Felix
Schomburg
a,
Bastian
Heidrich
b,
Sarah
Wennemar
c,
Robin
Drees
def,
Thomas
Roth
g,
Michael
Kurrat
de,
Heiner
Heimes
c,
Andreas
Jossen
g,
Martin
Winter
bh,
Jun Young
Cheong
*ai and
Fridolin
Röder
*a
aBavarian Center for Battery Technology (BayBatt), Weiherstr. 26, 95448 Bayreuth, Germany. E-mail: jun.cheong@uni-bayreuth.de; fridolin.roeder@uni-bayreuth.de
bMEET Battery Research Center, Institute of Physical Chemistry, University of Münster, Corrensstr. 46, 48149 Münster, Germany
cProduction Engineering of E-Mobility Components RWTH Aachen University Bohr 12, 52072 Aachen, Germany
delenia Institute for High Voltage Technology and Power Systems, Technische Universität Braunschweig, Schleinitzstraße 23, 38106 Braunschweig, Germany
eBattery LabFactory Braunschweig, Technische Universität Braunschweig, Langer Kamp 8, 38106 Braunschweig, Germany
fFraunhofer Research Institution for Battery Cell Production FFB, Bergiusstraße 8, 48165 Münster, Germany
gTechnical University of Munich, School of Engineering and Design, Department of Energy and Process Engineering, Chair for Electrical Energy Storage Technology, Arcisstraße 21, 80333 Munich, Germany
hHelmholtz Institute Münster, IEK-12, Forschungszentrum Jülich GmbH, Corrensstr. 46, 48149 Münster, Germany
iDepartment of Chemistry, University of Bayreuth, Universitätsstraße 30, 95447 Bayreuth, Germany
First published on 13th February 2024
The battery cell formation is one of the most critical process steps in lithium-ion battery (LIB) cell production, because it affects the key battery performance metrics, e.g. rate capability, lifetime and safety, is time-consuming and contributes significantly to energy consumption during cell production and overall cell cost. As LIBs usually exceed the electrochemical sability window of the electrolyte, formation is required to activate and stabilise the electrochemical reactions. Enhanced battery technologies are poised to further expand voltage windows and harness conversion or metal electrodes to elevate energy density, thereby magnifying the significance of cell formation in the battery realm. Despite its critical importance, even the understanding of the formation process of conventional LIBs is still incomplete due to numerous influencing factors. Complex internal processes and the associated high experimental and simulation effort make it difficult to gain a thorough understanding of the process and hence to optimise it. This review paper provides a systematic overview of the formation process and its influencing factors. It is emphasized that material and cell design and the formation process are not independent, but must interlock with each other. Promising experimental and simulative methods to gain the required understanding of the interplay for a truly knowledge-based design of the formation process are highlighted. In the concluding discussion research gaps are identified and a perspective for development of tailored cell formation processes for current and future battery technologies is outlined.
Broader contextDriven by the transition towards renewable energy sources and electromobility, the demand for cost-effective and sustainable batteries is continuously growing. Simultaneously, the demand for higher energy densities entails safety risks and challenges in maintaining long lifetime. Addressing these challanges is crucial to overcome driving anxiety and concerns against novel technology. These issues are closely linked to the formation process, which is essential to establishing stable interphase layers and utilizing materials outside the electrolyte's stability window. Challenges intensify with new materials like high-voltage cathodes and lithium metal or silicon anodes. In addition, sodium-ion technologies, seen as promising alternatives, face hurdles in establishing stable interphase layers due to high solubility. In all these technologies, a well-designed formation process is pivotal, yet complicated due to many interdependent influencing factors. This review highlights the need for a knowledge-based process design to ensure safe and durable batteries. By reviewing the status, outlining theoretical and experimental methods, the necessary foundation is laid. |
Formation is the final active process step in LIB cell manufacturing. The process affects the quality of the freshly assembled cells and contributes significantly to the overall cost, accounting for up to 33% of the production cost.12,13 Formation typically involves multiple charge and discharge cycles. The formation cycling is required to activate the materials and establish protective interphase layers at the electrochemically active interfaces that enable a stable operation of the cell outside the electrolyte's electrochemical stability window. The drive to increase energy density will place higher demands on the formation process, as high-capacity and high-voltage materials utilize a wider voltage window and are subject to significant volume changes.
The formation process must be carried out carefully, as it is linked to influence key electrochemical and safety properties of LIB cells. However, overly cautious formation is time consuming and therefore at odds with efforts to accelerate battery cell production and reduce overall costs. There is a significant potential to improve the formation process to optimise cell quality and reduce overall costs.14 The industry uses highly sophisticated procedures, but details are guarded as they are considered essential intellectual property of the manufacturers. Despite its importance, the formation process is not well understood due to the limited number of comprehensive studies in academia.
This review comprehensively summarises, examines and analyses the various facets of the formation process. It describes the formation process, presents the current state of knowledge on the various influencing factors, and highlights the most relevant experimental and theoretical methods for future research. In the light of future battery technologies aimed at higher energy density, a summary and suggestions for the further development of the formation process are presented. This review contributes to the overall understanding of formation and provides suggestions for methods and open questions being relevant to enable a knowledge-based process design for current and future battery technologies.
The purpose of the formation process is to electrochemically activate the cell so that its subsequent performance is positively influenced. The formation process is critical for a number of reasons. Firstly, formation is the last process step in the production of a battery cell and any scrap that is produced during formation causes the loss of value of all previous process steps.13 Secondly, the formation process is very time consuming and energy intensive.15,16 Finally, the process can have a significant impact on cell performance metrics such as capacity, power capability, lifetime and safety.
The formation process is often a production bottleneck due to the relatively low currents applied to individual cells.17 Achieving high throughput requires significant investment in equipment and cell formation space. Manufacturers therefore aim to reduce the formation time.18
Formation time is highly dependent on cell quality requirements and cell-to-cell variation. For batteries with lower quality requirements, relatively short formation times are possible. Here the quality controls performed as part of the formation/conditioning phase are often less sophisticated.18 Therefore, the progress of formation and individual cell quality is often not known. In applications such as electric vehicles (EV), cell quality and cell-to-cell variation are very important because the lowest cell capacity limits the total usable battery capacity. Further cell-to-cell variation within a battery system promotes ageing and is a safety concern.19 Therefore, battery cell production for such applications requires much more sophisticated formation procedures, including quality management measures. The formation time is much longer and may include various intermediate cell tests.
The energy consumption of the formation process is significant.16 According to Kurland et al., Northvolt expects the cell production at Northvolt Ett to use 20% of the total production electricity consumption for the formation step.20 In general, the electricity consumption can vary significantly between laboratory scale and industrial plants. According to Erakca et al. this is because recuperation is commonly used in industrial production facilitities.21 Recuperation is the process of recovering energy released during battery discharge. Consumption also depends on the energy consumption of the electrical equipment, which is usually higher for laboratory equipment. Thus, the energy consumption of the formation process can vary from 0.6 Wh of Wh cell capacity produced in pilot plants to 42.6 Wh of Wh cell capacity produced in laboratory cell productions.21
The formation process is influenced by several factors. These include material and cell design, as well as the electrochemical conditions during formation. Some of these factors, such as electrode properties or moisture contamination, also depend on previous manufacturing steps. Therefore, the entire process chain and its interdependencies must be considered when optimising the formation process.22,23
During formation several processes take place that lead to significant changes in the cell properties. The main processes observed are shown in the centre of Fig. 1. These processes include solid electrolyte interphase (SEI) formation,28 cathode electrolyte interphase (CEI) formation,29 structural changes in the active material,30–32 copper (Cu) corrosion33–37 and aluminium (Al) dissolution.36,38–40
The interfacial layers formed on the negative and positive electrodes can be expressed in terms of low unoccupied molecular orbitals (LUMO) and high occupied molecular orbitals (HOMO), as shown in Fig. 2.41,42 Both electrodes can reach potentials that exceed the electrochemical stability window of the electrolyte triggering electrochemical side reactions.
Fig. 2 Schematic representation of the thermodynamic stability window of the electrolyte|electrode interface. Φ and μ represent the electrodes potential and chemical potential, respectively. The indices Neg and Pos denote the negative and positive electrodes, repectively. Stabilising passivation layers, i.e. SEI and CEI, are formed when ΦNeg < LUMO and ΦPos > HOMO, respectively. The passivation layers extend the operating stability range, allowing the use of low voltage materials on the negative and high voltage materials on positive electrode. Redrawn from Goodenough and Kim,41 as well as Horstman et al.42 |
Among the processes during the formation, SEI formation is the most widely studied. This process is initiated by an irreversible electrochemical reduction reaction that consumes electrical charge, lithium (Li), solvent and other electrolyte components at the negative electrode. The reaction products can precipitate on the electrode surface inhibiting further side reactions. The ideal passivation layer exhibits strong electrical insulation, high ion conductivity, mechanical and thermal stability.43–47 The most commonly reported SEI components are lithium fluoride (LiF),48–53 lithium carbonate (Li2CO3),48,53–58 lithium oxide (Li2O),48,53,55 litihium ethylene dicarbonate (LEDC)51,52,59 and lithium alkyl (bi)carbonates (ROCO2Li/RCO2Li).50,53,55,58
CEI formation is another side-reaction that is first initiated during formation. The process is initiated by an oxidation reaction that releases electrical charge to the positive electrode.60,61 Like SEI, CEI inhibits further degradation.62 The CEI is thinner than the SEI,55,63,64 suggesting a lower magnitude of charge transfer due to CEI formation compared to SEI formation. Typical CEI components are LiF,55,65–68 LixPOyFz,65,67–69 Li2CO365–67,69,70 and ROCO2Li.66–71 For both interface processes, reaction products can also be gaseous species.72–74
Despite the interfacial processes, structural changes have also been observed during the first cycles. Agglomeration of nanoparticles,75 particle cracking76–81 or exfoliation of particles82–84 can lead to significant changes of the active material. Structural changes have also been observed at the material level. Here, irreversible rearrangement of the material can cause significant heat release and voltage hysteresis during the first cycle.31,85–87
After the end of formation, the cell characteristics should stabilise. A commonly used indicator of this stabilisation is the coulombic efficiency (CE), which approaches 100% towards the end of formation.18 CE is therefore useful in defining a target criterion for the end of formation.
After formation, the cell properties have changed significantly. Firstly, the formation results in a loss of cyclable Li, which has been immobilised mainly by the side reactions forming SEI.88–90 Secondly, the loss of cyclable Li results in a reduced cell capacity91 and a relative shift in the electrode open circuit potentials.92 Thirdly, there is an increase in interfacial resistance.93–96 In addition, the grown passivation layer and structural changes are likely to lead to a decrease of porosity. Further, electrode restructuring leads to a decrease in conductivity as seen in ageing studies.97–99 Similarly, the electrolyte composition is changed after formation,100,101 which impacts the Li-ion diffusivity and ionic conductivity.99,102,103
All of these changes can have a significant impact on cell performance in terms of ageing,104–106 power capability,107–111 safety,112–118 energy consumption,12,13,17 cost12,13,17 and CO2 emissions12,13 as shown in Fig. 1.
Prior to the start of the formation process, initial wetting is carried out. This can be done by a high temperature soaking process, in which the cells are exposed to high temperatures, typically between 30–50 °C, for several hours to change the viscosity of the electrolyte and reduce the time between the electrolyte filling process and reaching the optimum wetting level to start the formation.121 Formation is initiated by the so-called pre-charging process,122 in which a low charging current is applied to the cell until a low voltage, e.g. 1.5 V, is reached. Low currents and voltage limits are necessary to protect the Cu-based current collector from corrosion.12 The voltage limit during pre-charging can also be increased to a state where most of the gas generation during the first cycle has been completed. This is followed by the degassing step. The gases are usually removed by applying a vacuum. The cell is then sealed.123 Generally, the housing is sealed as early as possible to minimise evaporation of volatile electrolyte components.119 If there is no pre-charging step, degassing is carried out after the formation is complete or before shipping to the customers. After degassing, a second wetting and possibly electrolyte filling process may be added to compensate for any electrolyte loss during pre-charging. This is followed by the formation cycling, which includes at least one and possibly several charge and discharge cycles with a defined current and voltage limit. Finally, the cell is aged. During ageing the cells are stored at a defined state of charge (SOC) and temperature. The ageing process completes cell formation. It also serves as a self-discharge test. In this context, the ageing process can also be considered as part of the end-of-line (EOL) cell quality test since an increased capacity loss during storage indicates unwanted self-discharge.
Without the ageing test, industrial formation procedures typically take less than 20 hours.4,119 The ageing test alone can take up to three weeks.12,17,119 For application with low quality requirements, formation procedures are significantly shorter.13,122,124 Once the formation procedure is complete, further EOL testing is performed.123
The EOL test protocol can vary, but can include electrochemical tests for discharge capacity, CE, internal resistance, self-discharge test if not already covered by ageing process, impedance, as well as leakage, weighing and optical tests. Depending on the material, size, type and format of the cell, the formation procedure varies in terms of sequence, duration, and quantity from one cell manufacturer to another. The main objective is to find procedures, cycling protocols and evaluation methods that reduce the overall process time, while ideally performing formation and cell quality testing simultaneously.
To address the high energy consumption and the move towards greener battery production, new formation equipment with a recuperation function has been developed to use the discharge energy from one cell to charge another. Controlling and precisely delivering power to the formation equipment for an entire Gigafactory is a significant challenge. Recently, Tesla announced an innovation to reduce the cost of the formation process by leveraging the production design to allow a group of cells, rather than individual cells, to undergo this formation process resulting in an 86% reduction in formation investment.125 In addition, Müller et al. proposed an optimised formation step for serially connected cells.126 Thus, serial formation techniques could potentially reduce processing costs. In principle, parallel connection of cells is also possible. However, this reduces the maximum current that can be applied and is therefore only useful for smaller cell formats.
While the potential of the negative electrode clearly exceeds the electrochemical stability window, the positive electrode exceeds the stability window only at high cell voltages.141 However, recent experimental work has shown that HOMO energies do not necessarily correspond to the oxidation stability of the organic electrolyte, and that the CEI layer is observed at potentials below those traditionally considered thermodynamically stable.142 Nevertheless, the CEI is receiving little attention today with respect to formation cycling. However, this should be reconsidered, in particular if high-voltage materials are investigated as discussed below.
As can be seen, the voltage level is an important factor influencing the formation process and the SEI structure. It has been shown that SEI formation takes place in different stages, while low potentials of the negative electrode seem to be favourable for the formation of a stable SEI for common electrodes and electrolytes. At the same time, charging rates must be limited to mitigate Li plating. These principles should be considered while developing formation strategies that can be muti-cycle or single-cycle strategies.
Conservative formation strategies use multiple full cycles at low C-rates and therefore can last several days.17,105 To include multiple cycles while accelerating the formation process, decreasing the difference between upper and lower cut-off voltage, i.e. sub-cycles, or increasing the C-rates is possible. Pathan et al. investigated different formation variants within ten 0.05C cycles between different cut-off voltages.147 A positive correlation was found between increased lower cut-off voltage and capacity retention (CR). Furthermore, CR was improved when the difference between the upper and lower cut-off voltages was the smallest. These results suggest that the main influencing factor is the cell voltage and formation at high voltages is beneficial. This is utilized also in other studies that aim to decrease formation time. An et al. investigated a fast formation strategy with four repeated sub-cycles between 3.9 V and 4.2 V at high C-rates, i.e. 0.2C, and showed improved surface layer resistances compared to two cycles at lower C-rates, i.e. 0.05C. With this strategy, formation time was reduced from 80 h to 21 h. These formation times were even further reduced in consecutive studies. Wood et al. were able to reduce the duration to approximately 14 h by increasing the C-Rate to 1C for the initial charging until 3.9 V and for the final discharge from 3.9 V to the lower cut-off voltage.17 Drees et al. compared the protocol of An et al.148 with another sub-cycling formation protocol.149 The new protocol used 0.8C and removed the constant voltage (CV) phases during the sub-cycles, resulting in a reduced formation time of only 10 h, without significant impact on the cell performance. Thus, multiple studies show that formation time can be reduced without significantly compromising the cell quality.
Weng et al. investigated the influence of these cycling strategies on lifetime in more detail by comparing the average lifetime of cells subjected to the fast formation strategy of An et al.148 and Wood et al.17 with a slower baseline formation strategy at different temperatures.104 The results indicate, that the fast formation strategy even improved cycling performance at both 45 °C and room temperature. However, swelling was observed at the end of life at 45 °C for the fast formation strategies.104
A comprehensive comparative study of five different formation protocols with different cycle numbers and currents was performed by Mao et al.105 The total formation time ranged from 10 h to 86 h. The main results of this study are shown in Fig. 5. Analysing the first cycle CE (Fig. 5A), rate capabilities (Fig. 5B), and cycle retention (Fig. 5C), the authors conclude that neither long, 86 h, nor short, 10 h, formation times provide the optimal electrochemical performance. Instead, intermediate formation times, i.e. 26 h and 30 h, resulted in the lowest electrode resistance after formation (Fig. 5D), lower capacity losses for 300 ageing cycles at a cell voltage of 3.8 V (Fig. 5E) and lower area specific resistance (ASI) of cells at different cycles (Fig. 5F). The morphology of material of three protocols was further investigated by scanning electron microscope (SEM) and X-ray photoelectron spectroscopy (XPS). The atomic compositional profiles of the four elements, i.e. surface Carbon (C), Oxygen (O), Li, and Phosphorus (P), were used to estimate the SEI thickness of the electrodes formed with the different formation protocols. Among the electrodes studied in detail, the medium formation time of 30 h showed the thinnest SEI layer thickness of 54 nm. Even though the effect of the cycling strategy is not obvious in some studies, the formation cycling can influence the SEI structure and properties. However, whether one structure or the other is beneficial is often not well understood, as evidenced by the results of Lee et al., who investigate the influence of SOC and number of formation cycles with a fluoroethylene carbonate (FEC) based electrolyte.150 Cells were charged to SOC 10% (3.5 V), 20% (3.7 V), 30% (3.9 V), and 100% (4.2 V) and then fully discharged (2.7 V). Cells formed with three cycles in a narrow SOC range at low voltages, i.e. 10% to 30%, showed improved CE after 180 cycles. These results are in contrast to many other studies, as they suggest that lower negative electrode potentials during formation result in a less stable SEI. This is probably due to FEC, which was not used in the other studies. Based on the morphological analysis, this effect was attributed to a higher proportion of interlinked particles on the particle surface, which is illustrated in Fig. 6. Based on these results, it is emphasized that the cycling strategy is not independent of the material and cell design but must be adjusted individually.
Fig. 5 (A) Comparison of first cycle CE and total efficiency during formation with respect to different formation protocols. Comparison of the (B) rate capablity and (C) cycle retention characteristics with respect to different formation times. Nyquist plots of cells with different formation times (D) after formation cycle and (E) after 300 ageing cycles at 3.8 V cell voltage. (F) Area specific resistance of cells with formation time of 30 h and 86 h as a function of the number of cycles. Reprinted with permission from Mao et al.,105 Elsevier, Copyright 2018. |
Fig. 6 Illustration of the effect of performaming formation cycels within a narrow votlage range at low SOC with a FEC containing electrolyte on the SEI on graphite negative electrode. Reproduced with permission from Lee,150 Elsevier, Copyright 2014. |
Therefore, also other active materials must be investigated. Rago et al.33 investigated the fast formation protocol of An et al.148 for cells with Si–Gr composite electrodes. For this material, no performance differences compared to five-cycle formation strategy with C-rate of 0.05C were found. However, the authors also reported that the negative electrode surfaces for different formation protocols had different optical appearances and different chemical SEI compositions. Faster multi-cycle-based formation resulted in suppressed Cu and Al dissolution and thinner SEI.33
The effect of the upper cut-off voltage on the CEI in a multi-cycle formation was investigated by Phraewphiphat et al. for nitrogen-doped Li2MnSiO4/C positive electrodes.151 Five formation cycles were used with 4.5 V and 4.6 V as the respective upper voltage limit during the formation cycling. The cells formed at the lower potential showed improved cycling stability. The improved ageing behaviour was attributed to the improved structural stability of the CEI. Zhang et al. observed that during normal cycling the CEI does not form as a conformal surface layer completely covering the interface.152 However, by brief electrical shorting of their NMC||Li cells, they were able to form a conformal surface layer.
For multi-cycle formation strategies, cell voltage was confirmed to be an important factor influencing for both CEI and SEI formation. In contrast, the influence of cycle number and applied current is less obvious. Most studies found no significant influence on the cell properties. However, some studies report an influence on interphase properties such as thickness, composition, and resistance. The influence of the choice of material and electrolyte makes it difficult to draw general conclusions und rather suggests that cycling strategy must be adjusted to material and cell design.
Drees et al. compared several multi-cycle formation strategies with fast single-cycle formation strategies.149Fig. 7 shows (A) a slow multi-cycle, (B) a multi-sub-cycle and (C) a fast single-cycle formation strategy that have been investigated. The slow multi-cycle formation consists of two full cycles with low currents, i.e. 0.05C, in the first cycle and moderate currents, i.e. 0.2C, in the second cycle. This conventional approach is very time consuming. The multi-sub-cycle formation approach involves multiple sub-cycles at high SOC to enable faster SEI growth while reducing the formation time compared to full cycles. Fast single-cycle formation uses constant current (CC) charging at 1.5C followed by a constant negative electrode potential phase. Using cells in a three-electrode setup,134 the potential of the negative Gr electrode was limited to 0.02 V vs. Li|Li+. The potential limit was chosen slightly above 0 V vs. Li|Li+ to accelerate SEI growth while effectively preventing Li plating. Charging was stopped at 80% SOC. The cells were then discharged at 1C. The single-cycle formation strategy was clearly the fastest formation as it took only 1.7 h. Nevertheless, the different formation strategies achieved similar discharge capacities, internal resistances, and electrochemical impedance spectra immediately after formation and during 250 cycles. Differences were observed in a high temperature storage test at 60 °C, 80% SOC and 28 days storage time. Variants with the highest initial charge rate resulted in the lowest cell degradation, while the fast single cycle formation was one of them.149 A similar single-cycle fast charge formation approach was upscaled from three-electrode coin cells to 5 Ah pouch cells using a look-up-table based current control.153 The maximum C-rate was 3C and the total formation time was below 1 h. These fast formed cells resulted in similar cycling performance over 1000 cycles to various slower reference formations. The fast formation approach using the look-up-table based method was also successfully applied to the formation of 25 Ah prismatic cells.27 These results indicate that single cycling formation is promising approach to reduce formation time and possibly also improve cell quality.
Fig. 7 Formation strategies with different speeds and number of cycles. (A) Conventional slow multi-cycle formation. Very slow first full cycle and moderate second full cycle. (B) Multi-sub-cycle formation with periodic sub-cycles at high SOC (low NE potentials).33,147–149 (C) Fast single-cycle formation with high currents at low NE potentials.27,135,149,153 Reprinted with permission from Drees et al.,149 John Wiley and Sons, Copyright 2022. |
The influence of the current within the first charging cycle seems to be the key influencing factor and was investigated further by Chiang et al.154 They investigated the influence of the current by comparing three single-cycle formation protocols with LCO||Gr-based cells.154 They also proposed a dual current procedure which limited the rate to 0.1C for 1 h and then increased the rate to 1.0 C for the remaining first cycle. Long-term cycling experiments showed that the rate of capacity loss was not affected by either protocol. In addition, Zhang et al. studied the influence of the current rate.155 They used the second cycle CE to evaluate the formation protocols of Gr||Li cells at different current densities and observed that as the current density increased, the first cycle CE increased while the second cycle CE decreased. According to these findings, a high current density contradicts the idea of a single-cycle formation, as more cycles are required to complete the formation process. They further investigated the cycling performance and found that it was slightly better for cells after formation at low or medium current densities. Moretti et al. observed a compositional difference when comparing the effect of three formation protocols.156 They studied a single current formation protocol, which includes a single 0.05C constant current followed by constant voltage (CC–CV) cycle between 3.9 V and 2.5 V, a dual current formation protocol, which includes a slow 3 h, 0.05C pre-charge step, a subsequent 3.6 V CV charge phase and a 0.1C discharge to 2.5 V, and an advanced industrial protocol, which includes multiple steps, i.e. a pre-charge to 3.0 V, a long rest and a 0.1C CC–CV cycle between 3.6 V and 2.5 V at 40 °C as well as a final 1 h rest at 20 °C. Comparing the effect of formation protocols on SEI in LFP||Gr coin cells, the authors observed the highest LiF content and cycling performance in cells formed with the dual current protocol. Antonopoulos et al. reported improved Li-ion transport rates for formation protocols with high initial charge rates. They applied a current rate of 1C until the negative electrode potential reached 0.24 V vs. Li|Li+, followed by a 0.1C charge.138 The authors also found improved C-rate capability for currents higher than 1C when applying 1C during formation until 0.24 V vs. Li|Li+.
Besides the formation of passivating layers, structural changes are also possible. Although most of the research on the formation process focuses on the interfacial layers and the decomposition of the electrolytes, the degradation effects of the active material particles have also been investigated. Cheong et al. adopted different current densities in the initial formation cycle and observed how the morphological evolution takes place for Co3O4 nanoparticles as shown in Fig. 8.75 It was found that a fast formation cycle, i.e. with a current density of 1000 mA g−1, resulted in initial agglomeration of nanoparticles, which ultimately led to improved structural integrity of the active material particle. In contrast, the formation cycle with a current density of 50 mA g−1 resulted in increased agglomeration after cycling.
Fig. 8 Structural evolution of Co3O4 nanoparticles after a single formation cycle with different current densities and subsequent cycing. Reprinted with permission from Cheong et al.,74 Elsevier, Copyright 2019. |
These studies indicate that the composition and properties of the SEI as well as of active material structures can be influenced by single-cycle formations strategies. However, it remains unclear whether faster current rates are beneficial or not.
As can be seen, it is not possible to draw general conclusions and make direct comparisons between different studies. This may be due to the influence of material and cell design. This can also be seen from the different current limits reported to mitigate Li plating.157 Drees et al. reported Li plating after 1.5C CC–CV charging, which correlated with reduced discharge capacity.149 Mao et al. reported Li plating at much lower rates. They used a 0.5C CC–CV formation and found increased impedances and decreased cycling stablity.105 In contrast, Münster et al. also found Li plating only at higher rates, i.e. a 2C CC–CV formation. However, this did not affect cycling stability compared to a slow 0.2C formation protocol that did not resulted in Li plating.132
Single-cycle strategies promise to significantly reduce process time compared to multi-cycle strategies. However, avoidance of Li plating and incomplete formation are important limiting factors. There is also evidence that charge rates affect cell properties such as charge rate capability and self-discharge, but the mechanisms behind this are not well understood. Formation strategies must always be tailored to the material and cell design. For single formation strategies, the design of the first charge allows to influence the SEI composition and properties, while dual current formation is a possible approach.
The quantity of gas generated is linked to the active material on both electrodes.160,161 Similar gas species are reported for common negative electrodes, i.e. Gr and silicon (Si). For both active materials the release of ethylene (C2H4), carbon monoxide (CO) and carbon dioxide (CO2) has been reported for pure ethylene carbonate (EC) or EC-based electrolyte solutions.58,161–166 Additionally, ethane (C2H6), methane (CH4) and hydrogen (H2) may be formed as by-products of EC on Gr negative electrodes. These gases are related to the specific SEI formation mechanism on Gr and Si-based negative electrodes. The gas species may differ for other negative electrodes.167 On the positive electrode side, mainly CO and CO2 are reported as by-products of oxidation reactions.72,73,168 According to density functional theory (DFT) simulations by Leung et al., EC decomposition on Li0.6Mn2O4 leads to CO2 production only at high voltages.169 In contrast, Jung et al. hypothesise that the CO and CO2 evolution is due to a chemical reaction of the lattice oxygen with the electrolyte by analysing the gas evolution of different NMC||Gr cells for different cut-off voltages.73 In addition, the CO2 evolution may result from material impurities,72e.g. due to the oxidation of Li2CO3 at high potentials of the positive electrode.170,171
Electrolyte composition has a significant effect on the total gas volume and relative species content. For example, it has been found that EC-based electrolytes generate a higher volume of gas compared to propylene carbonate (PC) based electrolytes.172 Increased amounts of CO2 and CH4 were observed for solutions containing EC and dimethyl carbonate (DMC), respectively.172 The choice and ratio of solvant as well as of additive on the gassing behaviour was also reported in other experimental studies.164,166,173–177
Bernhard et al. found that the C2H4 gas production starts at around 0.9 V vs. Li|Li+, with the rate of production peaking at around 0.5 V vs. Li|Li+ before decreasing until the end of the first charge cycle. Significant amounts of C2H4 were found only in the first cycle.178 These results were confirmed by Jung et al. at room temperature.179 However, at 40 °C or 50 °C the C2H4 gassing extended into the subsequent cycles. In addition, the total amount of gas increased at elevated temperatures. Jung et al. also found significant amounts of CO and CO2 gas within the first formation cycle, but the CO2 was attributed to the NMC positive electrode.179 Xiong et al. reported increased gas volume for NMC electrodes when stored at elevated temperatures.74,164,180
It was found that some gases are consumed by SEI reduction reactions,165,178,179,181 and, to some extent, cross over from the positive to the negative electrode.182 Ellis et al. investigated the influence of gas consumption by comparing cells with and without degassing.165 After an initial 0.05C pre-charge to 3.5 V, the gas was removed. However, the degassed cells resulted in the same performance as the non-degassed cells.165 In contrast, Bläubaum et al. reported improved cell performance when CO2 gas was added to the electrolyte prior to formation.183 A positive effect of CO2 on the SEI was also reported in other studies.173,174 Thorton et al. hypothesis that stabilising effect of CO2 origins from a reduction reaction, which forms stable solid species and C2H4.181 CO2 injection to actively modify the resulting SEI composition was also suggested by Spotte-Smith et al. in a theoretical study.162
The produced gas amount also depends on the formation current and voltage.182,184–186 Leißing et al. investigated the influence of different C-rates on gassing.184 Gassing was found to be caused by both SEI formation and unspecific decomposition. Both very slow and very fast formation led to an increase in non-specific gassing, but the species observed differed.185
The gas amount was also seen to limit the ability for fast-charging formation. When Drees et al. scaled a single-cycle fast charging strategy (Fig. 7C) from three-electrode coin cells to pouch cells, irregular plating was detected.135 By introducing a moderate pre-charge and early degassing step before continuining with the same fast charging protocol, Li plating was successfully prevented. The efficieny of the gas removal was improved by applying external compression on the pouch cells, which also resulted in lower overpotentials.
Significant gas evolution occurs during battery cell formation. The amount and composition of the gas is influenced by the formation protocol, the active material, the electrolyte composition, and cell design. Gas formation can be a limiting factor for fast formation strategies, especially with thick electrodes and larger cell formats. Possible countermeasures are adequate degassing steps, external compression and reduction of gas evolution by selection of suitable electrolytes. Gas injection may also be used as an active measure to influence reactions during the formation process.
Fig. 9 Cell voltage curves of the first formation cycle (CC-CV-CC) for different ambient temperatures and external pressures of (A) 1.7 kN and (B) 0.05 kN. Reprinted with permission from Heimes et al.,188 John Wiley and Sons, Copyright 2019. |
Currently, pressure is rarely studied in the context of formation. However, for novel materials that undergo significant structural changes and volume expansion, pressure requires more attention.
An et al. describe that elevated formation temperatures accelerate SEI formation, resulting in a reduction in formation time.45 SEI formation at 40 °C results in a layer consisting of compact, inorganic components instead of a less compact, organic structure. Bhattacharya et al. performed formation at 25 °C and 60 °C.189 They compared the morphology and composition of the SEI layers and concluded that formation at higher temperatures resulted in a more homogeneous layer due to higher diffusion rates. He et al. elaborate that formation at 50 °C resulted in lower total internal resistances during and after formation in comparison to a formation at 25 °C.190 At 50 °C, the formation time was reduced to 15 h, compared to the formation time at 25 °C of more than 16.5 h.
Heimes et al. discovered similar relationships by investigating the influence of temperature on formation time.188Fig. 9 shows results for formation at different temperatures and pressure. Fig. 9A and B show that as temperature increases, the overpotentials decrease and the formation time decreases. The authors describe that higher ambient temperatures increase the effective conductivity of the electrolyte in the separator, improve the solid diffusivity in the active material and reduce the charge transfer resistance, which increases the reaction rate and reduces the internal resistance. Therefore, the formation time can be reduced by increasing the temperature.
In contrast to studies that suggest an advantage of elevated temperature, Lee and Pyun describe that as the formation temperature increases, so do capacity losses.191 They conclude that at elevated temperatures the SEI is formed with more defects, resulting in more transport pathways for solvent co-intercalation. Furthermore, the formation of Li2CO3 is accompanied by gas evolution which can damage the Gr layer. This is in contrast to an extensive follow-up study, the microstructure of SEI was investigated when formed at different temperatures.189 This study showed that when formed at 60 °C, the SEI morphology was more homogeneous and the surface composition differed significantly. It was found that formation or pre-treatment at higher temperatures could also be beneficial.
Ellis et al. found no significant influence of the formation protocol on either SEI thickness or the long-term cycling behaviour when they varied the temperature and CV step time during formation cycling.192 Moretti also analysed the SEI and found no major compositional differences when comparing two formation protocols where the first charge was performed only at 20 °C and partially at 40 °C, respectively.156
Elevated temperatures have been shown to be useful in reducing formation time by increasing the side reaction rate. However, there are conflicting results regarding possible improvements in cell quality. Both negative effects due capacity loss and positive effects due to improved interfacial homogeneity have been observed.
Many high-voltage or high-capacity materials, such as Nickel, Li or Mn rich layered oxides, high-voltage spinel oxides, and high voltage polyanionic compounds offer potential for advancement for the positive electrode.31,193 However, this can lead to thick interphase layers as shown on Fig. 10A. Achieving stability under highly oxidizing conditions is critical to suppress unwanted side reactions. The goal is to construct interphases that provide chemical and mechanical stability to promote long-term battery performance and safety. Potential strategies include new electrolytes that are more stable against oxidation or form effective passivating surface layers.194–196 In particular, the development of additives represents a promising way to improve layer properties without compromising the gains in the negative electrode. In addition, significant heat generation during formation cycling was observed for example in Mn-rich layered oxides.31 This unusual heat release was attributed to structural rearrangement of the active material and was also accompanied by pronounced voltage hysteresis. This indicates that focusing on the negative electrode might not be suitable when using high-voltage positive electrode materials that also require passivating interphases and undergo significant structural changes. This significantly complicates the optimization process as both electrodes must be addressed simultaneously in a full cell setup.
Fig. 10 Next generation pose significant challenges on the formation process which can be seen in (A) Surface layers in high voltage Ni0.7Co0.15Mn0.15O2 as depicted by TOF-SIM mapping. Reprinted with permission from Li et al.,193 Royal Society of Chemistry, Copyright 2017. (B) Illustration of different degradation mechanisms in silicon electrodes associated with significant volume changes. Reproduced with permission from Zhang,205 Springer Nature, Copyright 2017. (C) Snapshots of kinetic Monte Carlo simulation of SEI formation on a lithium metal electrode. Reproduced with permission from Wagner-Henke et al.,218 Springer Nature, Copyright 2023. (D) Illustration of the aging mechanisms in sodium-ion batteries. Reproduced with permission from Le Ma et al.,230 John Wiley and Sons, Copyright 2023. |
Silicon-based active materials, such as Si or silicon dioxide (SiOx), offer higher energy density potential, which can be up to ten times the theoretical capacity of Gr.197–199 However, large volume changes of up to 300% place significant demands and can lead to pulverization, lithiation retardation, and unstable SEI growth, as can be seen in Fig. 10B. The SEI on Si tends to be thick, non-uniform, oligomeric and susceptible to HF attacks.200 In contrast to Gr electrodes, the interphase structure of Si evolves significantly during cycling.201 However, the formation of a stable layer is critical to prevent thick SEI and continuous capacity loss. This challenge can be addressed by the development of new electrolytes202,203 or additives,197,200,204 material design strategies,205 the cycling strategy206 and the surface properties.207–209 In addition, Si electrodes undergo significant morphological changes, including crystalline-to-amorphous phase transitions, especially during initial charge–discharge cycles.210–212 Overall Si-based negative electrode materials will significantly increase the demand for sophisticated formation strategies.
Li metal is promising to achieve the highest energy density in Li-based batteries due to its low potential and high specific capacity.213–217 Unlike other electrode materials, Li metal immediately forms a SEI in a corrosion-like process within a few μs,218,219 as can be seen in simulation results shown in Fig. 10C. Therefore, the initial layer formation cannot be influenced by formation protocols. Nevertheless, a controlled subsequent formation cycle is often performed,4 which induces high volume expansions during plating and stripping of Li, associated with significant morphological changes of the interfacial layer.220 This places high mechanical stress on the SEI layer and poses a significant safety risk as Li can penetrate the separator and cause a short circuit. The development of a formation process that creates a robust SEI layer holds promise for enabling stable cycling and managing these morphological changes, as it can prevent uneven Li deposition.220 Therefore, the design of electrolytes195,221–225 or additives226–229 that promote the formation of more stable and homogeneous SEI layers appears to be a promising approach to enable Li metal for next generation batteries.
The challenges of formation are not unique to Li-based batteries, but also apply to sodium-ion batteries, for example. This battery technology promises to be a next-generation alternative due to the abundance of sodium. However, their formation process presents significant hurdles. Solid products in sodium-ion batteries exhibit higher solubility compared to Li-ion counterparts, resulting in dissolution that necessitates interface reformation, leading to significant capacity loss230 as illustrated in Fig. 10D. Therefore, the key objective is to develop a SEI that is predominantly insoluble. Studies also highlight the positive impact of formation cycling on film properties.231,232 For example, electrochemical reactions in high voltage regions may not effectively form the SEI, emphasizing the importance of specific voltage ranges.231 In addition, studies show that different electrolytes require different approaches: for example, NaBOB-based can benefit from lower C-rates and higher temperatures,231 while ionic liquids can benefit from higher C-rates.232
As outlined above, for many next generation materials the formation process can be considered pivotal and similar challenges arises. However, the possible influencing factors and solution can significantly differ. The development of all materials would significantly benefit from a better mechanistic understanding of the underlying processes, while often a complex interplay between electrochemical kinetics and mechanical stress needs to be considered.
Cui et al. used a surface treatment of Gr with 0.5 mol L−1 Li2SO4 aqueous solution, which resulted in the formation of thinner, more stable and elastic SEI layer capable of accommodating structural changes during cycling.233 The study suggests that surface properties play an important role in the formation process. Further studies have found a correlation between Gr surface properties and first cycle specific charge loss, confirming the importance of the surface properties of active material. A key factor in determining the initial capacity loss is the active surface area.90,234,235 The study of Ng et al. also highlights the importance of oxygen groups as an influencing factor when the active surface area becomes greater than 0.2 m2 g−1.234
Different types of Gr in EC- and PC-based electrolytes were investigated by Spahr et al.236 They show that active surface area was a main factor determining the surface passivation.236 Further insights into the importance of the active surface area of Gr was gained by Märkle et al. who investigated SEI formation on pristine and heat-treated Gr.237 Using SEM, the surface coverage and extent of exfoliation of Gr particles was investigated for different specific currents. Differences were reported between electrolytes and Gr type. The pristine Gr showed no exfoliation due to the larger number of oxygen-containing surface groups and hence high active surface area. The heat-treated Gr suffered from exfoliation regardless of the current in EC-based electrolytes. However, the exfoliation was less severe at higher current rates. This is attributed to suppressed solvent co-intercalation due to faster SEI coverage237 and is in agreement with other studies.137,236,238 In PC-based electrolytes, both Gr types suffered exfoliation regardless of current density, but again was more pronounced for the heat-treated Gr.236,237Post mortem SEM images of heat-treated and untreated Gr after a single charge are depicted in Fig. 11.
Fig. 11 Post mortem SEM images of heat-treated (A) and untreated (B) graphite (SFG6) in 1 M LiPF6 in EC/PC 1:1 (w:w) after a single charge to 0.005 V with a constant current of 0.01 Ag−1. A region suffering from exfoliation is indicated with a white circle. Reprinted with permission from Spahr et al.,236 Elsevir, Copyright 2010. |
Further exfoliation was found to be influenced by temperature and electrolyte composition by Zhang et al. in LFP||Gr cells.239 When formation occurred at 20 °C with certain solvent combinations and Gr types with few electrochemically active sites, a reduced first cycle efficiency was observed due to solvent co-intercalation. Increasing the temperature improved the desolvation kinetics and no exfoliation was observed. The addition of vinylene carbonate (VC) to the electrolyte also effectively mitigated solvent co-intercalation.239–241
In addition to the surface properties, the surface heterogeneity of the active materials influences the electrochemical properties during and after formation. Placke et al. reported that the absolute and relative extent of the basal/non-basal planes has a critical effect on SEI layer formation and leads to different initial capacity losses.242 The effect of current on Gr exfoliation was further investigated by Goers et al.238 Varying the current resulted in different local current densities, which played a significant role in the local overpotential that affected the electrochemical performance.238 Such partial local current density variations were found to be a key parameter in the exfoliation behaviour of Gr and emphasised the crucial role of exfoliation in the initial electrochemical performance of active materials.238 Tang et al. investigated the orientation of the Gr and the influence of the Li salt with respect to their effect on the quality of SEI after the formation.243 They found that the electrochemical activity is higher on the basal plane of the Gr, resulting in slower formation of the SEI.
While a strong dependence surface modifications of Gr negative electrodes on SEI formation has been reported in experimental236–238,244 and simulative studies,245 the reduction of EC on pristine Si has been simulated to be unaffected by surface modifications.246 However, differences between pristine and oxidised Si on the SEI formation were observed by Yun et al.,161 who modelled the reduction EC-based electrolyte on the two types of Si. The results showed that although an interaction between Li and passivated oxygen atoms on the Si surface leads to an increased presence of Li2O on the SiOx negative electrode, overall the two negative electrodes showed similar behavior.161
An important factor influencing the formation is the surface properties of the active material where the interfacial layers are formed. Exfoliation and slower surface layer formation are possible negative effects. Specific surface area, surface groups, e.g. oxygen groups and material homogeneity have been shown to be important factors. Some problems such as exfoliation or structural rearrangements are material specific and may vary betwenn vendors.239 Therefore, surface and material pretreatment must be considered individually for each material group.
With thicker electrodes also the risk of Li-plating rises during formation.135 In addition, Li-plating is also associated with an increase in the volume of gas generated during formation, which can further exacerbate the problem in larger cells.
The increased polarisation due to thick electrodes can be addressed by structuring the electrodes,258e.g. via mechanical embossing,259 laser ablation,260–262 or a gradient film design,263 enabling the production of thick high energy electrodes without sacrificing significant power capability.257 The maximum current and the degassing steps must be adjustd to the electrode design, especially for high-energy cells.
A basic requirement for successful formation is a sufficient electrolyte availability.24 This is achieved by maintaining an electrolyte volume factor that exceeds the total pore volume of the cell components. Recent research underscores the importance of achieving a 98% wetting degree before formation, as the quality of cells post-formation cycling was seen to be closely tied to the degree of wetting.27
There are several comprehensive reviews covering different aspects of battery electrolytes. Xu provided a review of electrolytes in LIBs and post-Li batteries and electrode|electrolyte interactions.141,265 The status and progress toward the realization of high-voltage electrolytes was reviewed by Fan and Weng.266 Electrolyte design principles for achieving high-energy Li batteries are summarized by Wan et al.7 Zhang et al. provide a summary and guidelines on the development of non-flammable electrolytes.5 Chemical and enviromental design guidelines for organic-based electrolyt are provided by Flamme et al.267 Additives were reviewed by Zhang,268 Zhao et al.269 and Haregewoin et al.270 Aravindan et al. review conductive salts used in Li-based batteries.271 Advances and future directions in the design of electrolytes for fast-charging is provided by Lei et al.272 and Zhang et al.273 The decomposition of electrolyte components has been the subject to many earlier reviews covering the impact on the SEI43,46,88,145,274 and CEI.275–277 Heisaken et al.89 and An et al.45 provide details of the initial SEI formation.
This summary focuses on electrolyte aspects that influence the formation process.
It is also possible to combine multiple conductive salts. For example, Yan et al. dissolved a blend of different salts.281 The electrolyte with the blended salts resulted in an increased cycling stability and lower impedance, both of which are attributed to the formation of uniform and stable passivation layer.
Varying the salt concentration affects the SEI/CEI layer composition. By elevating the concentration of F-containing salts the content of LiF can be increased at both electrodes, leading to a improved impedance and cycling stability, even at high voltages.282–285 Borodin et al. observed that in the case of highly concentrated electrolytes, solvent molecules were effectively excluded from direct interactions with the positive electrode surface.286
As can be seen, adjusting the salt and its concentration can be an effective measure to enable the formation of stable interphases. However, it must be taken into account that the salt and its concentration also significantly influence relevant properties such as the conductivity of the electrolyte.
EC is the most widely used solvent. It is known to form mainly oligomeric products, which are unfavourable due to their poor mechanical and thermodynamic stability.290,291 Accordingly, solvent decomposition should be suppressed to prevent the formation of an ineffective SEI/CEI.7 Differences in SEI morphology for typical solvent combinations on LTO electrodes are shown in Fig. 12. As can be seen, the use of PC/EMC (Fig. 12A) resulted in a smoother but thicker SEI compared to EC/EMC (Fig. 12B).172
Fig. 12 SEM images of LTO electrodes after five formation cycels between 1.5 V and 2.8 V (CC-CV, with a CC rate of 0.1C and CV termination criterion of 1/20C) of the two solvent mixtures PC/EMC 1:1 and (B) EC/EMC 1:1. Reprinted with permission from Liu et al.,172 Elsevier, Copyright 2015. |
The decomposition reactions also depend on the components in the Li+ primary solvation shell.7,286 As the predominant species in solution changes as a function of the salt concentration,282,292 this also elucidates the influence of the conductive salt concentration on the resultant passivation layer compositions.7 Balbuena et al. found that the spatial distribution of the reduction reaction product is influenced by the local solvent concentration.293 Their ab initio molecular dynamics (AMID) simulations suggest that Li2CO3 is more likely to be found in regions of low EC concentration, whereas LEDC is more likely to be found in regions of high EC concentration.
Since the solvent is the main educt for the decomposition reactions during formation, the choice of solvent has a significant impact on the composition and morphology. In addition, decomposition is strongly associated with salt and also indirectly affects interphase stability by determining the solubility of the solid reaction products, as discussed below.
During cycling, the solubility of the SEI species leads to dissolution and redeposition.127,296,299,300 Due to the higher solubility of organic species, dissolution is more pronounced for solvent-derived SEIs, which may also result in accelerated aging.296,301–303 By choosing a suitable solvent, the solubility of the SEI species in the electrolyte can also be used to selectively dissolve unfavourable species and form an SEI with a higher content of inorganic species.304
Solubilities may also affect initial formation, as suggested by Ushirogata et al.305 Analysing the solubility of SEI species often reported in the near-particle region, the authors question the often assumed immediate deposition of SEI species on the particle surface, which they call surface growth mechanism (Fig. 13A). Instead, a near-shore aggregation mechanism is proposed, which precedes in two successive steps (Fig. 13B). Upon reaction, the SEI species first diffuse into the electrolyte and agglomerate in the bulk phase. Agglomerates coalesce before being deposited on the electrode. This theory is supported by another recent simulation study, which indicates that the passivation layer grows by the aggregation of precursors within the electrolyte bulk.306 The strong aggregation of dissolved LEDC307 and the varying solubility of SEI species297,298 confirm the importance of the solvation state308 and support the hypothesis of a near-shore aggregation mechanism.
Fig. 13 Schematic representation of the (A) surface growth mechanisms and (B) near-shore aggregation mechanisms. Reprinted with permission from Ushirogata et al.,305 IOP Publishing, Copyright 2015. |
The discussion indicates that solubility is an important aspect of the electrolyte to consider, as it can affect the stability of the interfacial layer as well as the formation process.
Additive | Effect | Ref. |
---|---|---|
Vinylene carbonate (VC) | Compact but rigid sei | 118 |
Improved thermal stability | ||
Polycarbonate-rich sei | 314 | |
Fluoroethylene carbonate (FEC) | Smooth, compact sei structure | 118 |
Mechanically strong (less factures) | 313 | |
Improved thermal stability poly(carbonate)- and lif-rich inner sei | 318 | |
Aminosilane-based (TMS-ON) | Reduction of transition-metal dissolution due to stable cei | 310 |
Enhanced sei integrity with low interfacial resistance | ||
HF scavenger | ||
Lithium difluorophosphate (lidfp) | Uniform, mechanically stable sei and cei | 324–331 |
High li2co3 and lif contents | ||
Mitigation of near-surface reconstructions | ||
Transition metal scavenger | ||
High ionic conductivity | ||
5-Methyl-4-((trifluoromethoxy)methyl)-1,3-dioxol-2-one (DMVC-OCF3) | Mechanically deformable SEI | 334 |
Lif-rich inner SEI | ||
5-Methyl-4-((trimethylsilyloxy)methyl)-1,3-dioxol-2-one (DMVC-OTMS) | Mechanically deformable SEI | 334 |
Improved structural integrity of the CEI and SEI | ||
HF scavenger | ||
Lithium fluoromalonato(difluoro)borate (lifmdfb) + fluoroethylene carbonate (FEC) | Uniform, thin, compact sei and cei | 332 |
Improved positive electrode integrity and reduction of transition-metal dissolution | ||
High mechanical stress tolerance of the sei | ||
Hexafluorocyclotriphosphazene (HFPN) + fluoroethylene carbonate (FEC) | Lowered gas evolution | 333 |
Less alkyl carbonates in the electrolyte after aging | ||
Smooth, homogeneous sei with less organic species | ||
Improved structural integrity of the negative electrode |
Widely used are VC241,291,313–317 and FEC,226,316–322 which have been shown to improve SEI properties. VC can also effectively surpress co-solvant intercalation, improving first cycle CEs.239–241 Under certain conditions, the reduction of FEC and VC results in identical products, which may explain the beneficial properties of the resulting SEI layers.323 VC can also effectively surpress co-solvant intercalation, improving first cycle CEs.240,241
Another commonly studied additive is lithium difluorophosphate (LiDFP). In Gr||NMC full cells, LiDFP has been observed to improve rate capability and aging, which is attributed to improved stability and ionic conductivity of the SEI due to higher Li2CO3 and LiF content.324,325 The addition of LiDFP in NMC||Li half cells was also found to improve the stability of the positive electrodes and the CEI. The formed CEIs are reported to be more uniform, reducing mechanical stress in the layer and its transfer resistance.326–328 In addition, it mitigates near-surface reconstructions of the active material and acts as a transition metal scavenger, enabling stable operation of the NMC over a wider voltage range.327–331
Recently, electrolytes with multiple additives have been tested to take advantage of synergistic effects.332–335 For example, by combining LiDFP with other additives, the improvements in terms of rate capability, CR and calendric ageing of the individual additives were further enhanced.326,335–337 Combinations of addtives were systemetically assessed by Song et al.338 and Wang et al.312,339,340 with regards to electrochemical performance metrics and gas evolution.
Computational first-principles studies can provide insight into the functionality of additives. Wang et al. show that VC has a higher reduction potential than EC and produces more stable intermediates during reduction than EC.341 The effect on the oligomerization process of EC was also studied, showing that the introduction of FEC or VC modifies the process.342,343 This modification results in an increase in the formation of polymeric species and a decrease in the formation of unstable oligomeric products.342 In addition, in the presence of LiDFP, the LUMO and HOMO levels of EC and EMC are higher, indicating facilitated oxidation and hindered reduction of the solvents at the negative and positive electrodes, respectively.328
To conclude, additives can be used to influence the reactions during formation and to achieve interfacial layers with improved properties. As can be seen in Table 1, there is a wide variety of additives. Some are associated with specific properties. However, there is little understanding of the underlying mechanisms, which hinders to provide design principles. Nevertheless, in contrast to solvents and salts, additives promise a targeted improvement of the interphase properties without affecting the electrolyte properties.
Fig. 14 (A) Key scientific challengens reagarding the design of electrolyte involved in improving the SEI and CEI to enbale high-energy LIBs. Factors that need to be considered in designing an electrode include electrolyte compositio, bulk properties and the solvation structure. (B) Reduction potentials of various electrolyte composnents calculated via DFT simulations by Chen et al.222 The colors denote a preferential reduction of an anion (red) and solvant (blue). The marker denote species undergoing ionic aggregation (circels) and completely dissociated Li+ species. Reprinted with permission from Wan et al.,7 Springer Nature, Copyright 2023. |
The development of electrolytes is further complicated by the complexity of SEI/CEI, the ideal composition of which is unclear. Overall, a passivation layer with a high inorganic and low alkyl carbonate content seems to improve the aging performance of conventional Gr-based electrodes.333 The advantages of inorganic over organic SEI species are mainly due to their low solubility and thermodynamic stability, both of which ensure protection of the electrodes, resulting in improved safety and aging.
Among the inorganic components, high LiF content is increasingly associated with favourable SEI/CEI properties, especially for high-capacity materials.344 LiF is characterized by high mechanical stability.7,344 The mechanical stability results from low Young's modulus and high interfacial energy with alloy negative electrodes,7,222,334 which tolerates high active material expansions. This prevents breakage of the passivation layer, which would lead to increased electrolyte consumption. The mechanical stability and high lithiophilicity also prevent dendrite growth.344 Further, LiF exhibits good ionic transport properties229,345 and electronic insulation capabilities,346,347 resulting in a thin layer. This results in high first cycle CEs and low interfacial resistances. Lastly, LiF posses a high thermodynamical stability due to a wide band gap.348 These intrinsic properties result in an improved ageing, fast-charge performance and safety.5,7
To increase the content of inorganics in the SEI/CEI various electrolyte compostitions were tested. Multiple recent studies point out that inorganic-rich SEI/CEIs can be acheived by preferentially decomposing anions instead of the solvent. To foster the decomposition of anions over solvents, an increased number of anions in the primary solvation shell is key.7 Elevating the reduction potential of the anion and lowering the solvent's reduction potential (Fig. 14B) by tuning the electrolyte composition can augment the formation further to yield more inorganic species.222 By applying these design principles, tailored electrolyte solutions that promote F-rich SEI/CEIs already have developed and shown to improve the electrochemical performance.222,223,349
Despite the progress, improvements are still needed to achieve the high CEs (>99.9%) required for long-term use of high-capacity negative electrodes and high-voltage positive electrodes in many applications.350 Therefore, the search for electrolyte components and formation protocols that further promote the formation of the F-rich inorganic passivation layer is an important task.7,47 This requires a more in-depth understanding of the influence of individual electrolyte components on SEI/CEI composition, which is currently incomplete. This is due to the complexity of the interactions, the many contributing parameters. As a result, optimizing the electrolyte design remains one of the major challenges for improving the formation process. To address this challenge, more systematic research is needed to gain a more fundamental understanding of how to tune the electrolyte to achieve improved SEI/CEI compositions.
Man et al. describe the relationship between the degree of wetting of the separator and the effective current density in the battery cell.352 Inadequate wetting impedes ion transport and is resulting in higher internal resistances and inhomogeneous current distributions in the cell. During formation, inhomogeneous current densities can lead to inhomogeneous SEI formation, resulting in varying SEI and CEI thicknesses and structures.352
Ladadec et al. investigated the effect of the separator material on the local compressive stresses.353 They found that a given compressive stress has a greater negative effect, e.g. pore closure limiting Li transport, on polyethylene separator than on others separators such as a polypropylene separator. Therefore, some separator materials may also be more sensitive to internal and external compressive stresses during formation. This is relevant when external compression is applied during formation.
Frankenberger et al. investigated the effect of lamination on the formation process and electrochemical performance. They found that laminated cells exhibited reduced capacity fade at all formation rates and cycling rates.354 The authors attribute the reduced capacity fade of the laminated cells to either inhomogeneous SEI growth or significantly reduced additional surface reactions at fast formation cycles.354
The separator mainly influences cell performance and maximum applicable rates, which is also relevant to the formation process, e.g. fast formation cycles. Furthermore, only an indirect correlation has been reported due to incomplete wetting. It can be speculated that wettability might also correlate with degassing properties. However, currently there are no dedicated studies on this aspect.
In the work of Lee et al., half cells, coin cells and cylindrical cells were formed at different cut-off voltages.139 While the half cells and coin cells did not show any increased capacity fade during cycling at any specific voltage cut-off voltage, the cylindrical cells at a cut-off voltage of 3.6 V showed a reduced CR.139 Bridgewater et al. show a slightly increased capacity loss in the first cycle, but an improved CR for multi-layer pouch cells compared to coin and single layer cells.356 Both results suggest that cell performance is affected by formation in different cell formats.
Drees et al. investigated a fast formation protocol with three-electrode coin cells and multilayer pouch cells.153 The cells showed a slight voltage deviation for the different cell formats during formation, but showed similar voltage curves in later cycles at 1C.153 These voltage deviations were attributed to temperature and manufacturing inhomogeneities.153
In addition, the cell format influences the liquid and gas transport pathways, which also affect the formation process, as shown in a study by Hagemeister et al.27 They observed differences in the capacity decay after formation between pouch and hardcase cells for different wetting degrees. While there was a clear correlation between the ageing behaviour and the degree of wetting for the hardcase cells, there was no correlation for the pouch cells.27
Heidrich et al. investigated the SEI at the positive and negative electrodes of coin cells, laboratory-scale pouch cells and larger multi-layer pouch cells and found variations in the composition and thickness of the passivation layers.63 These were associated with differences in the residual water content, mainly of the casing used for each cell format. Higher initial water content in the cells resulted in lower CE during formation.63
In smaller cells, gas generation is uncritical, and formation can be performed in a sealed cell.119 In larger cells, gas bubbles interfere with the formation process by blocking Li-ion transport and interfering with electrolyte contact with the electrode surfaces,135,357,358 necessitating gas venting.119,120 In industry various gas venting methods are applied.359–362 The procedure may differ depending on the cell format. Hard case formats are usually sealed after gas venting.363,364 Pouch cell formats and occasionally hard case formats can be sealed, then opened and sealed again.12,119,365,366
In addition to the effect of cell format, there are also geometric differences between the cell types used in formation studies. In most commercial cells, the negative electrode is larger than the positive electrode, known as negative electrode overhang. This causes additional equalisation processes and interferes capacity,367 CE,367,368 and self-discharge tests.369,370 However, the effect of size differences is often neglected in formation studies and can lead to confounded conclusions when comparing laboratory and industrial scale cells. Roth et al. investigated the effect of SOC and negative electrode overhang size on self-discharge behaviour after formation in single-layer pouch cells. They attributed the transient self-discharge to SEI growth and equalisation processes, and showed that both effects were less pronounced at low SOCs.371
The cell format and design can influence the formation process. However, detailed studies and a mechanistic insight into this influence are currently not available.
Characterisation of the formation process varies in level of detail and scope. Some characterisation techniques can be integrated into battery cell production and are then applied to each individual cell produced, others provide more detail but are typically limited to a small number of samples due to time-consuming preparation or the need for special setups. The latter are therefore most commonly used in research and development activities. This distinction is also used to group available studies:
1. In-line testing: non-destructive characterisation of full cells. Applicable as a quality control measure in cell production.
2. Sample testing: destructive experiments (ex situ, post-mortem or special cell design required). Sample size is limited to small numbers, but in-depth material characterisation is possible.
This chapter aims to illustrate the options available for characterising the various aspects and properties affected by the process. Existing methods with their main results related to the formation process are summarised in Fig. 15.
The capacity is influenced by the formation process, because during the process Li is consumed, which reduces the amount of cyclable lithium inventory (CLI). Therefore, differences in the loss of CLI can be assessed by comparing discharge capacities.
CE is defined as the ratio between the capacity during the charging process and the capacity of the subsequent discharge process:
Capacity is usually measured by current counting, i.e. integrating the applied current, in CC charge and discharge cycles using low C-rates, e.g. C/22.367 However, it is important that the SOC before charging is equal to the SOC after discharging, which is ideally 0% SOC. This can be achieved by discharging the cell prior to the CE test using the same discharge procedure as for the CE measurement. CR is the ratio between the capacity of two successive discharges:
CE and CR are both useful for assessing side reactions at the negative and positive electrode. CE is also a useful value for estimating the cycle stability of a battery cell without long-term battery cycling tests.375,377 However, it has been pointed out that very high measurement accuracy is required for cycle life assessment. According to Smith et al.377 CE measurement accuracies of at least 0.01% are required, while standard measurement equipment is not sufficient. The requirements for CE measurement accuracy may be lower when evaluating the progress of the formation process, as the changes in CE during formation are significantly greater as discussed above.
CE and CR can be used to study irreversible side reactions at the negative and positive electrode, as studied by Tornheim et al.378 However, it is necessary to distinguish between cells with different type of limitation. Typically, LIBs are inventory limited, which means that cyclable Li inventory (CLI) determines the charge and discharge capacity. This is illustrated in Fig. 16. It can be seen that for a fully discharged cell the CLI is only in the positive electrode (Fig. 16A) and for a fully charged cell the CLI is only in the negative electrode (Fig. 16B). Therefore, the discharge and charge capacities are only determined by the CLI of the negative and positive electrodes, respectively.
Reduction reactions at the negative electrode reduce the CLI of the negative electrode, while oxidation reaction at the positive electrode increase the CLI of the positive electrode, as shown in Fig. 16. Therefore, the reduction reaction at the negative electrode reduces the discharge capacity but does not directly affect the charge capacity. Conversely, the oxidation reaction at the positive electrode increases the CLI and thus increases charge capacity but does not directly affect discharge capacity.
As shown by Tornheim et al.,378 this can be used to distinguish between oxidation and reduction reactions. CE is only affected by the reduction reaction at the negative electrode, giving CE <1. CR is affected by both reactions, but in opposite directions. Reduction reactions at the negative electrodes result in a decrease in CR. Oxidation reaction at the positive electrode result in an increase in CR.
The situation is more complicated for site-limited cells. Here the capacity is determined by the available sites at one electrode. Cells using Li metal as the negative electrode are a typical example.379,380 Here, Li metal provides a large reservoir for CLI. Thus, as long as sufficient Li metal is available, the capacity is determined only by the available sites of the positive electrode.379,380
Another important issue is anomalies in the measurements, such as these caused by the aforementioned negative electrode overhang.367 Here the measured CE depends on the initial state and resting time prior the CE measurement, which can be classified as a short-term path dependency phenomenon.381 As this anomaly can significantly distort the measured value, it must be considered carefully when using CE to evaluate formation progress.
While the capacity measurement mainly evaluates the loss of CLI due to the formation process. Evaluation of the CR and CE provides further insight into the stability and passivation properties of the interfacial layers and allows differentiation between positive and negative electrode passivation. Despite the limitations for inventory limited cells and the bias due to negative electrode overhang, it is a simple yet useful method to evaluate the progression and success of the formation process.
There are three methods for determining self-discharge: the capacity loss method, the voltage hold method and the voltage decay method.369,384 The methods are depicted in Fig. 17. The capacity loss method (Fig. 17A) compares the capacity before and after a period of storage, allowing the separation of reversible and irreversible capacity losses. Disadvantages of this method are that it does not provide information during storage, which limits its ability to identify early rejects, and it requires the cell to be cycled before and after storage.369 The voltage hold or float current method (Fig. 17B) directly measures the current required to maintain a given cell voltage.370,385 This measurement shows time-resolved self-discharge and allows early rejects to be identified. However, it requires continuous and accurate voltage control and the measurement equipment is occupied for the entire test duration.369 The voltage decay method (Fig. 17C) tracks the cell voltage over the storage period. The voltage can be measured at intervals and does not require a battery cycler. Although the voltage drop during storage already provides information about self-discharge, it has been shown that calculating the actual self-discharge current from the differential capacity vs. voltage plot provides time-resolved information comparable to the voltage hold method.369,385 Liao et al. recently developed a detection method that can quickly assess the self-discharge rate of the battery cells by observing the OCV difference during the battery resting process. The OCV threshold is then compared with the OCV difference, and the ratio of the threshold is normalised. When the threshold ratio is compared with the self-discharge analysis results, it is concluded that the OCV detection method is reliable and can therefore be used to quickly analyse the self-discharge behaviour of the battery cells.386
Fig. 17 Illustrations of different self-discharge measurement approaches: (A) Voltage decay method, (B) voltage hold method, and (C) capacity loss method. Reprinted with permission from Roth et al.,369 IOP Publishing, Copyright 2023. |
Overall, the voltage decay method offers greater flexibility in terms of measurement equipment, while still providing a detailed and time-resolved self-discharge measurement. This makes the decay method well suited for cell manufacturing and formation analysis. The voltage hold method, on the other hand, provides reliable and direct measurement of self-discharge. Separation of reversible and irreversible effects can only be achieved by the capacity loss method, making it a suitable complement to voltage-based methods. When interpreting self-discharge measurements, it is also important to consider the influence of intermediate check-ups and cell design on relaxation behaviour, which can distort the evaluation.371,387,388
To conclude, self-discharge measurements also provide information on the passivation properties of the interfacial layers. In contrast to CE and CR measurements, it is much more time consuming, but offers a higher sensitivity. Sometimes it is the only in-line method that shows significant differences for different formation processes.149
Often, EIS spectra are used primarily to qualitatively assess differences between different formation process conditions and material impacts. In general, higher impedances indicate less favourable conditions. Similarly, EIS spectra can be used to assess the completion of SEI formation.139
In addition, EIS allows to distinguish between different process due to varying contributions at different time constants.389,390 In addition, EIS can enable to distinguish between different process contributions, such as interface or diffusion processes.389 However, sometimes process separation is difficult due to similar time constants.389,391 An overview on different processes and their typical frequency domains is provided by Krewer et al.392
To separate the process contributions, equivalent circuit models are often fitted to the impedance spectra. By applying distribution of relaxation time (DRT) analysis to the frequency data, improved identification of the underlying process contributions and definition of more robust models can be achieved.391 Separation of processes via equivalent circuit models allows, for example, to evaluate cell-to-cell differences and to assess the SEI film resistance separate from other process contributions.393
However, some challenges remain. For example, an increase in film resistance could be related to thicker surface films but could also indicate slower ion transport within the film. In addition, interface resistivity is strongly related to specific surface area, which is often unknown and subject to significant uncertainty.
EIS can also be used in combination with other electrochemical techniques. For example, EIS has been combined with cyclic voltammetry.190 Here, it is used to identify distinct regions for subsequent EIS analysis.
EIS provides complementary information to the previously discussed measurements, as it does not evaluate the interfacial layer based on the loss of Li inventory, but directly on the interfacial properties, e.g. the interfacial or film resistances. In particular, a combined evaluation of CLI and interface resistivity using impedance spectra can be used to infer film properties such as ion transport.
DVA can be useful to identify undesired SEI formations. This was shown by Müller et al. who identified differences in the differential voltage curves of NMC622||Gr-based cells formed with and without a CV step and were able to correlate these observations with ex situ analysis.394 It was revealed that the SEI appeared to be more heterogeneous and porous in the cell formed without a CV step, which was also reflected in the appearance of the differential voltage profiles.
To investigate Li plating/stripping, an operando analysis via DVA was introduced by Petzl and Danzer.395 It associates a particular minimum in the DVA curves with stripping of previously plated Li. However, some challenges to the general applicability of using DVA to quantify Li stripping have been raised in the recent work of O’Kane et al.396
In conclusion, DVA is another technique complementary to CLI based techniques. It provides additional information and can be applied as a manufacturing process control measure,397 but the interpretation of measurement results is more demanding.
The determination of the internal resistance can be carried out using a direct current (DC), i.e. galvanostatic intermittent titration technique (GITT)401,402 and intermittent current interruption (ICI),403 or a alternating current (AC), i.e. EIS.401,404 Barai et al. compared the internal resistance of 20 Ah LFP||Gr-based LIB pouch cells obtained with GITT and EIS.405 The determined resistances were found to be in good agreement.405 In pulse-based resistance investigations, however, the time-based measurement resolution of typical cyclers, limits the distinction of different process contributions to the overall resistance.405
In summary, measuring the internal resistance provides information about the cell quality. Resistance can be determined using both DC and AC measurements. While AC-based determination is more time-consuming and requires more precise equipment, DC-based techniques can be easily performed in-line by typical cyclers and provide comparable information on the total cell resistance. However, an EIS measurement is required to differentiate between individual process components.
Comprehensive discussions of the analytical techniques that can be used to investigate LIBs can be found in other available reviews. Verma et al. published a review giving insight into many methods, which can be used to characterise the SEI.88 The review by Zampardi and La Mantia summarises current and emerging techniques applied to analyse the CEI.406 Waldmann et al. provided an overview article on post-mortem analysis methods of LIBs in general.407 Here we will focus on discussing some tests specifically informative about the formation process.
Three-electrode cells allow the individual impedance of the negative and positive electrodes to be evaluated. For example, in the formation study by Chiang et al. the CEI and SEI were examined separately and it was shown that the impedance of the negative electrode was affected by different formation processes, while no significant changes were found for the positive electrode.154 Huang et al. aimed to investigate impedance contributions, but assembled a three-electrode setup by plating Li metal reference electrode in situ on a Cu wire located inside the cell prior to formation.106
Schomburg et al. introduced a novel DVA-based approach to determine the growth behaviour of SEI using three-electrode setup with Li metal reference electrode.408 This allowed to derive continuous SEI growth curves for different electrolytes. The growth curves are shown in Fig. 18. As differences between electrolytes were observed, this method is useful for analysing new electrolyte formulations. The data can also be used to parametrise SEI growth models.408 Potentially the method can also be used as an early quality control measure in production lines, as the DVA has also been used in ageing studies based solely on full cell voltage.409–414
Fig. 18 SEI thickness growth identified by a combined DVA and model-based analysis method for two electrolytes. Reprinted with permission from Schomburg et al.,408 John Wiley and Sons, Copyright 2022. |
Similar to three-electrode cells, half cells can be used to study electrode formation individually. Half cells are used in combination with EIS and cyclic voltametry measurements to study initial SEI formation95,131 as well as effects of material pretreatments415 and electrolyte compositions.416 Solchenbach et al.96 used EIS at 0% SOC before and after formation to obtain the SEI resistance RSEI for various electrolyte solutions in of Gr||Li half cells. The measured impedance spectra are depicted in Fig. 19A and B. The low SOC (blocking condition) was chosen because charge transfer is slow in fully discharged Gr electrodes, which allows the clear separation of RSEI by fitting a transmission line model (Fig. 19C) to the impedance data, which is not possible at high SOCs (non-blocking condition). Although there are variations due to electrolyte dependence, the resistances between the non-blocking (Fig. 19D) and blocking (Fig. 19E) conditions are similar, suggesting that the SEI resistance plays a significant role in the overall intercalation resistance of a Gr electrode under non-blocking conditions.96
Fig. 19 (A) Normalized impedance of graphite||Li half cell at 0% SOC (ca. 2.01 V vs. Li+/Li) at 25 °C after formation (2× C/10 CC full cycels) for the base electrolyte solution 1 M LiPF6 EC/EMC 3:7 (blue squares) and with 1 wt% of the respective additives FEC (yellow triangles), VC (red diamonds) and DiFEC (green triangles) (B) Enlarged view of high-frequency area of the entire impedance spectrum. (C) Used transmission line model (TLM) with two R/Q-elements to fit RSEI to the impedance data. The value of RPore was obtained by analysing additional impedance data and was kept fixed during the RSEI fit. (D) Intercalation resistance RInt (=RSEI + RCT) obtained by fitting the impedance data in non-blocking conditions at 40% SOC. (E) SEI resistance RSEI determined in blocking conditions at 0% SOC. Reprinted with permission from Solchenbach et al.,96 IOP Publishing, Copyright 2021. |
Symmetric cells are often used for EIS experiments and are a useful tool for formation studies. This setup allows the contribution of the negative and positive electrode to be separated. For example, it can be used to investigate whether the SEI has formed completely without interference from the impedance of the positive electrode. Momma et al. provided a detailed review of symmetrical cells.417 Heidrich et al. used electrodes from NMC622||Gr-based cells after formation in symmetrical cells for EIS experiments.418 This showed that the impedance of the negative electrodes did not increase much as the impedance of the positive electrodes as a result of long-term cycling, implying a different growth behaviour of the SEI and the CEI. Further, symmetric cells can be used to sudy the impact of electrolyte compositions on the SEI and cell performance. Analysing EIS and cyclic voltametry mesurements of various electrolytes in Li||Li cells, Hobold et al. found a correlation between a fast Li+ exchange trough the SEI to a stable cycling performance.419
Three-electrode, half cell, and symmetrical cell setups can be used to examine individual electrode contributions separately. This is useful for identifying material-dependent changes. Symmetrical cells are suitable as a post-mortem analysis technique to analyze full cells after formation but can also provide information on the evolution of the passivation layer.
SEM is a powerful and widely available microscopic technique. After formation, typical applications of SEM include morphological studies of the electrode particles, including their coverage, deposition of residues or plated Li. For example, Märkle et al. used SEM in their study of different Gr types and electrolytes to identify the surface coverage and extent of exfoliation of Gr particles.237 Schweidler et al. were able to correlate CEI growth and particle cracking during the first five cycles.76 For identification of particle cracking they relied on SEM images. Fig. 20 shows SEM images of the LNO electrodes during initial charging with increasing electrode potentials, i.e. (A) pristine, (B) 3.8 V, (C) 4.1 V and (D) 4.2 V. The technique reveals that at higher voltage particle cracking occurs.
Fig. 20 Top view SEM images of the LNO cathode taken (A) after 1 h OCV and (B) after charging to ∼3.8 V, (C) ∼4.1 V and (D) ∼4.2 V vs. Li+|Li in the first cycle. Reprinted with permission from Schweidler et al.,77 John Wiley and Sons, Copyright 2020. |
Transmission electron microscopy (TEM) enables to gain deeper insight at smaller length scales. However, unlike SEM, TEM often requires more elaborate sample preparation and is generally not designed to examine large number of samples. Fig. 21A shows one of the earliest TEM images of the SEI on Gr.420 Today, cryogenic transmission electron microscopy (cryo-TEM) allows the study of sensitive chemical phases without damaging the interfacial layer,421 allowing the microstructure of the SEI to be studied. For example, Huang et al. found that in carbon black||Li half cells, a predominantly amorphous SEI of approximately 2 nm forms on the carbon black during an initial C/10 cycle, as shown in Fig. 21B.422 Extending TEM with electron energy-loss spectrocopy (EELS) also provides high resolution structural and chemical information.423 For example, Chen et al.222 obtained and analysed EELS spectral images of Si negative electrodes after cycling for a designed and a conventional electrolyte. The comparison of the images shows that the SEI of the designed electrolyte (Fig. 21C) is thinner compared to the traditional electrolyte (Fig. 21D). In addition, the LiF content near the particles of the designed electrolyte (Fig. 21E) is higher than that of the conventional electrolyte (Fig. 21F). Recent progress in establishing in situ TEM analyses of battery electrodes has been discussed in a review article by Yuan et al.424
Fig. 21 (A) Early TEM micrograph of SEI on a discharged artificial graphite negative electrode. Reprinted with permission from Besenhard and Winter,420 International Union of Pure and Applied Chemistry, Copyright 1998 (B) Cryo-TEM images of a carbon black after a single cycle in EC/DEC electrolyte. The image was taken in the fully delithiated state. Reprinted with permission from Huang et al.,422 American Chemical Society, 2019. (C) and (D) are high-angle annular dark-field imaging images of Si particle surfaces with a designed electrolyte with 2.0 Mol LiPF6 in tetrahydrofuran(THF)/2-methyltetrahydrofuran (MTHF) (mixTHF) 1:1 (C) and a conventional electrolyte with 1.0 Mol LiPF6 in EC/DMC (D). The red numbers indicate the 5 positions at which the EELS spectral images were analysed, with 1 being at the layer surface and 5 being at the inner layer. The LiF distribution of the analysed area is shown in percentage for the designed electrolyte in (E) and for the conventional electrolyte in (F). Reprinted with permission from Chen et al.,210 Springer Nature, 2020. |
The combination of SEM and TEM can provide interesting insights due to their distinctly different scopes and capabilities. For example, Bhattacharya et al. studied formation of SEI on Gr-based electrodes when they applied different voltage scan rates in a linear sweep voltammetry experiment.425 SEM images showed a difference in layer morphology, with the SEI being more uniform and tubular.425 By performing TEM measurements, the authors were then able to determine the microstructure of the formed SEI, which consisted of nanocrystalline domains within an amorphous structure.420,425 Cheong et al. also used a combination of TEM and SEM to study the formation on the alternative negative electrode material, i.e. Co3O4 nanoparticles. By combining the datasets from both methods, changes in the SEI morphology became apparent.75 TEM visualised the SEI layer after formation and after cycling and determined its thickness.75 SEM, which covers a larger area, showed the SEI around the particles.75
When electron microscopes are equipped with energy-dispersive X-ray spectroscopy (EDX) capability, the elemental composition of samples can often be determined directly after imaging, making the technique even more valuable.426 For example, among other analytical techniques, Stockhausen et al. used SEM and EDX. Results are shown in Fig. 22. While Gr surface did not show any significant structural changes as shown in Fig. 22A, SEI deposits are visible on the Gr surface in the EDX as shown in Fig. 22B. The scope and limitations of electron microscopy in the context of LIB research have been reviewed by Wu et al.426
Fig. 22 SEM (A) and EDX (B) images of a graphite surface in NMC622||graphite cells with EC/DMC 1:1 + 3% VC electrolyte after formation, respectively. Reprinted with permission from Stockhausen et al.,101 Elsivier, Copyright 2022. |
Atomic force microscopy (AFM) is an alternative method for morphological analysis, e.g. for determining the thickness of the SEI. Using special cell designs, the method can be applied in situ, as described for amorphous carbon,427,428 HOPG-based electrodes,52,428,429 MCMB428 and Si-based electrodes.222,430 As an example, the evolution of the SEI morphology on a thin layer of amorphous carbon during an initial charge with a consistent voltage scan rate is shown Fig. 23.427 The AFM images show minor changes of the surface morphology at potentials higher than 1.5 V (Fig. 23B). The surface roughness and features exhibit their most significant changes between 0.8 and 0.6 V vs. Li/Li+ (Fig. 23E and F) due to formation of organic decomposition products. During the remaining charge, the morphology changes at a slower rate. Such a potential dependent SEI growth of the SEI was observed is also observed for the other mentioned electrode types.
Fig. 23 Evolution of the surface morphology on a carbon model electrode and a 1 Mol LiPF6 in EC/DEC 1:1 electrolyte obtained via in situ AFM during the initial charge with a constant voltage scan rate of 1 mV s−1. The voltage scane was stopped and held at OCV (A), 1.5 V (B), 1.25 V (C), 1.0 V (D), 0.8 V (E), 0.6 V (F), 0.5 V (G), 0.4 V (H), 0.3 V (I), 0.2 V (J), 0.1 (K) and 0.02 V (L) vs. Li/Li+ for several minutes to gather the AFM data. The blue asterisk marks the same feature in troughput all images. Reprinted with permission from Steinhauer et al.,427 American Chemical Society, 2017. |
Furthermore, distinct variations in SEI growth behaviour were observed when different electrolytes were used. This highlights the relevance of AFM in the development of novel electrolytes and the elucidation of the structural evolution of the SEI. An overview of the application of AFM in the context of LIB research is given in the review by Weidong et al.431
While SEM and TEM are mostly complementary techniques as they cover significantly different length scales, AFM is a frequently used alternative and especially suitable for in situ analysis.
XPS is a very surface sensitive and frequently applied technique that can be used to determine surface layer composition. An example schematic of an F 1s XP spectrum is shown in Fig. 24 for a Gr-based negative electrode after formation. It is illustrated that each of the peaks of the overlapping areas can be assigned to a chemical state, i.e. LiPF6 or LiF, on the surface. The area under each peak reflects the number of atoms in the respective chemical states. Thus, XPS enables both identification and quantification of chemical components.
By combining XPS with sputter depth profiling (SDP), it can also be used to measure thickness. For example, Mao et al. determined the layer thickness for electrode samples.105 The chemical composition was monitored before and after parts of the layer had been removed by sputtering. This allowed the composition of the SEI to be determined and the thickness to be approximated. Several advanced XPS-based approaches have been developed to determine the thickness of the SEI based on depth profiles that rely on sputtering432 or a variation of X-ray energy.55 By measuring XPS at multiple positions spatial homogeneity of the SEI and CEI can be deduced.433
XPS is usually performed ex situ. However, efforts are being made to establish in situ XPS measurements and are summarised in the review by Wi et al.434 Another recent development in photoelectron spectroscopy is the lab-scale hard X-ray photoelectron spectroscopy (HAXPES),435 which uses X-rays of higher energy than conventional lab-scale XPS, thereby increasing the depth information. This allows thicker layers to be studied without the need for SDP, which is in particular relevant to study the formation of Si-based electrodes, which are known to form thicker SEI.436
Time-of-flight secondary ion mass spectrometry (ToF-SIMS) is an alternative technique for mapping of elements and chemical species. Due to its excellent lateral resolution, it provides valuable additional insights. For example, Veryovkin et al. used ToF-SIMS to analyse Gr-based negative electrodes before and after formation, as well as after long-term cycling.437 It allowed them to observe distinct differences between the cycling protocols. However, the authors point out that the large number of data sets obtained with ToF-SIMS could not be evaluated in detail. This may be one reason why ToF-SIMS-based analyses of SEI are not yet as common as XPS analyses. An advantage of ToF-SIMS over XPS is the ability to make continuous measurements during depth profiling experiments, which allows to follow the compositional changes as the layer is removed.438 ToF-SIMS has been shown to be very powerful in combination with XPS. Gauthier et al. analysed the surface layer formed on LTO-based electrodes.439 While surface composition information obtained by the two approaches is mostly comparable, some results are also synergetic. For example, the observability of titanium in XPS, with an information depth of approximately 10 nm, allowed the sputtering depth of their ToF-SIMS experiments to be estimated. Furthermore, ToF-SIMS detected titanium-containing species on the positive electrode, which was not possible with XPS alone. Sui et al. used a combination of ToF-SIMS with SEM-EDX to study the surface properties.440 This allowed them to map the elemental composition of a SEM image with EDX and scan the surface with the highly surface sensitive ToF-SIMS to obtain a high resolution map of surface species distribution.
When it comes to characterising the SEI and CEI using vibrational spectroscopy, such as Raman and Infrared (IR) spectroscopy, a recent review by Weiling et al. gives a comprehensive overview of the scope of each method.441 They conclude that these techniques will play a key role to monitor chemical compounds, and highlight the importance of in situ techniques. Vibrational spectroscopy techniques certainly can add complementary information. This can be seen in very early studies by Aurbach et al., where IR spectroscopy and XPS was used.442 XPS showed that halides, such as LiF, were part of the layer, while IR spectroscopy enabled the authors to identify Li alkyl carbonates as the main components.
Several studies demonstrate the synergistic effect of sample-based characterisation, which can help to understand fundamental aspects of the interphase layers. For example, a broad combination of surface analytical methods was used by Bhattacharya et al.189 Optical microscopy was used operando to monitor surface damage, SEM gave further insight into the morphology, EDX provided elemental distribution, TEM, XPS and Fourier-transform infrared spectroscopy (FTIR) each provided information on the chemical composition and structure. Also Lu et al. highlighted the importance of combined morphology and composition analysis.93 They further included EIS in their analysis, which enabled to show that not only thickness, but also the composition of the SEI contributes to the measured impedance.
Surface analysis techniques are very important in characterising the formation process, as they provide a detailed insight into the composition, thickness, morphology, and heterogeneity of the interphase layers. However, each method has its advantages and disadvantages. Often only a combination of different methods provides the comprehensive picture needed to gain the mechanistic insight required for knowledge-based process optimisation.
Archimedes principle is an in situ method to determine the gas volume as result of formation cycling and was introduced by Aiken et al.443 The technique was later used by Self et al.175 and Leißing et al.184,185 in combination with gas chromatography mass spectrometry (GC-MS) to identify the species generated when using different electrolyte additives during formation. By equipping commercial LIB cells with a gas sampling port, Schmiegel et al. were able to monitor the volume and nature of the evolving gas in situ at defined cell voltages.444
As gas evolution is closely linked to the decomposition reactions, the combination of surface and gas analysis can provide insight into the reactions and the build-up process during the formation step. For example, Ellis et al. used a combination of GC-MS, the gas volume determination using Archimedes principle and surface analysis using XPS analysis.165 The authors showed measurable differences in surface composition when the cells were not degassed after formation, which could be attributed to follow-up reactions of the formation gases within the cell.
In addition, analysis of the electrolyte, e.g. by liquid chromatography (LC), can provide further insights into the underlaying degradation. Numerous studies have been carried out using, e.g. liquid ion, or with mass spectrometry. Comprehensive reviews of electrolyte analysis have been provided by Nowak et al.445 and Stenzel et al.446 Nuclear magnetic resonance (NMR) spectroscopy is a powerful analytical technique that has been widely used to understand degradation pathways in LIBs. Ex situ and, with special cell designs, in situ measurements are possible,447e.g. to study the ageing of electrolytes. A review and an overview of the historical development of NMR in the context of battery research has been provided by Hu et al.448
Gas analysis is most important for understanding the reaction mechanisms during. Since the gas is also involved in the reaction, it is important to get a detailed understanding of the gases produced, where GC-MS are suitable methods. Similarly, analysis of the electrolyte composition by NMR can contribute significantly to the understanding of the reaction processes. A complete picture requires identification of surface layer composition, electrolyte composition change, and gas evolution and composition.
Li plating is most commonly detected by optical inspection, e.g. visually or by microscope after opening the cell. Visual inspection is straightforward and reliable as no special equipment is required. Despite its simplicity, visual inspection is usually sufficient to judge whether Li plating has occurred during formation, as can be seen for two electrodes shown in Fig. 25. It can even give a rough estimate of the amount and therefore the severity of Li–metal deposits. Janakiraman et al. have written a comprehensive review of established methods for characterising Li plating in LIBs.449
Fig. 25 Photographs of graphite-based negative electrodes from multi-layer pouch cells after formation at either (A) 0.2C or (B) 2.0C with clear differences in the visual amount of Li plated. Reprinted with permission from Münster et al.,132 IOP Publishing, Copyright 2015. |
A specific example of detection of plating in formation studies is given by Mao et al. who found a visual increase in plated Li for reduced formation time.105 Only with very slow formation was no plating observed at all. After long-term cycling, which showed only minor differences between formation protocols, a smaller part of the electrodes appeared to be covered with Li. The authors concluded that some Li plating was reversible.
The post-mortem detection of Li plating is possible not only by optical examination but also by NMR spectroscopy. Münster et al. varied the formation C-rate between 0.2C and 2C.132 After 2C formation, optical inspection showed that large parts of the Gr-based electrode were covered with deposited Li, which could also be confirmed by SEM images. Using NMR spectroscopy and inductively coupled plasma-optical emission spectroscopy (ICP-OES), the difference in the amount of plated Li was quantified to be up to 5% of the total capacity.
Glow discharge optical emission spectroscopy (GD-OES) can also be used to determine the depth-resolved elemental composition of samples. GD-OES has been shown to be particularly useful for detection of Li on negative electrodes.450 Ghanbari et al. studied differently aged Gr-based electrodes with GD-OES and were able to estimate the Li distribution of the samples.451 They were also able to distinguish between different plating morphologies, i.e. island-like deposits vs. more homogeneous, layer-like coverage of the particles. This approach was later adapted and revised by Flügel et al. to be applicable to Si-Gr composite electrodes, where Li plating is less understood than for purely Gr-based negative electrodes.452
Li plating must be excluded during formation. Therefore, usually optical inspection is sufficient for formation studies. However, to effectively mitigate Li plating during formation, additional information on the plating morphology may be helpful to identify the root cause. This is where advanced techniques such as GF-OES can be useful.
Since the formation process is affected on considerably different time and length scales, different modelling techniques must be used. Frequently used computational techniques, together with their typical length and time scales at which they are applied, are summarised in Fig. 26.
Fig. 26 Overview of the simulation methods used to depict phenomena during the formation process as well as their typical time and length scales. Adopted with permission from Röder et al.,496 John Wiley and Sons, copyright 2019. |
Wang et al. published a comprehensive review of modelling approaches for the SEI.453 The review by Horstmann et al. focuses on multi-scale models.42 Quantum mechanical methods used to investigate the main mechanisms involved in SEI reduction reactions of organic solvents are reviewed by Ramos-Sanchez et al.454 Theoretical work on the CEI is summarised in a section in Xu's critical review.275 Atkins et al. also devote a section to the simulation of interphases, discussing both interphase layers, i.e. SEI and CEI.455
This section highlights modelling approaches specifically related to the formation process. The section is structured according to the relevant areas of application, which are the decomposition reactions at the solid|liqud interfaces, interphase layer growth, interphase layer nanostructures and gas evolution.
The electrochemical stability window of state-of-the-art electrolyte has been studied in many DFT simulations.341,456–458 In addition, quantum chemistry (QC) calculations can be used to obtain binding energies and redox potentials of common salts, solvents and additives and to study the redox reactions.459 Examples for computed reduction potentials are shown in Fig. 27. Such calculated values can be used as an initial assessment to screen and pre-select promising electrolyte components.
Fig. 27 Reduction potentials of electrolyte components according to the first principal simulation. Reprinted with permission from Wang et al.,453 Springer Nature, Copyright 2018, who compiled the data of Delp et al.459 |
The reaction mechanisms of the formation process are very complex, making them difficult to describe mathematically. There are many possible reaction pathways, often involving multi-step reactions with many intermediates. AMID simulations were used to investigate reduction pathways. For example, the decomposition mechanisms of EC163 and LiPF6460 in the electrolyte bulk and EC in the proximity of electrodes, i.e. Si246,461 and Gr,462 were investigated with this simulation method. Further, AMID simulations were used to compute redox potentials of additives.463 AMID simulations have been also used to gain a better understanding of the reaction mechanisms at the positive electrode. For example, Leung studied the initial stages of EC decomposition on ideal Li0.6Mn2O2 surfaces at constant potential using DFT and AMID.169 Computerised reaction networks (CRN)464 and their combination with machine learning techniques (CRN-ML)465 can be used to construct large reactions networks from first principle calculations.
Combined DFT and AMID simulations were used to calculate electron transfer through SEI components,461 thermodynamic and transport properties of electrolytes,466,467 decomposition potentials of solvents468 and additives.463 Furthermore, these methods can be used to study the influence of the local solvent concentration,293,469 the degree of lithiation246 or the electrolyte composition at the interface.286,470 These aspects must be considered together to predict the SEI composition.
There are two main tasks that computational methods can perform with respect to interfacial reactions. They can be used to quickly screen possible electrolyte components by examining their decomposition potentials. In addition, AMID simulations can be used to investigate complex reaction networks, even near the interface. The combination of these aspects may allow computational design of composition of the interphase layers.
Continuum models can be used to describe the growth of interphase layers in a lumped fashion. These models usually simplify the reaction mechanisms and kinetics by reducing them to a single mechanism and assuming homogeneous interphase properties. The growth rate is limited by various passivation mechanisms. Understanding the growth limitation of the interphase layer is essential.11
Several transport mechanisms that determine the self-limiting nature of the SEI are discussed. These mechanisms are illustrated in Fig. 28 and were discussed by Horstmann et al. in their review on multi-scale models of the SEI growth in detail.42 In general, these mechanisms are also applicable for the formation process. Equations and references for these mechanisms are summarized in Table 2.
Fig. 28 Schematic of considered leakage mechanisms to reaction site enabling SEI growth. Reprinted with permission from Single et al.,479 John Wiley and Sons, Copyright 2018. |
Mechanism | Equation | Parameters | ||
---|---|---|---|---|
Symbol | Term | Unit | ||
Electron tunneling (ETM)476 | i 0 | Exchange current denstity of the reduction reaction | A m−2 | |
α | Reaction symmetry | — | ||
η | Overpotential | V | ||
T | Temperature | K | ||
κ ET = κET0e−βLSEI | ||||
κ ET 0 | Reaction probability in direct contact | — | ||
β | Tunnelling barrier | m−1 | ||
Electron conduction (ECM)471,472 | J k | Molar diffusion flux of species k | mol cm−2 s−1 | |
c k | Molar concentration of species k | mol m−3 | ||
Solvent diffusion (SDM)473 | D s | Diffusion coefficient of the solvent | m2 s−1 | |
c s | Solvent molar concentration | mol m−3 | ||
z | Coordinate variable | m | ||
Interstitial diffusion (IDM)479 | c 0 | Average Li concentration in the SEI | mol m−3 | |
D | Diffusion coefficient of the Li interstitials | m2 s−1 | ||
ϕ | Potential of the active material vs. Li|Li+ | V | ||
T | Temperature | K |
The electron tunneling mechanism (ETM) originates from quantum mechanics, where it describes the possibility of an electron passing through a zone that is normally prevented by an energy barrier. The mechanism can be implemented using a modified Tafel equation.
The electron conduction mechanism (ECM) assumes that the rate is determined by the transport of charge-carrier driven by potential gradients across a dense SEI.471 Colclasure et al. implemented this mechanisms in a single-particle model that includes reaction kinetics and transport of species within the SEI film.472
The solvent diffusion mechanism (SDM) assumes that the transport of the solvent or salt from the electrolyte bulk to the reaction site is the limiting factor in continuous SEI growth. The process is modelled by assuming Fick's law of diffusion for the solvent components.473
The diffusive interstitial mechanism (IDM) has been proposed by Shi et al.474 Small radicals, such as Li atoms, are charge carriers. Shi et al. calculated the concentration of Li interstitials within a compact Li2CO3 layer to be between 4.9 × 109 and 2.3 × 1016 cm−3 depending on the electrode potential. The Li interstitials enable the propagation of decomposition reactions in the SEI layer.42,342,474
Several simulative and experimental studies suggest that the initial growth after the first atomic layer of SEI is established on the surface, is limited by ETM.408,475,476 The tunneling barrier of SEI has been investigated in several studies. Lin et al. performed DFT simulations to calculate the electron tunnelling barrier of Li2CO3, Li2O, and LiF and a Li metal electrode.475 They concluded that the initial capacity loss is driven by ETM until a thickness of approximately 2–3 nm is reached, which was also found to change under compression/expansion.475 Leung et al. concluded that electron tunnelling is sufficiently suppressed by a 7 Å insulating oxide SEI in a combined experimental and first-principles study.477 A thickness of about 10 Å was calculated by Benitez et al. using AMID and DFT simulations to sufficiently block electron transport through pure LiF or Li2O films on Si negative electrodes and varying degrees of lithiation by Benitez et al. using AMID and DFT simulations.461
Beyond the tunneling regime, knowledge of the limiting mechanism is less profound. Using the experimental long-term storage data at different electrode potentials published by Keil et al.,383,478 Single et al. contrast the discussed mechanisms by comparing their plausibility.479 The authors conclude that IDM is the only mechanism that adequately explains the experimentally observed dependencies.479 The analysis of the same data was later extended to include the time dependence of the SEI growth mechanisms, focusing on the comparison between SDM and IDM.480 It was shown that IDM explains both the SOC and time dependence, while SDM reproduces only one of these dependencies.480 They conclude that IDM is the main driver for the SOC-dependent SEI growth.480 This is in agreement with the DFT study by Soto et al., which identified IDM as responsible for electron transfer.342
Schomburg et al. analysed SEI growth curves obtained during formation using DVA.408 Using an SEI growth model that considers a concurrent ETM and IDM for two electrolytes, the model-based analysis indicated that the tunnelling barrier shifts as a function of electrolyte composition. It was also shown that the SEI growth obtained during a slow initial formation cycle cannot be explained by ETM alone. As shown recently, growth functions can be also identified with data driven models such as neural ordinary differential equations.481
A dependence on the current direction has been reported by Attia et al.482 To describe this phenomenon, Das et al. introduced a model based on the assumption that the SEI can be considered as a mixed ion-electron conductor.483 By this and by introducing a Li concentration dependence for the electronic conductivity in the SEI, the growth of the layer is promoted or suppressed depending on the lithiation current direction.
Typically, SEI growth models assume a solid layer of a single SEI species and a limiting transport mechanism. However, it is conceivable that the evolution of the SEI morphology may also influence the limiting transport processes. To investigate this possible influence, Single et al. introduced a one-dimensional framework that takes into account the porosity of the SEI.484 This approach allows to simulate the evolution of SEI thickness and morphology along the interphase thickness and incorporates different rate-limiting transport mechanisms, i.e. SDM in the SEI pores and IDM or ETM in the solid SEI phase.484 The model predicts the often reported two-layer structure with a constant ratio between the thicknesses of the inner and porous outer layers.484
Growth models typically neglect mechanical stress. While these variables can be neglected for comparison in storage tests, they play an important role in formation cycling. Mai et al. modified a Pseudo 2D (P2D) battery model using finite-strain theory to account for the interplay between large electrochemical-mechanical deformations at the particle and electrode levels. This was used to model the influence of porosity changes on the mechanical stress of each cell component.485 Kolzenberg et al. introduced a chemo-mechanical model that incorporates the mechanical-stress inducted SEI growth.486 The model is in good qualitative agreement with the experimentally observed SEI cracking during lithiation and healing during delithiation of Si particles, leading to accelerated capacity fade during cyclic ageing.486 However, the model has not yet been tested for formation cycling. Other continuum-level models that consider degradation mechanisms due to mechanical stress have been reviewed by Zhang.205
Model-based analysis of SEI properties after formation is also possible. Witt et al. have applied such an approach to investigate long-term evolution of the SEI properties.487 The model comprises a P2D and SEI model capable of calculating EIS and electrochemical discharge curves.487 This allows to determine physically insight such as SEI thickness or interfacial SEI|liquid interface area.
Simulation of SEI growth can be used to design formation cycling protocols. However, it remains a challenge to identify suitable growth limiting processes. The situation is even more complicated for materials with significant volume expansion, such as Si, because the mechanical properties of the interface layers must also be considered. The goal of the formation simulation methods discussed is to reduce the number of experiments required to design formation cycling protocols that are tailored to the material and cell design.
Although first-principle studies provide spatially resolved information on the SEI and have made progress on the time scale,497 they are still limited to the ns range and are therefore unable to describe complete formation cycles. Compared to ab initio simulations, kMC allows longer time scales to be simulated while preserving atomistic detail.162,493,498
Methekar et al. developed a 2D kMC model to theoretically investigate the effect of process parameters, such as charging voltage, current density and temperature, on the initial SEI layer formation and suggest that there is an optimal temperature to minimise initial capacity loss.490 This model was extended by Ramos-Sanchez et al. to a three-dimensional coarse-grained kMC (CG-kMC) model, allowing them to track the growth rate as a function of layer thickness.454 The model was used as an example to simulate the layer growth for a multi-step reduction of EC to LEDC using a fixed adsorption rate and DFT-based reaction rates.454 Esmaeilpour et al. performed a simulation study of SEI formation using a 2D kMC model with reaction rates obtained from QC calculations.306 Their simulation shows qualitatively good agreement with experimental observations of the SEI morphology for a solution-mediated growth scenario.306 Similarly, based on reaction mechanisms obtained by AIMD and a stochastic analysis method that accelerates CRN-ML simulations,494 Spotte-Smith et al. used a kMC model to perform a comprehensive study of the initial SEI formation and evolution for varying potentials and electron tunneling barriers.
Röder et al. dynamically couple a kMC surface film growth model with a macroscopic continuum battery model to simulate SEI formation.495 In this way, the structure of the passivating film directly influences the reactions at the particle surface. In a first simulation study, they investigate the effect of two particle sizes on SEI formation for an electrolyte composed of EC and LiPF6. The simulations show faster SEI growth on larger particles. In subsequent work, the same multiscale approach was extended by combining the kMC model with a P2D model and an advanced coupling method, allowing to account for distribution of concentrations and potentials.492,496 The novel multiscale model was used to simulate various formation protocols. Simulation results are show in Fig. 29. While slow charging led to locally heterogenous structures and homogeneous distributions of species along the x-axis (Fig. 29A), the opposite was observed for fast charging protocols (Fig. 29B).496
Fig. 29 Results of a multi-scale simulation of SEI growth. Shown are the composition and film structure of a kMC instance at t = 269 s during the first charge applying (A) a low C-rate of 0.07C and (B) a high C-rate of 2C at different positions x along the negative electrode thickness. The positions shown are left (x = 9.2 µm), middle (x = 27.6 µm) and right (x = 46 µm), which are close to the current collector, middle of the active material layer and close to the separator, respectively. In addition, the highest and lowest values for the species SEI1 from four parallel kMC instances are shown. Finally, the potential at the electrode surface (s), left and right of the film (s) and the electrolyte (s) is shown. Reprinted permission from Röder et al.,496 John Wiley and Sons, copyright 2019. |
In contrast to the continuum models used to simulate SEI growth, the models discussed in this section provide a detailed insight into the molecular structure of the interphase layer. They allow to combine insights from complex decomposition mechanisms and SEI growth mechanisms. Theoretically, it enables to reveal differences in the SEI structure without the need for sophisticated experimental surface characterization methods. The goal of these simulations is to tune the electrolyte choice and cycling protocol to achieve a desired interphase layer structure. However, it is usually not known how the ideal SEI and CEI should be composed to achieve long life and safety. These aspects need to be worked out in order for the methods to reach their full potential.
In first-principle and kMC simulations, gases are considered as products and reactants of the reactions. For example, Leung predicted the formation of CO gas when examining two-electron reduction pathways of EC and assuming fast electron tunneling.163 In the case of slow electron tunneling, the reduction of EC was calculated to yield C2H4 and CO2 instead.163 This is consistent with the findings of Alzate-Vargas et al. who reported the presence of C2H4 in the later stages of their simulation, covering the first 100 ns of SEI formation.497 They also discuss the relevance of the consideration of gas molecules, as they did not observe the formation of gas bubbles.497 However, given the uncertainties regarding potential entrapment, temporary influence of these molecules on SEI growth, or change in density they suggest to consider the gas molecules in the simulations.
Gas evolution and its effects on battery performance have been simulated in mechanistic continuum models. These models assume that the generated gas is inert and reduces the volume fraction of electrolyte. Seo et al. introduced a model that discusses how gas evolution from electrolyte decomposition and impurities causes swelling in Li-ion batteries.499 Their model explains the non-linear increase in cell resistance with changes in volume fraction caused by gas evolution. They apply the model to study the effects of these changes on discharge curves and heat generation.499 The P2D model has been extended by Rashid and Gupta to capture the combined effect of SEI formation and gas evolution on battery performance caused by cyclic degradation.500 Here, both growing SEI and generated gases reduce the electrolyte volume fraction.
Although the modeling approaches presented have been applied to study long-term effects of gas evolution, they are also relevant to simulate gas evolution and its effects during formation. As discussed in this review, gas evolution is critical for fast formation processes. Therefore, the amount of gas produced and its transport through the cell during formation must be considered for the model-based design of formation protocols and in particular for the degassing steps. This aspect is currently poorly addressed in the literature.
Experimental methods suitable for studying the formation process are also reviewed. It is shown that a combination of non-destructive and destructive characterisation techniques can provide valuable insights into the formation process and quality parameters of battery cells. A combination of methods is required to deduce details on the interphase layer properties as a function of the structure and composition. Several techniques are presented and discussed, which can be used as a guide for a detailed study of relevant phenomena of interest.
Computer-aided engineering techniques have become essential for understanding and optimising the formation process. However, few of the available tools are systematically used to support process development. It is shown that the multitude of influential parameters affecting the process on different time and length scales is a key challenge. Different modelling techniques are reviewed that allow the representation of relevant formation process phenomena.
It was shown that single cycle formation protocols are a promising way to achieve faster formation procedures. However, a key challenge with single cycle protocols is that Li plating must be avoided, and formation may not be complete.
A common definition of when formation is complete is needed for systematic comparison and improvement. Otherwise, no comparison of formation time is possible. This, combined with the lack of systematic variation, limits our current understanding of these formation protocols. Based on our analysis of non-destructive characterisation techniques, we propose to use the CE and CR to evaluate the progression of the formation process. These values can be used to convolute the CEI and SEI contributions. Possibly the change in CE in two consecutive cycles can be used for a general material independent definition of the end of formation. However, measurement accuracy and measurement artefacts caused by electrode overhang in large cell formats are major challenges for this assessment. Therefore, the use of CE and CR alone may not be sufficient and must be accompanied by other methods such as EIS and self-discharge testing.
There is a well-established toolbox of experimental methods that can be used to characterize the influence of the formation process in detail. In this context, surface analysis techniques play a key role. It has been shown that a combination of different techniques is required to gain a comprehensive understanding of the interphase layers. The goal must be to select a comprehensive set of methods that can reveal composition, molecular structure, thickness, ion transport, and passivation capabilities. For example, without a comprehensive set of methods, it is not possible to deduce the difference between improved ionic conductivity and a thinner film.
Due to the large parameter space and time-intensive characterisation, computational approaches are promising to reveal multi-scale and multi-physical correlations or to pre-select promising material candidates, e.g. electrolyte additives. Although simulation studies have already contributed valuable insights to the current understanding of the formation process, it is striking that key influencing factors, such as pressure or gas evolution, are hardly investigated in theoretical work. While simulations have explored the factors influencing battery ageing, their application to the formation process is still limited. Similarly, gas evolution and transport at the electrode and cell level have not been extensively studied in theoretical work. However, gas distribution might be limiting factor for fast formation protocols and should therefore be investigated in more detail. Control of the gaseous species may even prove to be an active means of modifying the formation process, offering further potential for optimisation.
A major problem for knowledge-based optimisation of the formation process is that the relationships between process, structure, and performance are not fully understood yet, despite the progress in understanding key physical aspects and the effects of SEI species on cell performance. To adress this, further knowledge regarding the relationship between electrolyte components is essential, as also emphasised in other articles regarding ageing89 and fast-charging.501,502 To be able to activly form desired passivation layer, also systemtically acquiring more insights on the involved reactions and initial dynamic structural evolution is important.
A methodical comparison of theoretical and experimental work is required. This requires a temporal recording of different physical properties and ideally a quantitative comparison with simulation results. Tracking and comparing the temporal evolution of gaseous by-products could be a simple yet powerful method to parameterise and validate mathematical models. Validation of the temporally and spatially resolved evolution of SEI structure and morphology appears more complex. However, EC-AFM, if possible, in combination with XPS images at selected time points, can provide useful and detailed information on the temporal evolution of the SEI, but is time consuming.
An improved multi-physical understanding also holds potential for model-based operando analysis of formation data for cell quality assessment. It has been shown that formation data can already be used for quality assessment and parameterisation of interphase growth models. Such approaches could even be considered as an integral part of the EOL, potentially leading to a reduction in manufacturing time and cost.
The limitation of commonly used small sample sizes needs to be addressed, particularly given the significant uncertainties associated with the handmade battery electrodes often used in academia. Valid evaluations of reject rates in relation to the formation protocol are often not possible in academia due to limited sample sizes. Larger sample sizes are essential to investigate the effect of the influencing factors and their interdependencies, particularly to assess the impact on battery ageing. The use of cells produced in larger production lines and larger sample sizes is one way to overcome this problem.
There is little information from industry on formation processes. This is likely to lead to differences between academic and industrial perspectives on the formation process, creating a research gap and limiting the transferability of research results to real production lines. Industry expertise and insight is needed to bridge the gap between academic studies and practical implementation.
Additional process, e.g. prelithiation or surface medication, should be investigated in more detail in the context of cell formation as they offer a possibility to influence of formation prior to the formation processes without compromising material and cell format choices. Prelithiation of the negative electrodes can compensate for initial irreversible capacity losses.503 This showed promising improvements in cell quality and performance and should be carefully investigated to better understand its effect, particularly in terms of improved SEI formation for electrodes with increased Si content.249 This is also interesting in terms of gassing, as the SEI is partially built before the actual formation cycle.
Finally, it has been shown that many of the current challenges are even more important for next generation materials such as Li, Si or high-voltage positive electrode materials. These materials are often associated with significant implications for the formation process design. A knowledge-based understanding and a systematic use of computational methods can significantly reduce time and cost for the development of material-tailored formation processes which improve the cells’ sustainability, durability, and safety.
Interdependencies between the electrochemical conditions during formation and the materials selected, as well as the lack of a standardised evaluation of formation, often make it difficult to directly compare existing studies, which sometimes prevents general conclusions from being drawn. Therefore, it is highlighted that a key challenge for a knowledge-based design of the formation process lies in an interlocking of material and cell design on the one hand and the process design on the other hand. This is also illustrated in Fig. 30. The discussed experimental and theoretical methods must be applied systematically to reveal the mechanisms and then utilize the knowledge by computational process engineering.
Most important future perspectives are summarised below:
•There is no agreed criterion for when the formation process is complete. Therefore, a chemistry independent criterion for the end of formation is needed.
•Gas evolution during formation is high and affects the overall process dynamics. In some cases, degassing even plays a rate-limiting role. Degassing must therefore be considered as an active part of the formation process.
•The effect of passive materials, i.e. separator design, and cell formats, i.e. pouch or cylindrical cells, also influence the formation process and need to be investigated.
•A better understanding of the relationship between the SEI/CEI structure and the performance metrics is required.
•Different computational methods are available at different scales, covering the most relevant aspects of formation. They should be systematically combined to optimise the process.
•Systematic and standardised characterisation of battery cells after formation with sufficient sample quality and size is essential to allow meaningful comparison of different formation strategies and data-driven analysis.
•Next generation materials often increase the significance of the formation process. Thus, formation must be studied more extensively for these materials.
Reflecting on the current state and future directions of the formation process, further advances are needed at both the materials science and process engineering levels to tailor optimal formation processes to specific chemistries, cell formats and applications. More systematic research is needed to identify key relationships between the formation process and performance metrics, and to pave the way for novel, affordable, high quality, safe and sustainable cell generation with knowledge-driven formation process design.
Footnote |
† Cell conditioning is also occasionally called cell finalisation. Both refer to the same part of the manufacturing process. Throughout the rest of the script, we only use cell conditioning. |
This journal is © The Royal Society of Chemistry 2024 |