Commercial fast-charging stations subject electric car batteries to high temperatures and high resistance that can cause them to crack, leak, and lose their storage capacity, write engineers at the University of California, Riverside in a new study published in Energy Storage.
To remedy this, the researchers have developed a method for charging at lower temperatures with less risk of catastrophic damage and loss of storage capacity.
Mihri Ozkan, a professor of electrical and computer engineering and Cengiz Ozkan, a professor of mechanical engineering in the Marlan and Rosemary Bourns College of Engineering, led a group that charged one set of discharged Panasonic NCR 18650B cylindrical lithium-ion batteries, found in Tesla cars, using the same industry fast-charging method as fast chargers found along freeways.
They also charged a set using a new fast-charging algorithm based on the battery’s internal resistance, which interferes with the flow of electrons.
The internal resistance of a battery fluctuates according to temperature, charge state, battery age, and other factors.
High internal resistance can cause problems during charging.
The UC Riverside Battery Team charging method is an adaptive system that learns from the battery by checking the battery’s internal resistance during charging.
It rests when internal resistance kicks in to eliminate loss of charge capacity.
For the first 13 charging cycles, the battery storage capacities for both charging techniques remained similar.
After that, however, the industry fast-charging technique caused capacity to fade much faster- after 40 charging cycles the batteries kept only 60% of their storage capacity.
Batteries charged using the internal resistance charging method retained more than 80% capacity after the 40th cycle.
At 80% capacity, rechargeable lithium-ion batteries have reached the end of their use life for most purposes.
Batteries charged using the industry fast-charging method reached this point after 25 charging cycles, while internal resistance method batteries were good for 36 cycles.
“Industrial fast-charging affects the lifespan of lithium-ion batteries adversely because of the increase in the internal resistance of the batteries, which in turn results in heat generation,” doctoral student and co-author Tanner Zerrin said.
Worse, after 60 charging cycles, the industry method battery cases cracked, exposing the electrodes and electrolyte to air and increasing the risk of fire or explosion. High temperatures of 60 degrees Celsius/140 degrees Fahrenheit accelerated both the damage and risk.
“Capacity loss, internal chemical and mechanical damage, and the high heat for each battery are major safety concerns, especially considering there are 7,104 lithium-ion batteries in a Tesla Model S and 4,416 in a Tesla Model 3,” Mihri Ozkan said.
Internal resistance charging resulted in much lower temperatures and no damage.
“Our alternative adaptive fast charging algorithm reduced capacity fade and eliminated fractures and changes in composition in the commercial battery cells,” Cengiz Ozkan said.
“The proposed adaptive fast charging provides a novel perspective for the design of fast charging technology for electric vehicles with better safety performance and longer battery lifespan,” Bo Dong, a doctoral student and paper co-author said.
The researchers have applied for a patent on the adaptive internal resistance fast-charging algorithm that could be licensed by battery and car manufacturers.
In the meantime, the UCR Battery Team recommends minimizing the use of commercial fast chargers, recharging before the battery is completely drained, and preventing overcharging.
Principles of battery fast charging
An ideal battery would exhibit a long lifetime along with high energy and power densities, enabling both long range travel on a single charge and quick recharge anywhere in any weather. Such characteristics would support broad deployment of EVs for a vari- ety of applications.
Unfortunately, the physics of each of these re- quirements results in tradeoffs ; for example, thicker electrodes needed for high energy density suffer more acutely from the con- centration and potential gradients resulting from fast charging [17,18].
The combination of physical properties of materials and devices with temperature dependent behaviour deﬁnes the oper- ating envelope for batteries. As ambient temperatures decrease, charge rates and recommended maximum voltages are typically kept low to improve safety and performance, making operating temperature a key barrier to fast charging .
The risk of lithium plating during charging signiﬁcantly increases with decreasing temperatures, impacting capacity retention. The temperature threshold below which lithium plating becomes likely depends on many factors, including the cell parameters, age and C-rate.
Although many authors have reported lithium plating at temper- atures below 25+C [19e21], it can also occur at higher tempera- tures, particularly when high C-rates or high energy density cells are used [22,23].
Additionally, the efﬁciency of fast charging equipment is often strongly dependent on temperature, with power conversion efﬁciencies of 50 kW chargers reported at up to 93% and as low as 39% for operation at 25+C and -25+C, respectively, primarily due to the derating of power levels requested by BMSs at lower temperatures .
Efﬁciency in this case is deﬁned as the ratio between the instantaneous DC power delivered by the charger to the vehicle and the instantaneous AC power supplied to the charger by the grid. The level of this temperature-dependency varies between charger models, with some still able to maintain about 70% efﬁciencies at -25+C and others failing to function at all .
This section looks to provide insights into the fundamental phenomena involved in fast charging and their resulting impacts.
A Li-ion battery typically includes a graphite anode, a lithium metal oxide cathode with a layered, spinel, or olivine structure, a liquid electrolyte containing a mixture of organic carbonates, salts, and additives, as well as copper/aluminium current collectors and a porous polymer separator.
Processes that take place within the battery, whether within electrodes or at key interfaces, are central to enabling reliable operation and fast charging  and are dependent on factors such as ion transport and temperature.
As shown in Fig. 2, when a Li-ion battery is charged, ions move from the cathode, through the electrolyte, to the anode. Key mechanisms that inﬂuence this journey are ion transport
1) through the solid electrodes,
2) across the electrode/electrolyte interface for both anode and cathode, and
3) through the electrolyte, including Liþ solvation and desolvation .
In an ideal situation, these primary phenomena involved in battery charging would be favoured. However, battery operating conditions can lead to a range of side reactions that compromise performance and lifetime.
In addition, the thermal behaviour of the battery is strongly dependent on the conditions: factors such as high charging/discharging currents,
high cell resistance or high concentration polarisation increase the heat generation rates, affecting the efﬁciency and safety of the process.
The anode receives a lot of the attention during the charging process, since multiple studies have explored the role of cathode degradation and cathode electrolyte interphase (CEI) layer growth and concluded that these processes, while important, are generally not rate-limiting for traditional Li-ion systems [22,25,26].
At the anode, lithium metal can begin to deposit on the surface of the graphite under certain conditions during charging. This can result from a lack of accessible intercalation sites within the electrode if lithium reaches the surface when the surface concentration is already high. Understanding the site accesibility is a complex pro- cess, with studies showing that lithium diffusion into the electrode along the edge plane of graphite is signiﬁcantly more likely compared to the basal plane of graphite [27,28].
Under certain circumstances, the lithium may continue to deposit preferentially where deposits already exist, forming needle-like structures known as dendrites which pose a risk of piercing the separator and causing a short circuit.
As the cell is repeatedly cycled, the dendrites may also break and become electrically disconnected from the anode, reducing the amount of cyclable lithium. However, even relatively uniform lithium plating removes signiﬁcantly more lithium from active availability than typical solid electrolyte interphase (SEI) layer formation on graphite, leading to dramatic capacity losses.
The reactivity of lithium metal and the chemical instability of electrolyte components near the Li/Liþ potential result in the for- mation of gas by-products and new SEI material, removing active materials from the system .
Factors that inﬂuence lithium deposition and resulting struc- tures include lithium diffusion rates within the anode , limited ion transport in the electrolyte leading to concentration gradients across the anode and salt depletion at the current collector , and reactions at electrode/electrolyte interfaces . Both mate- rials and cell designs have been studied in detail to manage the impacts of fast charging on these phenomena . Managing
internal resistances, including those associated with the transport through the electrolyte and electrodes as well as charge transfer at interfaces, is important to good charging characteristics, perfor- mance, and lifetime , with low temperatures increasing these transport resistances and requiring more sophisticated control strategies .
In studies of electrolyte additives to mitigate lithium plating under fast charging conditions, Liu et al.  found signif- icant differences in cell capacity losses between C/2 and 1C charging rates at 20+C which were attributed to lithium plating.
They concluded that many of the phenomena occurring at the anode could be combined into a single guideline relating the area speciﬁc resistance for the anode to the onset current of lithium plating for the conditions tested.
Additionally, work on charge transfer by Jow et al. demonstrated that reducing the anode resis- tance, which dominates overall cell resistance, is important for improving fast charging performance .
Regarding temperature, work by Waldmann et al.  explored the main aging mechanisms at both low (0+C) and high tempera- tures (45+C) at 0.5C in lithium nickel cobalt aluminium oxide (NCA)/graphite systems to better understand implications related to different aging mechanisms. Cells that experienced aging at higher temperatures by consumption of active materials through SEI growth show higher temperature onsets for thermal runaway in accelerated rate calorimetry testing than cells that have experi- enced low temperature aging even at relatively mild charging rates. While dendrites and the potential for short circuits are viewed as the typical hazard resulting from lithium plating, the reactivity of lithium metal is also a signiﬁcant consideration.
Reactions between lithium and the electrolyte not only contribute to the growth of the SEI layer but can also be highly exothermic and may play a part in the evolution of thermal runaway. Historically, high temperatures, such as those above 30+C were avoided for battery operation, due to the increased kinetics of SEI formation and other side reactions relative to functional reactions.
However, when batteries are sub- jected to extreme fast charging, higher temperature has a beneﬁcial effect in spite of the increase in kinetic rates of SEI formation, most likely by avoiding operating conditions that would induce lithium plating .
Finally, recent work by Yang et al.  explored the dependence of the optimal cell temperature on the charging C-rate and cell energy density. The results of their modelling study showed that while both extremely low and extremely high tem- peratures were generally damaging, fast charging shifted the bal- ance in favour of higher temperatures, particularly for high energy cells.
While the intersection of phenomena, materials chemistry and component design are described in the following sections, anode thickness and its inﬂuence on transport provides a useful lens to explore their linkages.
While thin electrodes are generally consid- ered to represent ideal transport , when electrodes are sufﬁ- ciently thick, it becomes critical to ensure sufﬁcient Liþ concentration at the electrode/electrolyte interface throughout the anode to keep the overpotential stable and reduce the chances for lithium plating [28,34].
Strong heterogeneities, whether in electrode microstructures  or in lithium concentration gradients observed during the charging process , have recently been shown to have signiﬁcant inﬂuence on lithium deposition reactions.
Non-uniform SOCs within graphite anodes have also been shown to inﬂuence lithium plating , with reduced pore tortu- osity highlighted as a key lever to avoid electrolyte phase transport limitations and the resulting overpotential increase [17,37].
In relatively thick anodes subjected to fast charging conditions, the lithium salt may become depleted near the current collector, leading to non-uniform electrode utilisation and an increase in the local current densities near the separator [17,29,31].
Local salt depletion at the anode has also been shown to inﬂuence the morphology of lithium deposition, inducing the formation of den- dritic structures in place of the more uniform mossy deposits formed when sufﬁcient amounts of lithium are available in the electrolyte [29,38].
To avoid the phenomena that result in plating or dendrite growth, Gallagher et al. recommended operating at current densities below 4 mA cm-2 but acknowledged the complex interactions of operating parameters in providing such guidelines . Due to the difference in mass loadings and other characteris- tics between cells, the general applicability of this recommendation may be questioned.
Alternative anode materials are also the subject of signiﬁcant research, to assist in addressing the challenges related with lithium plating on graphite anodes during fast charging described above, as well as to increase energy density. Since most studies have been conducted on graphite, much remains to be studied for other systems , including the effects of localised current densities on lithium metal anodes .
In a Li-ion battery, the heat generation occurs as a result of both reversible and irreversible processes . The irreversible heat generation Qirr is given by Ref. :
Qirr ¼ ðVbat – UÞI (1)
where U is the open circuit potential, Vbat is the cell voltage, and I is the current (I > 0 during charging). The difference between Vbat and U represents the total overpotential of a battery induced by processes such as the charge transfer reactions at the electrode/elec- trolyte interfaces , the diffusion and migration of Liþ ions across the electrolyte , diffusion and migration of Liþ ions in the electrodes  and Ohmic losses . A large contribution to the irreversible heat generation is resistive (joule) heating:
Qjoule ¼ I2R (2)
where R is cell resistance. As fast charging requires higher charging currents, more heat is generated due to the quadratic dependency of irreversible heat generation rate Qirr on the current. The heat
generated/consumed in the reversible process, also known as entropic heat, originates from the reversible entropy change DS during electrochemical reactions . Once the entropy change DS is measured [48,49], the reversible heat Qrev can be calculated ac- cording to :
battery lifetime roughly doubles when the average battery tem- perature (during storage and cycling) is reduced from 35+C to 20+C. The effects of lower temperatures were not explored.
3.1.1. Temperature distribution during cycling
In Li-ion batteries, whether pouch, cylindrical or prismatic, heat can dissipate from some locations more easily than from others: for example, the poor through-plane conductivity of battery materials such as polymer separators causes higher heat accumulation in the core compared to the regions closer to the surface. Furthermore, the current densities and the heat generation rates are not equal at different locations. These inhomogeneities are aggravated for large format cells. For cylindrical cells in particular, the temperature in the battery core is signiﬁcantly higher than at the surface  as shown in Fig. 3. For pouch  or prismatic  cells, the tem- perature is higher in the regions close to the tabs  as shown in Figs. 4 and 5 due to the higher current densities in those locatons. Furthermore, the temperature near the positive tab is often higher than near the negative tab due to the higher ohmic resistivity of the aluminium current collector on the cathode side compared to copper on the anode side. The heat generation rate among the different regions is not uniform due to the non-uniformly distrib- uted current during cycling. The non-uniformity of current density can be caused by a temperature gradient and the location of tabs [58,59]. High current density near the tabs may lead to local overcharge or overdischarge, potentially resulting in localised failure . The inhomogeneities in the temperature and current distribution can also lead to different local rates of side reactions and therefore different local degradation rates . Zhu et al.  recently used micro-Raman spectroscopy as a temperature-sensing platform to investigate the effects of temperature hot spots on lithium dendrite growth in Cu/Li and Cu/LCO optical cells, showing that high local temperature gradients dramatically increased local current densities, leading to dendrite growth and eventual short- circuiting during charging. In Li-ion cells with graphite anodes charged at low temperatures the relationship between hot spots and dendrite growth rates may be more complex due to the interaction between temperature and the local potential of the graphite anode.
Uneven heat generation is not solely a cell-level effect. Design decisions on the pack level, particularly those concerning the pack layout and the thermal management system design, have a strong inﬂuence on the temperature variations within a pack. Wu et al.
 used a coupled electrochemical-thermal model to investigate heat generation in battery packs containing cells connected in parallel. Finite interconnect resistances were shown to result in unequal interconnect overpotentials and, in consequence, lead to load imbalances and heterogeneous heat generation between cells.
These effects were aggravated during pulse loading due to the additional heat generation caused by rebalancing between pulses. Over time, different degradation behaviours between individual where T is the absolute temperature, n is the stoichiometric num- ber of electrons involved in the electrochemical reaction and F is Faraday’s constant. To understand heat generation in batteries, Nazari et al.  employed a mathematical model to simulate the heat generation in lithium iron phosphate (LFP), lithium manga- nese oxide (LMO) and lithium cobalt oxide (LCO) batteries with graphite anodes. The results revealed that the total heat generation in all cells investigated is of the same order of magnitude. At low C- rates the reversible heat generation is dominant and at high C-rates the irreversible heat is dominant. Li-ion battery lifetimes vary greatly with cell temperature. To understand this correlation, the US National Renewable Energy Laboratory has developed aging models [52,53] of Li-ion cells that consider the impact of temper- ature and charge/discharge cycle on battery life. It was found that
cells can also be expected to strongly affect the homogeneity of heat generation in an aged pack due to the unequal rise in cell re- sistances. The resulting temperature gradients may, in turn, further aggravate the differences in the aging behaviours between cells. The design of the thermal management system to minimise tem- perature inhomogeneity is discussed in Section 6.
Many degradation mechanisms that take place in Li-ion batte- ries show temperature dependence. The temperature effect is correlated with Arrhenius equations [20,63] or modelled with empirical equations ﬁtted to experimental data  in the past studies.
At high temperatures, the SEI layer on the anode grows faster, becoming more porous and unstable . At the other end of
the spectrum, low temperatures lead to slower diffusion and intercalation with the possibility of lithium plating and subsequent lithium dendritic growth . On the cathode, the temperature increases during cycling may lead to binder decomposition , phase transitions , metal dissolution and CEI growth.
The electrolyte, typically ethylene carbonate (EC)/dimethyl carbonate (DMC) (1:1 wt ratio)-LiPF6 (1 M), can decompose to release gaseous species, for example CO2 , when operated at elevated temper- atures. In general, most degradation mechanisms are accelerated at high temperatures.
Lowering the temperature can suppress the degradation rate but low temperatures also undesirably slow down the diffusion of active species and change the reaction chemistry, leading to accelerated degradation if metallic lithium starts depositing on the anode.
Furthermore, operation at low tempera- tures leads to lower energy efﬁciency  and more heat genera- tion due to the higher overpotential resulting from the increased transport limitations and slower kinetics.
The SEI growth at the anode/electrolyte interface is one of the main degradation mechanisms in Li-ion cells under most cycling conditions. The passive surface layer adds onto the cell impedance and diminishes the power.
The lithium consumed by the SEI is not recoverable, resulting in capacity fade. The electrochemical reduction of the electrolyte which generates the complex mixture of SEI with lithium species such as LiF, Li2CO3, (CH2OCO2Li)2, is highly temperature dependent .
The SEI layer formed initially may serve a protective function, but upon further cycling becomes unstable and grows to block pores of the electrode or even pene- trate the separator, which causes an impedance rise from the decrease in active surface area and safety issues.
At elevated temperatures (60+C or above), the dissolution and decomposition of SEI species, Li2CO3 for example , has been found to disrupt the integrity of the protective thin ﬁlm to form separate islands.
Besides forming the SEI on the graphite anode, thermal decomposition of the carbonate-based electrolyte yields many other possible products, including carbon monoxide, carbon diox- ide, hydrogen, ethylene, alkylﬂuorides, ﬂuorophosporic acids, etc. Gachot et al.  recovered lithium methyl carbonate from the separators in Li/chromium-based-oxide (Li/CBO) cells cycled at 55+C. LiPF6 is not stable at 60+C, and its dissociation into PF5 dis- solves the protective SEI and exposes fresh graphite surface .
The decomposition reactions are thermally activated and were observed to become signiﬁcant at 85+C by comparison of nuclear magnetic resonance (NMR) spectra of 1.0 M LiPF6 in EC/DMC/ diethyl carbonate (DEC) . The release of gaseous by-product during SEI formation and electrolyte decomposition also compro- mises the mechanical properties of the anode and the cathode.
As side-products from parasitic reactions accumulate, electrode delamination and particle cracking are also detrimental to a battery after prolonged exposure to elevated temperatures. Delamination was observed by Pieczonka et al. for a polyvinylidene ﬂuoride (PVDF) bound LiNi0.5Mn1.5O4 cathode from the Al current collector after storage in electrolyte at 60+C for three weeks and during cycling .
Volume changes in the electrode material, exacerbated by fast charging/discharging and temperature variation, induce stress which generates cracks and eventually delamination when the binders cannot hold the electrode fragments together . The carbon additives in the cathode have been found to be electro- chemically active towards PF- intercalation which leads to structural changes in the cathode, with more severe effects observed at 45+C compared to 25+C .
Other mechanisms which cause structural instability, for example the interaction of electrolyte oxidation products and corrosion from hydroﬂuoric acid (HF), are mostly accelerated by higher temperatures.
In the extreme cases when the temperature rises beyond the safety threshold, thermal runaway may occur. Thermal runaway is a process whereby an exothermic reaction rate and the resulting temperature increase run into an uncontrolled positive feedback. The series of reactions during thermal runaway include SEI decomposition, anode/electrolyte and cathode/electrolyte reaction, electrolyte decomposition and reaction with the binder .
The breakdown of the SEI layer due to either overheating, overcharging or physical abuse is associated with exothermic decomposition of metastable components such as (CH2OCO2Li)2, followed by further exothermic reactions between the anode and the electrolyte .
Common polymer separators melt at above 130+C, although some separators are designed to maintain structural integrity and pre- vent a short circuit.
For example, trilayer polypropylene- polyethylene-polypropylene (PP-PE-PP) separators can be engi- neered to exhibit a shut-down process when temperature rises beyond a threshold. As the PE layer melts down, the separator becomes nonporous and is no longer ionically conductive, stopping the electrochemical function of the cell.
At the same time, the PP layers maintain their mechanical properties due to their higher melting point, preventing separator shrinkage and a short circuit .
Alternatively, a method of inducing thermal shut-down by introducing an extra material layer between the cathode and the current collector has been proposed by Chen et al.  The material consisted of graphene-coated spiky nickel nanoparticles in a polymer matrix and exhibited high electrical conductivity at moderate temperatures.
However, at high temperatures (70+C in the study, but tunable depending on the choice of polymer matrix and nickel-to-polymer ratio) its electrical conductivity dropped by seven to eight orders of magnitude within a second, effectively shutting down the cell and preventing thermal runaway.
If thermal runaway is not successfully stopped before the separator meltdown occurs, the cell is internally short-circuited, and consequently the cathode breaks down. During thermal runaway, electrolyte decomposition at 250 e 350+C is severe with rapid release of gaseous species, building up the pressure of the cell and eventually venting ﬂammable vapour into the environment .
Fast charging strategies
While many of the material level solutions discussed in Section
4.1 do indeed show promising results, most are not expected to reach the market on a wider scale in the near future. Many re- searchers have therefore turned towards cell and pack level ap- proaches which can often be implemented in real world systems in a signiﬁcantly shorter time. Charging strategies, which determine how the current density is varied during the charging process, are an important category of such solutions.
Types of charging protocols
Standard protocols. Numerous charging protocols have been proposed for Li-ion batteries. CC-CV is by far the most common one. It consists of a constant current charging phase where the battery voltage increases up to a cut-off value (CC phase), followed by a constant voltage hold until the current falls to near-zero (CV phase).
The CV phase allows for the concentration gradients within the electrode particles to disperse and is usually necessary to obtain high capacity utilisation without exceeding the maximum voltage. However, since the current gradually decreases during the CV phase, the charging time is signiﬁcantly increased compared to CC- only charging.
The simplicity and ease of implementation of CC-CV has made it the standard charging protocol in most applications. However, many other protocols have been shown to achieve reduced charging times, increased efﬁciencies and/or improved capacity or power retention. Fig. 8 illustrates some of the alternative protocols proposed for fast charging.
Zhang et al.  studied the effects of the CC-CV charging protocol on the reversible capacity and anode potential evolution in experimental LCO/graphite cells with embedded reference elec- trodes as they were charged at ambient temperatures from -20+C to 10+C with C-rates between 0.16C and 1.2C.
The results showed a positive correlation between increasing charging current in the CC phase and increasing time taken by the CV phase to complete. The authors observed that increasing the current beyond a certain value did not bring any further reduction in overall charging time due to this correlation.
Both increasing the charging current and decreasing the ambient temperature was shown to result in a reduction in anode potential, with all test schemes causing the anode potential to become negative versus Li/Liþ at some point during charging.
Negative anode potential was assumed to inevi- tably cause lithium plating, which was suggested as the main reason for the reduction in charged capacity at high currents and low temperatures.
However, the occurrence of plating was not conﬁrmed through other methods. Ouyang et al.  demon- strated that the shape of the voltage proﬁle of large format LFP cells changed signiﬁcantly with the number of cycles at 0.5C CC-CV at -10+C, resulting in an elevated average voltage in the CC stage and longer duration of the CV stage.
In comparison, the voltage proﬁle of cells charged at 0.2 and 0.3C at the same temperature retained a similar shape. The increased average voltage as a result of repeated cycling at higher C-rates may contribute to increased degradation rate.
Multistage constant current (MCC) protocols. Many researchers have proposed that adjusting the current levels during the charging process may limit cell degradation while reducing the charging time. Such approaches are often motivated by reducing heat gen- eration, avoiding conditions that enable lithium plating or reducing mechanical stresses when the diffusion of Liþ ions is constrained.
MCC protocols are one of the earliest types designed speciﬁcally for fast charging. Such protocols consist of two or more constant cur- rent stages, often followed by a CV stage. Higher current levels are usually chosen for the earlier CC stages since the anode potential is less likely to become negative in the beginning of the charging process. Nevertheless, some authors have used the opposite approach with current levels increasing in later CC stages due to the lower cell resistance.
Zhang  observed faster capacity loss in cells charged with the latter MCC-CV approach compared to CC-CV and constant power – constant voltage CP-CV protocols with the same average C-rate of 1C. The study concluded that CP-CV resulted in best capacity retention when the cell was fast charged (1C), while CC-CV was less damaging for cells charged at 0.5C. Another work  reported an 83% capacity retention after 4200 cycles in an LFP cell charged with a 25-min MCC-CV protocol with two CC stages of decreasing current. Waldmann et al.  studied the anode potential evolution in reconstructed commercial NCA/graphite cells.
Three MCC charging protocols were then tested in fresh cells of the same type. Current level was decreased just before reaching a voltage at which negative anode potential was observed in the reconstructed cells. All three protocols were faster than 0.25C CC- CV protocol and led to a comparable rate of capacity fade.
How- ever, this protocol design method is both time consuming and expensive as it requires reconstructing the cell with a reference electrode and measuring the anode potential in the range of tem- peratures and currents of interest in the practical application.
Spingler et al.  postulated that lithium deposition leads to local volume changes which can be directly measured in pouch cells in operando using laser triangulation, removing the need to measure the anode potential. Based on this assumption, they designed an MCC protocol with multiple stages in which current level was decreased every time the maximum local expansion reached a predetermined value. Cells cycled with this protocol experienced dramatically improved capacity retention compared to a CC-CV protocol with the same average current. However, the method still requires extensive experimentation and is limited in applica- tion to pouch cells.
Pulse charging protocols. Pulse charging protocols, where the charging current is periodically interrupted by short rest periods or discharge pulses, are also common in literature. The strategy aims to reduce concentration polarisation, reducing the risk of local anode potential becoming negative or reducing mechanical stresses due to uneven insertion and extraction of lithium in the solid particles.
Aryanfar et al.  used a Monte Carlo simulation of lithium dendrite growth to show that pulse charging can also potentially inhibit dendrite propagation. One of the early studies  reported that a 1C pulse charging protocol could reduce the charging time from 3.5 h for a 1C CC-CV protocol to approximately 1 h due to the absence of the CV stage that accounted for most of the total charging time in CC-CV.
Higher discharge capacities were observed for the pulse protocol relative to CC-CV which was attributed to better active material utilisation. However, since the capacities of the cells were not determined in a standard charac- terisation test at the beginning of life and the rate of capacity fade did not vary substantially, the discrepancy in capacities could be caused by manufacturing differences.
The same study  compared SEM images of both electrodes after 300 cycles, ﬁnding cracks in the cathode particles of all cells but signiﬁcantly less SEI formation on the anodes of the pulse charged cells. Abdel-Monem et al.  monitored capacity fade and impedance changes in LFP cells cycled with eight different fast charging protocols, all belonging to CC-CV, MCC or pulse charging categories.
They observed comparable rates of capacity fade until the 700th cycle, when the capacity of the cell charged by CC-CV started deterio- rating considerably faster. The rate of capacity fade between pulse charging and MCC remained similar, however a higher rate of impedance rise was observed for the MCC protocol than for pulse charging.
The discrepancy in the rates of capacity fade and impedance rise was interpreted as a manifestation of different aging mechanisms occurring for each protocol. Chen et al.  proposed a sinusoidal-ripple-current (SRC) charging strategy using the minimum impedance frequency signal to minimise heat gen- eration.
Experiments revealed that the SRC strategy could offer signﬁcant improvements in charging time, efﬁciency, temperature rise and lifetime compared to 1C CC-CV, and slight improvements compared to a square pulse charge with the same parameters. The method was only validated with one type of cell at a relatively high
starting temperature of 28.5+C, and it is unclear if the ambient temperature was controlled in any way during the tests. Amanor- Boadu et al.  used the Taguchi orthogonal arrays method to optimise the pulse charging parameters to maximise charging ef- ﬁciency, again ﬁnding the minimum AC impedance frequency to be optimal for charging.
Signiﬁcantly shorter charging times and higher efﬁciencies were reported for the pulse charging protocol compared to CC-CV with the same average current in the CC phase at 23+C, however the temperature rise was higher and the effects on cycle life were not studied. Yin et al.  designed a module charger that performed pulse parameter optimisation on-line to limit battery polarisation. The authors implemented the charger on four cells connected in series, showing that the charge time could be reduced compared to 2C CC-CV, although the resulting tem- perature rise was slightly higher.
No cycle life experiments were performed, while the complexity of implementation was acknowledged by the authors. As mentioned in Section 3.1, pulse loads may lead to increased and uneven heat generation in packs connected in parallel due to the rebalancing events between pulses .
The implications of these effects have not been extensively studied in the context of charging protocol optimisation.
Boostcharging. Boostcharging is characterised by high average current in the beginning of charge, followed by a CC-CV part with more moderate currents. The ﬁrst, boostcharge stage could simply comprise a CC proﬁle (making the protocol identical to MCC-CV), a CV proﬁle where the cell is immediately brought to a set maximum voltage by means of high initial current (CV-CC-CV), or an entire CC-CV proﬁle (CC-CV-CC-CV) .
The boostcharge stage should, in any case, allow higher currents or higher maximum voltage compared to the following CC-CV part in order to reduce the overall charging time. Notten et al.  tested a few variations of the boostcharge protocol on both cylindrical and prismatic LCO cells with a 5 min boostcharge period.
Compared to a 1C CC-CV protocol, the charging time was reduced by about 30e40% with no notice- able acceleration in capacity fade for cylindrical cells. For prismatic cells, a smaller reduction in charging time was reported along with slightly higher rates of capacity fade. Interestingly, some of the boostcharging protocols seemed to achieve marginally better ca- pacity utilisation at the beginning of life compared to the slower 1C CC-CV.
It is, however, unclear whether this could be due to manufacturing differences as cell capacities were not measured by a standardised test at the beginning of life. Keil and Jossen  arrived at contradicting conclusions after experimentally comparing cycling data from three types of 18650 cells of different chemistries charged by CC-CV, boostcharging and pulse charging protocols. An increased rate of capacity fade was reported for boostcharging compared to CC-CV with a similar charging time, while no signiﬁcant differences in capacity fade rates were observed for pulse charging.
The authors concluded CC-CV was suitable for fast charging high power cells, however mentioning that MCC could be useful in conditions when lithium plating is likely to occur.
Variable current proﬁles. A number of more complex variable current proﬁles have also been proposed for fast charging. Sikha et al.  investigated a Varying Current Decay VCD) protocol, designed to charge faster than a conventional CC-CV protocol while being less damaging than pure CV charging.
The proposed protocol achieved improved capacity utilisation in the early cycles, but still resulted in a signiﬁcantly higher rate of capacity fade compared to CC-CV with a similar average current. Possible overcharge of the cathode was suggested as the reason for accelerated degradation.
Building on a similar idea, the Universal Voltage Protocol (UVP)  was designed to reduce both charging time and energy losses due to heating, thus improving charging efﬁciency. The UVP is derived from a set of CC-CV charging curves using an optimisation algorithm for a speciﬁed charging time and target terminal voltage.
A variable current proﬁle is then calculated from the UVP and cell resistance. As the cell ages, the current proﬁle needs to be recalculated due to changes in resistance while the voltage proﬁle re- mains the same as for a fresh cell.
Irrespective of the cell age, the current is always very low and rapidly increasing in the beginning of charge due to the cell resistance being highest at 0% SOC and then rapidly dropping. Maximum current is reached at relatively low SOCs, and subsequently it is gradually reduced due to the mass transport of Li becoming more constrained in increasingly lithiated graphite particles.
The study demonstrated that a very high charging efﬁciency could be maintained even for highly aged cells. The capacity of the tested LCO/NMC cell dropped to 80% after 370 cycles with the UVP protocol compared to only 100 cycles achieved by the 2C CC-CV protocol with a similar charging time.
While the voltage proﬁle was shown to be universal with respect to cell age, the effects of the charging conditions, most notably temperature, on the cycling performance were not investigated. Ye et al.  developed a multistage constant heating rate charging protocol aiming to minimise charging time and temperature rise while maintaining the same charge capacity as in CC-CV.
The heating rate was decreased in stages as the battery SOC increased, producing a variable current proﬁle. Reduced charging time and temperature rise compared to CC-CV were reported at ambient temperatures ranging from 10+C to 40+C, although the effects on cycle life were not studied. Another charging strategy concept, Constant Temper- ature – Constant Voltage (CT-CV) was proposed by Patnaik et al. .
The temperature was set to a pre-deﬁned maximum value during the CT stage and then decreased during CV, producing a current proﬁle similar to boostcharging. The method was demon- strated to achieve about 20% shorter charge time compared to CC- CV with the same temperature rise, however the cycling performance was again not investigated.
It should be stressed that while temperature plays a key role in determining degradation rates, the assumption that higher temperature rise is always detrimental may not apply in all cases. Temperature gradients within packs and cells are crucial but often not considered in charging protocol studies that use surface temperature as the main indicator of degradation rate.
Schindler et al.  took the novel approach of combining previously published physically-motivated fast charging strategies. Four current-neutral current proﬁles were superimposed, sepa- rately and in combination, on a 1C CC-CV protocol.
The proﬁles are shown in Fig. 9 and included:
a) AC pulse, motivated by a decrease in cell resistance due to additional irreversible heating;
b) cold derating, a proﬁle in which a very low current is used initially due to high resistance at 0% SOC and the current is brought to a higher constant level as soon as the resistance drops;
c) overpotential reserve, a proﬁle in which the overpotential is always maintained below a certain safety limit to prevent negative anode potential, producing a gradually decreasing current proﬁle; and
d) pulse charging. Capacity fade was monitored in eighteen cells charged by different protocols, and compared to four cells charged by standard CC-CV. The best performing cell, charged by a combination of all proﬁles, retained 80% capacity after 800 cycles.
In comparison, cells charged by CC-CV alone experienced the same capacity fade after only about 400 cycles, while the worst performing cell, charged by CC-CV combined with cold derating, completed about 330 cycles.
In general, protocols that included the overpotential reserve proﬁle resulted in best capacity retention. All tests were conducted at a temperature of 25+C and the results should not be generalised to other conditions.
While many fast charging protocols have been proposed based on an array of physical motivations, most of them have only been validated at standard temperatures and for certain cell chemistries or form factors.
Since high currents induce higher mechanical stresses in electrode particles along with more severe current and temperature distributions, care must be taken when generalising the results of experimental studies to different cell types.
At the moment, the applicability of many charging protocols to different conditions is difﬁcult to establish without further experiments, which are often time consuming and expensive to conduct. There is a clear need for accurate cell and pack models that would enable the design of charging protocols without the need for extensive and complicated laboratory testing.
As the EV industry grows in colder climates, more research on charging protocols suitable for low temperature fast charging will also be necessary. More consider- ation should also be given to the cell temperature proﬁles induced by different charging protocols, as it is the cell temperature rather than ambient that determines performance. Finally, studies on the effects of the fast charging protocols discussed in this section at the pack level are still lacking.
Implications on thermal management
Section 3.1 outlined the effects of fast charging on the temper- ature uniformity within cells and packs along with potential local degradation effects that may evolve. Fast charging is normally accompanied by high heat generation rates and signiﬁcant in-homogeneities.
At the same time, high charging currents applied at low temperatures may be detrimental to battery lifetime and safety. As such, effective and ﬂexible thermal management strate- gies are critical to enabling fast charging in all conditions.
The requirements on the Battery Thermal Management Systems (BTMSs) can vary greatly depending on temperature. While high thermal conductivity is crucial when cooling a hot battery pack, low tem- perature operation prompts the desire for good insulation to retain the heat produced by the battery. Passive thermal regulators that adjust their thermal conductance depending on temperature could be one way to address this problem .
The cooling media commonly used for EV battery packs can be divided into air, liquid and phase change materials (PCMs). Air cooling systems are low cost and relatively simple, but fail to ach ieve sufﬁcient cooling rates or good temperature uniformity due to the low heat capacity and thermal conductivity of air [234e236] and are therefore unsuitable for fast charging.
Liquid cooling can be 3500 times more efﬁcient than air  but its drawbacks include high cost, complexity and potential of leakage. Liquid immersion cooling, whereby batteries are immersed directly in the cooling medium, has been receiving attention from researchers due to its effectiveness and ability to achieve good temperature homogenization .
To prevent short circuits, it is vital that the cooling liquid is dielectric. Examples of such liquids include deionised water and mineral oils. While liquid immersion cooling has not yet been widely adopted in EV battery packs , some researchers have shown promising results.
For example, the two-phase thermal management liquid Novec7000 (3 M, USA) has been shown to achieve very good surface temperature uniformity in an immersion-cooled cylindrical Li-ion battery, particularly when the battery temperature was high enough to induce boiling . The same liquid has also been reported to effectively maintain the temperature of a battery module around 35+C even at a 20C discharge rate .
In PCM cooling, the latent heat of phase change of the cooling medium is used to absorb heat produced by the battery. However, the method has signiﬁcant drawbacks: at high ambient temperatures, the PCM could melt completely without any heat being produced by the battery, and the low thermal conductivity of the liquid PCM would then act as a barrier to heat transfer . On the other hand, Hmery et al.  coupled PCM with a liquid cooling system to enable solidiﬁcation of the melted PCM.
The coupled system was shown to effectively cool a battery module during a 2C fast charge.
Since fast charging inevitably leads to higher temperature gra- dients, both within a pack and within a cell, effective and uniform cooling becomes even more critical than under standard charging conditions.
Heat generated during charging is more difﬁcult to dissipate from the centre of a cell compared to the regions closer to the outer surfaces due to poor thermal conductivity across the electrodes .
At the same time, it is the outer surface of cells that is typically in contact with the cooling medium, further amplifying the temperature gradients. Similar observations apply to battery modules and packs.
The dense packing of cells in EV battery packs to minimise the volume aggrevates the issue , making the cooling system design critical to pack safety and longevity. Lu et al. modelled the performance of a forced air cooling system with either 15 or 59 air ﬂow channels, showing the latter option to achieve more uniform temperature distribution .
The diameters of the channels were adjusted so that both systems required the same volume within the pack and the same heat ﬂux was imposed. Xia et al.  reviewed a number of cooling system conﬁgurations, concluding that parallel and mixed series-parallel conﬁgurations were more effective in minimising temperature gradients between cells compared to parallel conﬁgurations in ﬂuid (air or liquid) cooled packs. However, cooling efﬁciency and uni- formity are sometimes in contradiction.
Hunt et al.  reported substantially lower rates of capacity fade for tab cooled pouch cells compared to surface cooled ones, which was attributed to lower temperature gradients between cell layers. Zhao et al.  modelled the effects of surface and tab cooling on the temperature distribution in a prismatic Li-ion cell.
The study demonstrated that tab cooling resulted in a substantially more uniform temperature distribution across the cell thickness compared to surface cooling. However, surface cooling was able to reduce the average cell temperature more effectively than tab cooling. The model assumed a set coolant temperature for both methods, which may not be representative of the conditions in operation.
Finally, recent developments in increasing EV charging powers have been accompanied by some interest in external cooling technologies that could be provided by the charging stations. Such an approach, if successful, could help reduce the cost and weight of onboard cooling systems.
Ford Global Technologies LLC have applied to patent a charging station that communicates with the connected EV and supplies cooled air to the vehicle radiator during fast charging . Lightning Energy patented a charging station with coolant pipes integrated into the charging connector .
The connector locks in place to prevent coolant spillage if removed, and coolant is cycled through by the controller in the charger in response to battery information from the car. Tesla patented a similar idea, with an automated charging connector placed beneath the vehicle and being capable of supplying both cold and hot liquid to optimise the battery temperature during charging .
None of these technologies have been commercialised at the time of writing and their ability to provide sufﬁciently uniform cooling for fast charging applications has not been proven. Leakage of the thermal ﬂuid is a potential issue.
Preheating in cold conditions
As explained in earlier sections, low temperature fast charging of Li-ion cells proves particularly difﬁcult. While relatively few studies have attempted to address the issue through designing appropriate charging protocols, signiﬁcant research effort has been dedicated to exploring different preheating strategies.
In this section, only the methods that can result in rapid heating of the entire cell are discussed. This is because high speed is an imperative requirement for any preheating strategy that could be integrated with fast charging.
Internal preheating methods are favourable due to higher efﬁciency (as less heat is lost to the surroundings) and better unifor- mity [246,247]. Ji and Wang  used a coupled electrochemical- thermal model to compare the performance of four heating methods: self-heating by discharging the battery, convective heating using a fan and a resistance heater powered by the battery, mutual pulse heating, and AC heating.
Mutual pulse heating is a strategy whereby a battery pack is divided into two groups of equal capacity and charge is exchanged between the two groups in pul- ses, taking advantage of the resistance to generate heat.
Self-heating by discharging was found inefﬁcient, as can be expected considering the cell generates both heat and power during discharge. Convective heating resulted in relatively fast but inefﬁcient and non-uniform heating.
The efﬁciency of the mutual pulse heating strategy was, on the other hand, shown to be much higher and mainly limited by the efﬁciency of the DC/DC converter used. Only 120 s were needed to raise the temperature of the simulated 2.2 Ah 18650 cell from -20+C to 20+C using this method.
AC heating could reach even higher heating speeds, achieving the same tem- perature rise in 80 s with a 10 mV sinusoidal voltage wave at a frequency of 1000 Hz. However, the impacts on degradation and cyclability were not studied.
For AC heating, the time to reach the desired cell temperature can theoretically be minimised by using high current amplitudes, yet exceeding a safe current limit can lead to non-uniformity (as the resistance may vary within the cell), exceeding maximum cell voltage, and to the anode potential becoming negative . Geet al.  used an ECM-type model to estimate the maximum alternating current that could be applied depending on the frequency and temperature while maintaining a positive anode po- tential and therefore preventing lithium deposition.
The study found that the maximum permissible current amplitude increased with increasing temperature and signal frequency. Based on these observations, a method for designing a stepwise AC heating strategy with the current amplitudes increasing with the average cell temperature was proposed.
The frequency of the signal was kept constant at 100 Hz as it was demonstrated that gradually increasing the current had a stronger effect on the heating rate than lowering the frequency.
The 1 Ah pouch cell used in the study could be heated from -20+C to 5+C in 800 s according to both the model and experimental data. However, the effects on the cyclability were not reported. The method for designing the AC proﬁle required placing a reference electrode within the cell in order to obtain EIS spectra of both electrodes needed to ﬁt the model.
The authors argued that preparing and testing one cell per cell type would not be a considerable cost to manufacturers. However, even assuming all cells of a certain type can be reproduced with the same equivalent circuit with the same parameters, it is crucial to note that these parameters would change as the cells age, and may change differ- ently for different cells.
Another study  proposed a similar AC heating proﬁle with gradually increasing current amplitudes, again conﬁrming that adjusting the amplitude was more beneﬁcial than adjusting the frequency as the cell temperature increased.
Based on an electro-thermal coupled model, the optimal frequency resulting in highest heat generation for the study case was calculated at 1377 Hz, slightly lower than the minimum impedance frequency.
The frequency optimisation process used a simple model requiring only overall cell data without the need for a reference electrode. However, it was assumed that only cell voltage limits needed to be adhered to; the effect of the AC signal on the anode potential was not considered.
The commercial 18650 cell of 2.75 Ah capacity was heated up from -15.4+C to 5.6+C in 338 s. No signs of aging were observed in incremental capacity (IC) curves after 300 heating tests. Capacity retention in preheated and subsequently cycled cells was not explored. Zhu et al.  experimented with different frequencies and amplitudes for a preheating AC proﬁle on a 30 Ah LFP pouch cell, although the maximum amplitudes were low relative to the cell capacity (60 A, or 2C).
With amplitudes limited to this value, a low frequency signal was more successful in heating the cell rapidly compared to higher frequencies. The most effective AC proﬁle tested was able to heat the cell from -25+C to 5 +C in 1800 s using a frequency of 0.5 Hz and an amplitude of 60 A. SEM images of the anode after 240 heating tests did not feature signs of lithium deposition.
While these were compared to images of an anode subjected to 36 DC charge/discharge cycles at 0.5C and -25+C which showed a clearly altered morphology, no attempt was made to study a cell subjected to the AC preheating procedure with subsequent cycling.
A thermocouple inserted in one of the cells enabled the authors to quantify the variation between the internal and surface temperature of the AC heated cell at about 5+C, providing a useful insight into the homogeneity of the AC heating method.
Designing Li-ion cells speciﬁcally to enable rapid preheating is another approach to address low temperature fast charging. Electrically insulated thin nickel foils inserted between two single- sided anode layers of a cell provide a means to raise the tempera- ture when necessary using standard direct current [16,250,251].
Current through the foils can be controlled using an activation switch. Turning the switch on directs the current to the nickel foils, causing rapid heat generation. When the switch is in the ‘off’ position, the current ﬂows through the electrodes only, and the nickel foils are electrically isolated. Yang et al.  modelled this cell design, using 10 Ah and 40 Ah prismatic cells as examples. The study demonstrated that heating uniformity and speed improved substantially with the number of nickel foils inserted, even when the total thickness of the foils was unchanged.
The total thickness of nickel foils was 200 mm which reduced the volumetric energy density by only about 0.58%. The same group  followed up with an experimental study on a 9.5 Ah NMC622 pouch cell with two nickel foils.
The results showed that the cell could be preheated and charged to 80% SOC in 15 min even from -50+C. Cycling tests at 0+C ambient revealed a cycle life of 4500 cycles before reaching 20% capacity fade for a cell preheated and subsequently fast charged at 3.5C, compared to only 50 cycles for a cell that was not preheated.
Although it is generally accepted that internal heating methods offer both better efﬁciency and more homogeneous temperature distribution, still little research has been done to evaluate the effects of internal preheating in conjunction with fast charging on cycle life. Since current ﬂows preferentially through paths of lower resistance and therefore higher temperature, it is possible that even small temperature gradients produced by internal preheating methods would be exacerbated if immediately followed by a fast charge. Since internal temperature is difﬁcult to measure experimentally, cycle life testing or reliable models will be needed to fully assess the performance of AC preheating.
Nickel foil preheating, while potentially promising, would require a non-standard cell design with additional weight, instrumentation, and possible reliability issues. As the technology was proposed only recently, more research on its performance and cost will be needed before po- tential commercialisation.
Effects of fast charging on thermal runaway characteristics
Lithium plating [30,96,252] and the temperature rise due to the heat accumulated in the cells or bus-bars at the end of fast charging [253,254] pose potential safety risks in addition to causing accel- erated degradation.
Research has shown that the thermal runaway behaviour of batteries changes after experiencing fast charging .
By conducting ARC tests on a fast-charged high energy pouch battery, it was found that the self-heating temperature and the thermal runaway triggering temperature drastically reduced for cells sub- jected to fast charging compared to fresh cells.
These effects do, however, seem to be reversible if sufﬁcient rest time is allowed. Waldmann et al. [256,257] conducted ARC tests on batteries rested for different periods of time after fast charging. While a signiﬁcant change in thermal runaway characteristics was observed for cells after only a short rest time, the thermal runaway behaviour of cells subjected to long rest times was similar to that of a fresh cell. This phenomenon can be explained by the reintercalation of lithium into the anode during the rest time, as well the reaction of the plated lithium with electrolyte to form new SEI [96,115].
With the rest time extending, the active plated-lithium content participating in the thermal runaway process is reduced, and the thermal runaway characteristics of the battery recover to their normal state.
Thermal runaway is known as being composed of a series of chain reactions . In a fresh battery it is generally triggered by an internal short circuit, and the maximum temperature is quickly reached with the participation of the electrolyte [259 e 266].
To clearly describe the chain reactions of fast-charged batteries, the evolution process of thermal runaway can be divided into three stages, as shown in Fig. 10. For the fast charged battery which ex- hibits abnormal thermal runaway behaviour, the reaction between lithium and electrolyte is dominant in the thermal runaway pro- cess, as opposed to that of fresh batteries.
In the ﬁrst stage (60+C < T < 110+C), the plated lithium reacts with the electrolyte
and heats the battery. The SEI ﬁlm on the plated lithium surface is continuously decomposed and regenerated, while the temperature rise rate remains relatively low. In the second stage (thermal runaway triggering process), the plated lithium is consumed in a large amount in the violent reaction with the electrolyte, causing a sharp increase in temperature.
The separator collapses, and the cathode and anode connect with each other. In the third stage (thermal runaway developing to the highest temperature), other reactions are involved. The reaction of the anode and electrolyte, the reaction of the anode and the cathode, and the reaction of the cathode and electrolyte are triggered due to the sudden rise of temperature. As a result, the battery reaches the maximum
temperature of thermal runaway.
Overcharge-induced thermal runaway
Some cells in a fast-charged battery pack may become over- charged due to the inconsistency among the cells [34,267,268], potentially leading to thermal runaway [269,270]. This process consists of four stages .
Stage 1 (100% < SOC < 120%): The battery voltage exceeds the cut-off voltage and increases slowly. The excess anode material, normally incorporated in Li-ion cells for safety, can still withstand excessive embedded lithium in the beginning of overcharging [271 e 273].
Some side reactions of the battery material may be triggered and the temperature and internal resistance of the bat- tery slightly increase.
Stage 2 (120% < SOC < 140%): A dissolution reaction of transition metal ions such as Mn2þ may be triggered in positive electrodes of certain chemistries, due to the excessive deintercalation of lithium [274,275].
At the same time, electrolyte oxidation may also begin since the battery potential exceeds the stable operating window. The negative electrode of the battery gradually becomes unable to withstand further transfer of lithium atoms, leading to signiﬁcant amounts of Li depositing on its surface [275,276].
The deposited lithium reacts with the electrolyte to form new SEI ﬁlm, increasing the internal resistance of the cell [276,277]. The joule heating effect of the overcharge current causes a signiﬁcant increase in the bat- tery temperature.
Stage 3 (140% < SOC < 160%): The heat produced by the exothermic reactions of the battery material begins to be compa- rable to and subsequently dominant over the joule heat due to the further removal of lithium from the positive electrode, the large deposition of lithium in the negative electrode, and the increase of the electrode potential.
Oxidative decomposition of the electrolyte produces a greater amount of heat [278,279], accompanied by the generation of gas, causing the battery to expand. The lithium- electrolyte reaction also becomes more pronounced as the amount of lithium deposition increases. When the SOC is close to 160%, a large amount of Mn2þ is dissolved if present in the cathode material; the SEI ﬁlm gradually collapses [280,281]; the structure of the cathode material changes, causing the battery voltage to reach a maximum value and then begin to decrease [282 e 284].
Stage 4 (140% < SOC < 160%): The oxidative decomposition of the electrolyte generates a large amount of gas, resulting in the battery rupture instantaneously. The battery separator is displaced by the shock, and a large internal short circuit occurs in the battery. The contact between the positive and negative electrodes of the battery produces a severe redox reaction . Thermal runaway occurs eventually.
High-precision thermo-electrochemical coupling models have been created to investigate overcharged batteries, based on the amount of internal materials and reaction kinetic parameters of the battery, and proposed two design methods to help protect the battery from overcharging :
1) Raising the electrolyte oxidation reaction potential from 4.4 V to 4.7 V. This results in a more stable electrolyte and the SOC of thermal runaway increased to 183%. It can be achieved by adding functional additives or redox shuttle additives [285,286], i.e. chemical species that can be reversibly reduced and oxidised at potentials slightly higher than the maximum safe cathode poten- tial, thus preventing it from increasing further . Due to the range of requirements for such compounds, including stable elec- trochemistry over many cycles, appropriate redox potential and non-reactivity in both oxidised and reduced forms with any of the components of a Li-ion cell, still more research is needed to ﬁnd suitable redox shuttle species .
2) Increasing the temperature at which the battery thermal runaway occurs to 300+C can delay the occurrence of large-scale internal short circuit inside the battery, and increase the SOC of thermal runaway to 180%. This measure could potentially be real- ised by optimising the battery pressure relief design or using a diaphragm with a higher heat exchange stability [80,260] to post- pone the rupture of the battery .
More information: Sandeep S. Sebastian et al, Adaptive fast charging methodology for commercial Li‐ion batteries based on the internal resistance spectrum, Energy Storage (2020). DOI: 10.1002/est2.141