooling is easy to operate but is not enough to cope with the temperature elevation of battery. Liquid cooling is performant but complex and presents the shortest lifespan. As for PCM, the leakage danger is a problem in application, thus a container is necessary.
Nowadays, many researchers have focused on developing a combined battery management system (CBMS), in which more than one of the above thermal management methods are applied. In this way, different methods can complement mutual strengths and weaknesses to present excellent performance. Ling et al. [263] developed a hybrid thermal management system for LIB-integrated PCM with forced-air cooling, as depicted in Figure 23. This CBMS successfully prevented heat accumulation and maintained the maximum temperature under 50 • C, in which the PCM dictated the maximum temperature rise and temperature uniformity within the battery pack while forced air convection succeeded in recovering the thermal energy storage capacity of PCM. Fathabadi et al. [264] proposed a novel battery pack design including a hybrid active-passive thermal management system. Their developed CBMS could be divided into two parts: the active part applied distributed thin ducts, air flow and natural convection as cooling media while the passive part utilized PCM/EG as the cooling component to optimize the thermal performance of the proposed battery pack. Xie et al. [265] developed a CPCM-based thermal management system coupled with air cooling and their results showed that the cooling effect of the integrated system was evidently better than that of the air cooling system. Furthermore, it was more energy-efficient due to the relatively lower air flow rate.  Figure 22. (a) The fabrication procedure of diamond foam and the generated composite phase change material (CPCM), adopted from Ref. [259]; (b) The synthetic route of polyethylene glycol/expanded graphite (PEG/EG), adopted from Ref. [262].
In summary, each type of BMS has its own merits and drawbacks, herein a comparison is conducted among these different types of BMS, shown in Table 4. Most scholars agree on the fact that air cooling is easy to operate but is not enough to cope with the temperature elevation of battery. Liquid cooling is performant but complex and presents the shortest lifespan. As for PCM, the leakage danger is a problem in application, thus a container is necessary.   As shown in Figure 24, Song et al. [266] investigated a CBMS that integrated liquid cooling and PCM together. Their results demonstrated that both the battery temperature ramp-up rate and the battery steady-state temperature were significantly reduced by the conjugated cooling, in comparison with single PCM or liquid cooling condition. Ling et al. [267] investigated an optimization method, the response surface methodology (RSM) with PCM and liquid cooling integrated for LIB, which helped retain the PCM mass by up to 94.1% and the volume by up to 55.6%. Compared with the conventional LCS, the hybrid system is not only highly efficient but lightweight and flexible for the batteries with arbitrary shapes. Zheng et al. [268] proposed a BMS combining liquid cooling and PCM for a fast-charging LIB pack and their results depicted that the improved system controlled the temperature well in an 8C-rate charging battery pack. system (CBMS), in which more than one of the above thermal management methods are applied. In this way, different methods can complement mutual strengths and weaknesses to present excellent performance. Ling et al. [263] developed a hybrid thermal management system for LIB-integrated PCM with forced-air cooling, as depicted in Figure 23. This CBMS successfully prevented heat accumulation and maintained the maximum temperature under 50 °C, in which the PCM dictated the maximum temperature rise and temperature uniformity within the battery pack while forced air convection succeeded in recovering the thermal energy storage capacity of PCM. Fathabadi et al. [264] proposed a novel battery pack design including a hybrid active-passive thermal management system. Their developed CBMS could be divided into two parts: the active part applied distributed thin ducts, air flow and natural convection as cooling media while the passive part utilized PCM/EG as the cooling component to optimize the thermal performance of the proposed battery pack. Xie et al. [265] developed a CPCM-based thermal management system coupled with air cooling and their results showed that the cooling effect of the integrated system was evidently better than that of the air cooling system. Furthermore, it was more energy-efficient due to the relatively lower air flow rate. Figure 23. Schematic of a combined battery management system (CBMS) integrated phase change material (PCM) with forced-air cooling. Adopted from Ref. [263].
As shown in Figure 24, Song et al. [266] investigated a CBMS that integrated liquid cooling and PCM together. Their results demonstrated that both the battery temperature ramp-up rate and the battery steady-state temperature were significantly reduced by the conjugated cooling, in

Firefighting
As displayed in Table 5, fires are generally classified into five different kinds (classes A, B, C, D and K) [276]. Nevertheless, due to the complexity of battery components and the diversity of usage conditions, the classification of LIB fire is controversial and it may vary among classes A, B, and C. In view of the uncertainty, there are currently no unified and specific requirements for the suppression To conclude, it is worth noting that both the application of FR additives and the thermal management on battery are compliant before the appearance of battery hazard, that is, the normal operation process. Even the safety device acts as a protective method for battery thermal hazard. Whereas, once thermal hazard occurs, it is essential to take certain countermeasures such as the warning of hazard and subsequent firefighting methods to reduce the damage inflicted by the hazard and suppress further deterioration.

Warning
There are two aspects from the perspective of early warning in order to solve the safety problem associated with LIB. The first is for a single battery; real-time monitoring of voltage, current, resistance and other data should take place prior to the occurrence of thermal hazard, allowing for an alarm to be raised when an abnormality appears. The second is to carry out management monitoring for a battery pack, that is, to analyze and alarm the abnormalities occurring during the operation of the entire battery pack. The early warning includes monitoring the changes in voltage, current, resistance and other electrical performance parameters, the temperature change and the escaping gases. An early warning and monitoring system is disclosed by Xie et al. [269] for battery packs. During normal cycling of a battery, surface temperature, voltage, current and impedance may be monitored to determine if abnormalities exist in the battery. Hermann invented a method for detecting LIB failure within a battery pack based on variations in the measured electrical isolation resistance of the battery pack [270]. Besides, escaping gas from LIB is becoming a growing concern since the volatile organics emitted are flammable and their unpredictable release represents a safety risk. Davion et al. [271] illustrated the results that the sensor could detect escaping gas prior to thermal events. With enough prior warning from the sensor, there is a desire to determine whether it provides suitable control at the fringes of battery performance. Somov et al. [272] applied an autonomous wireless sensor system for early fire and gas leak detection of LIBs. The system consisted of two modules: a gas sensor module and a power management module. At the same time, there have been more studies on the thermal runaway and fire releasing gases of LIBs; therefore, the number of gas detection methods that can be used for thermal runaway and fire warning has also increased [273][274][275]. The problem of thermal runaway and fire in LIBs is now more serious, therefore some advanced means are needed to predict the occurrence of disasters. Accurate warning of thermal runaway and fire shall be achieved through more quantitative data analysis.

Firefighting
As displayed in Table 5, fires are generally classified into five different kinds (classes A, B, C, D and K) [276]. Nevertheless, due to the complexity of battery components and the diversity of usage conditions, the classification of LIB fire is controversial and it may vary among classes A, B, and C. In view of the uncertainty, there are currently no unified and specific requirements for the suppression of LIB fires, and effective fighting technology for LIB fire is still a challenge. Suppressing LIB fire involves extinguishing the open flame and decreasing the battery temperature. If the battery temperature is sufficiently high after the open flame is extinguished, there is still a possibility that the battery will reignite.
To investigate how LIB fire can be suppressed effectively, researchers and institutions have conducted extensive studies to date. The National Technical Information Service (NTIS) assessed the fire hazards of LIBs and tested the extinguishing effect of different Halon products on battery fire [277,278]. Their results showed that the Halon extinguishing agents could suppress the open fire, while the battery temperature would still increase after the flame was extinguished. Later, Egelhaaf et al. studied the fighting effect of water combined with a surfactant and a gelling agent on battery fire [279]. They measured the battery temperature and the water consumption and found that water was capable of extinguishing the battery fire and additives helped to apparently reduce the amount of water required to fight the fire. Additionally, the Fire Protection Research Foundation (FPRF) performed a full-scale suppression test for the fire involving EV battery [280]. They proposed that although the water jet could quickly extinguish the battery fire, the toxic or flammable smoke and gas would still release after suppression. The FAA conducted experiments to screen the effective fire extinguishing agent for battery fire [281]. Experimental results indicated that water-based extinguishing agents (water, AF-31, AF-21, A-B-D, and Novec 1230 (C 6 F 12 O)) were the most effective, especially the evident cooling effect, while nonaqueous extinguishing agents (gas, dry powder, etc.) were the least effective. The Det Norske Veritas and Germanischer Lloyd (DNVGL) carried out a series experiments to evaluate the extinguishing effect of the F500, Fireice, PyroCool, aerosol and water on battery fire respectively [282]. Their results showed that all extinguishers used in this work could put down battery fire if they were sprayed immediately upon the detection of a thermal spike. Among, water exhibited the best cooling effect for a battery with thermal runaway. Hui et al. selected carbon dioxide, superfine powder and heptafluoropropane to conduct the battery fire suppression test series [283] in which heptafluoropropane was verified to be the most effective in controlling battery fires and suppressing the thermal runaway reactions. Further, Wang et al. investigated the efficiency of heptafluoropropane fire extinguishing agent (HFC-227ea), and they found that heptafluoropropane could suppress individual battery and even small-format battery pack fire [284]. However, the battery might be reignited due to the fiercely exothermic reactions inside battery after the fire was extinguished. Liu et al. investigated the suppression efficiency of dodecafluoro-2-methylpentan-3-one (C 6 F 12 O) in large-capacity battery fires [39]. They proposed that C 6 F 12 O first showed a negative inhibitory effect, which was then transformed into an inhibitory effect as the dose increased, and they also calculated the proper dose of C 6 F 12 O under their conditions. Besides, Luo et al. studied the fire extinguishing effect of a water mist containing additives system on a power battery [285]. Their results indicated that 5% F-500 solution and 5% self-made solution yielded a more significant extinguishing effect than pure water on battery fire. Table 5. Classification of fires.

Class Description
A Fires in ordinary combustible materials such as wood, cloth, paper, rubber and many plastics.
C Fires that involve energized electrical equipment. D Fires in combustible metals such as magnesium, titanium, zirconium, sodium, lithium and potassium.
K Fires in cooking appliances that involve combustible cooking media.
In summary, the previous studies mainly focused on the comparison of fire extinguishing effects among different extinguishing agents, such as Halon, water, carbon dioxide, foam, dry powders and so on. However, some deficiencies still exist and challenging questions are being proposed regarding fighting battery fire. For example, the spray time of the agent, dose of agent, damage degree of the battery, cost and scenario, etc. shall be quantitatively or qualitatively estimated. Furthermore, standard specifications and guidelines for the firefighting of battery fire shall be formulated and promulgated as soon as possible to meet the requirements of the market.

Summary and Outlook
With the global energy policy transitioning from fossil energy to renewable energy, LIBs, as a type of energy storage method, show great advantages over other kinds of batteries. Therefore, they have attracted considerable attention. However, the safety concerns regarding LIB are still a main obstacle that hinder large-scale applications of LIBs in many fields. Benefiting from the fast spread of information today, incidents of fire or explosions induced by LIBs are often reported ranging from mobile telephones and EVs to airplanes. These incidents greatly threaten consumers' lives and property, while damage human's confidence on LIBs, thereby restricting the progression of their use. Therefore, paying more attention to the thermal hazards of LIBs and the corresponding countermeasures is of great importance.
Generally, the thermal hazards of LIBs can be caused by several abusive factors, e.g., physical, electrical and thermal factors, manufacturing defect and battery aging. The physical factor can trigger electrical abuse, and the electrical abuse releases heat which will further induce thermal abuse; namely, thermal hazard and even thermal runaway. During the process of battery thermal hazard, a series of destructive reactions among battery components such as the decomposition of electrodes/electrolytes, the reaction between electrodes, the reaction between electrode and electrolyte, etc. are induced, substantial heat is released and quantities of combustible gases are generated. Besides battery components, the thermal hazards of a single battery and a battery pack are also reviewed. For the former, the thermal hazards that are generally exhibited are high-temperature, ejection, combustion, explosion and toxic gases during thermal runaway. While for a battery pack, thermal failure propagation provides a thermal hazard in addition to thermal runaway. Thermal failure propagation will aggravate the thermal hazard further, resulting in a serious incident. In addition, the influence of low-pressure environment and cathode chemistry on the thermal hazard of LIBs is discussed here. However, it still lacks sufficient examinations on the thermal behavior of LIBs under low pressure, the hazard difference between normal condition and low-pressure condition is worth noting in future research.
In an aim to decrease the thermal hazards of LIBs as much as possible, certain management or countermeasures must be implemented. In current work, the measures discussed were (1) the usage of safety devices inside battery; (2) the application of FR additives; (3) the thermal management of battery; (4) the warning of thermal hazard and (5) firefighting upon the occurrence of a hazard. However, challenges still remain. Effective FR additives and thermal management systems still need to be developed to meet the requirement of practices including performance, cost and so on. The current warning method is not sufficiently sensitive to give the humans the time necessary to escape danger or take the appropriate measurements. Moreover, standard guidelines for the fighting of battery fire remain to be formulated and promulgated. In short, to decrease the challenges of battery thermal hazards and create LIBs that better benefit humankind, there is still a lot of work to do.