The State of Understanding of the Lithium-Ion-Battery Graphite Solid Electrolyte Interphase (SEI) and Its Relationship to Formation Cycling

Abstract An in-depth historical and current review is presented on the science of lithium-ion battery (LIB) solid electrolyte interphase (SEI) formation on the graphite anode, including structure, morphology, composition, electrochemistry, and formation mechanism. During initial LIB operation, the SEI layer forms on the graphite surfaces, the most common anode material. The SEI is essential to the long-term performance of LIBs, and it also has an impact on its initial capacity loss, self-discharge characteristics, rate capability, and safety. While the presence of the anode SEI is vital, it is difficult to control its formation and growth, as they depend on several factors. These factors include the type of graphite, electrolyte composition, electrochemical conditions, and temperature. Thus, SEI formation and electrochemical stability over long-term operation should be a primary topic of future investigation in the LIB development. This article covers the progression of knowledge regarding the SEI, from its discovery in 1979 to the current state of understanding, and covers differences in the chemical and structural makeup when cell materials and components are varied. It also discusses the relationship of the SEI layer to the LIB formation step, involving both electrolyte wetting and subsequent slow charge–discharge cycles to grow the SEI.

This paper is a comprehensive review of the science of the LIB anode SEI layer and its relationship to electrolyte wetting, formation cycling, and cell lifetime. It spans from the basic science of the SEI formation interfacial physics and reaction mechanisms to the applied science of reducing formation cycle time and increasing LIB lifetime. The time period covered is from the discovery of lithium-ion intercalation in graphite (1979) up to the present day, and it offers insights into the SEI formation mechanism, chemical and morphological properties of the SEI, and relationship to formation cycling and cell lifetime. This paper will appeal to the entire LIB research community and the broader energy storage community as a whole. Given that it deals with an intricate combination of surface chemistry, electrochemistry, and reaction mechanisms, it will also appeal to chemists and chemical scientists in other fields.

Background
The anode SEI layer is formed from the so-called "lithium inventory" of the cathode and electrolyte salt, which is the total amount of lithium available for building the SEI and initial charging of the cell, and there is a delicate balance between the ideal surface area 3 resistance thereby decreasing power density. [1,17] During the first full cycle, 10% of the original capacity is generally consumed in irreversible SEI formation. [18] Therefore, the total surface of the anode should be minimized from an energy density or cell cost standpoint. However, the minimization comes with a performance tradeoff -low anode surface area means lower power density (capacity at high C rates) with solid-state diffusion limitations. In contrast, high anode surface area is beneficial to power density, but much greater lithium inventory is consumed when passivating the surface to form the SEI layer, thereby decreasing energy density. Section 7 includes an overview and understanding of carbon/graphite properties and related SEI formation.
There is also a secondary connection of the SEI layer to LIB safety, and it comes into play once the anode is fully passivated. To avoid lithium plating or dendrite formation at the anode during charging over the life of the cell, capacity is often kept about 10% more than that at cathode [18] (N/P ratio of 1.1 where "N" is the negative electrode, or anode during cell discharging, and "P" is the positive electrode, or cathode during cell discharging) to prevent internal electrical shorts. Therefore, this extra anode material must also undergo SEI layer passivation adding to the cell cost and diminishing the total cell energy. Optimizing the N/P ratio is important for minimizing initial lithium inventory loss and decreasing initial irreversible capacity. For long-term capacity retention and Coulombic efficiency, optimizing only the capacity ratio would be insufficient because SEI continuously grows and consumes electrolytes and lithium ions when it is not well formed. [19,20] Better understanding of the state-of-the-art graphite SEI layer composition and morphology is an important step towards growing improved SEI layers that prevent continuous decomposition of electrolyte on the graphite surfaces. The anode SEI layer is composed of precipitates from reduced decomposition of solvents, salts, lithium ions, and impurities in the electrolyte due to their instability at the anode potential operating window. [1,21] It forms mostly during the first charge, but the formation continues slowly and gradually after first cycle until the SEI layer is fully developed, adding to the complexity of modern LIB formation protocols. An optimized SEI layer is expected to 4 have negligible electrical conductivity and high electrolyte diffusion resistance while having high lithium ion selectivity and permeability. Once it is properly formed, further decomposition reactions with salts and solvents are prevented since electrons cannot transfer to or through the layer (the increased electronic resistance increases the potential on the graphite surface and shifts the surface potential to within the stability window of the electrolyte). However, in reality, the SEI layer gradually thickens during repeated charge-discharge cycles due to electron exposure to electrolyte or electrolyte diffusion to the graphite surface, although the layer thickness growth after a few charge-discharge cycles is not nearly as great as the amount during the first cycle. The gradual thickening of the layer further consumes Li ions, solvents, and salts and increases cell resistance.
This continuous SEI layer growth during the formation cycling process lowers cell capacity and Coulombic efficiency.
It is worth mentioning briefly that a "SEI-like" layer forms on cathodes, as well, by oxidation reactions of electrolytes at high potentials [22,23], but its impact on cell performance is generally less. Recent studies involving lithium-manganese-rich (LMR) NMC materials (Li 1+x Ni y Mn z Co 1-x-y-z O 2 ) for EV applications, show high capacities when operated at high voltage. [3,14,[24][25][26][27][28] This cathode material has an operating window of 2~4.8V vs Li/Li + and capacities of 200~250 mAh/g [24], but only 150 mAh/g within the typical operating voltage window (3~4.2V). [15] As the voltage approaches 4.7V vs Li/Li + (or even less), decomposition takes place on cathode surface during charge or storage by oxidation of electrolyte solvent organic carbonates (ethylene carbonate, dimethyl carbonate ethyl methyl carbonate, propylene carbonate, etc.). [21,22,29,30] Since these carbonates have oxidation (highest occupied molecular orbital (HOMO)) and reduction potentials (lowest unoccupied molecular orbital (LUMO)) around 4.7V and 1V vs Li/Li + , respectively [21], they are decomposed by electro-reduction at the anode below 1V and by electro-oxidation at the cathode above 4.7V during charging or storage. The oxidation potentials of these carbonates are further reduced at elevated temperature (LIBs in vehicles or portable devices experience locally increased temperatures) to 4V at 40 o C and 3.8V at 60 o C. [22,[30][31][32][33][34][35] Ethers and esters are not typically stable above 4V. [36,37] Other species in the electrolyte, such as lithium compounds that are partially reduced at 5 the anode and diffuse to the cathode, have even lower oxidation potentials. Wursig et al. reported SEI formation at 4.3V vs Li/Li + and even at 25 o C on various cathode materials. [38] Hence, at high potentials, cathodes suffer from increases in resistance from SEI-like passivation layers as well as from loss or migration of active materials such as Mn and Co. When initial charge-discharge cycles and storage time are extended, the resistance at the cathode increases even more than that of the anode. [39] Despite the importance of understanding the formation, composition, morphology, and long-term structural and chemical evolution of the anode SEI layer, these topics are not yet fully understood because of analysis and measurement difficulties. In fact, the SEI layer formation mechanism is much less understood than the resultant chemical and physical properties themselves. The SEI is quite thin, a few hundreds of angstroms, and sensitive to moisture and oxygen in the air that may convert SEI components into different forms before or during analysis. [40][41][42] Because of the environmental sensitivity, SEI analysis requires inert and well-controlled conditions. Functional properties for an ideal SEI layer are high electrical resistance and high lithium selectivity and permeability. Physical ones are a thickness close to a few Å, high strength, tolerance to expansion and contraction stresses (the SEI layer must accommodate expanding and contracting sub-surfaces during charging and discharging, respectively), insolubility in the electrolyte, and stability at a wide range of operating temperatures and potentials. Actual SEI layers seem to not yet have enough of these properties because it has been found that they keep growing over repeated charge-discharge cycles. This growth is closely related to lithium loss from both the electrolyte salt and cathode lithium inventory, as well as lithium diffusion resistance at the liquid interfacial zone adjacent to the SEI layer and within the SEI itself. The lithium consumption and diffusion resistance cause an increase not only in the overall cell resistance but also in anode potential. The increase in the anode potential is attributed to a lower number of Li ions in the electrode after consumption at the SEI. This increased anode potential also induces a similar increase in cathode potential to maintain a charge cutoff potential. When the cathode potential increases and reaches a certain point, the cathode crystal structure rearranges 6 and distorts due to oxygen loss and transition metal shifting. [3,14,24,25] The electrolyte also becomes less stable at higher cathode potentials, which leads to solvent oxidation on the cathode surface (for example LiPF 6 in ethylene carbonate (EC): dimethyl carbonate (DMC)) is oxidized around 4.5V vs Li/Li + during charging). [43][44][45][46][47] This gradual SEI growth on the anode negatively affects cathode potential and stability. Therefore, forming a robust and stable SEI layer on the anode carbon/graphite is essential for long LIB lifetime and high capacity retention.
SEI formation generally takes days because scan rates are slow, ~±C/5 down to ~±C/20, [13,15] to form a denser SEI structure rather than a highly porous one. After the first charge-discharge cycle, formation cycles generally repeat at different scan rates and/or different temperatures to build quality SEI layers. The longer the times and greater the number of charge-discharge cycles, the more expensive the process becomes, which also either lowers cell production rate or increases capital expense (i.e. more cycling stations required). If the electrical energy is not "recycled" (i.e. using the energy of one cell after charging it from the primary electricity source to charge another adjacent cell), the cost further increases. Reducing the time for SEI formation would provide higher production rates without needing extra space, equipment, and energy, eventually reducing battery pack and plant costs.

Energetics of Anode SEI Formation
In LIBs an aprotic salt solution with low-molecular-weight organic solvents are the most widely used electrolytes. These electrolytes undergo decomposition at the graphite anode, and the SEI layer is formed from these decomposition products, which then dictates initial performance of the cell and long-term capacity fade characteristics. Therefore, the question is can the electrolyte decomposition be minimized or controlled to provide predictable performance of the cell. 7 the electrolyte is the difference between the energy of the LUMO and HOMO. This window is shown as E g . If A is above the LUMO energy, then it will reduce the electrolyte, and, likewise, if C is below the HOMO energy, it will oxidize the electrolyte.
The energy separation between the anode and cathode needs to be as high as possible to increase the energy density of the redox pair. The organic electrolytes used in LIBs have oxidation potentials around 4.7 V vs. Li + /Li and reduction potentials close to 1.0 V vs. 8 should have high Li-ion conductivity and negligible electronic conductivity. The electronically insulating property of the SEI prevents further reduction of the electrolyte on the graphite surface, while the ion conductive nature allows permeation of lithium ions to the graphite surface and provides pathways for the desired ion intercalation. To avoid cracking of SEI layers due to stress from a volume change of graphite during intercalation and de-intercalation and to avoid further passivation reactions, the molecular force between the SEI layer and graphite surface should be strong. Physically, the SEI layer should be strong or flexible enough to accommodate the volume change (expansion during charging and contraction during discharging) of the anode during the cycling process. Ideally, the SEI layer should be uniformly distributed over the graphite surfaces.
The chemical composition of the SEI should contain stable and insoluble compact inorganic compounds such as Li 2 CO 3 rather than metastable organic compounds such as ROLi and ROCO 2 Li (where R is an low-molecular-weight alkyl group), [48,49] [50] Inorganic products are hard to dissolve but can also diffuse into an electrolyte when surrounded by soluble organic products. Li 2 CO 3 is generally abundant on a graphite anode than LiO 2 because of low concentration of lithium on surface of the graphite anode. [51] The concentration of LiO 2 can be increased on lithium metal anode.
From a historical standpoint, the SEI has been thought of as having a bilayer type structure. The layer near the interface of the electrolyte is assumed to be porous and less dense, composed of a large portion of organic components, and filled with electrolyte. This outer, organic layer may undergo further reduction, so its morphology may change in subsequent cycling. The inner layer adjacent to the graphite is presumed to consist mostly of inorganic compounds that protect the anode surface and prevent reduction.
Thus, it is assumed to have a denser morphology with lower porosity. In recent studies, the SEI structure shows a bilayer structure in general, but in reality is more complicated. [51][52][53][54] For example, according to the results of Takenaka's hybrid Monte Carlo (MC)/molecular dynamics (MD) reaction simulation, inorganic salts such as Li 2 CO 3 are abundant near the anode surface and distributed within the whole SEI film, becoming Li 2 CO 3 junctions for the organic lithium carbonates and stabilizing the SEI film. [54] Other recent computational studies have also shown detailed and complicated structures, even though they were based on many simplifying assumptions. Considering real-world LIB systems involving side reactions, impurities, and uneven current distribution, it is likely that SEI structures are even more complicated than those depicted by fundamental simulations.
Recently Lu et al. studied the morphological evolution of the SEI during the formation process. [55,56] Figure 2 shows SEM micrographs at two different magnifications of the graphite anode surface from their study at different degrees of polarization during the first charge. The SEI thickness increased as the formation cycle proceeded (i.e. as the potential of the anode moved towards the intercalation potential). According to their model, the SEI at the beginning of the formation process contained mainly loosely held organic polymer compounds. As the potential was lowered, the SEI layer transformed into a more compact structure of inorganic salts. Figure 2 and Figure 3 captures this morphological evolution. [55,57] The SEI layer formed at the graphite basal plane differs in morphology and chemical composition from that formed at the edge plane. The SEI formed at the basal plane does not need to have ionic conductivity, but it does need to be electronically insulating and impermeable to other electrolyte components. Since lithium ions cannot intercalate into graphene layers across the basal planes, these planes are ionic insulators and do not contribute to reversible capacity. Hence, SEI formation at these locations should be  The chemical composition and morphology of the SEI are affected not only by the electrolyte, but also by the chemical compositions and morphologies of carbon/graphite surfaces. In the case of the 1M LiPF 6 in EC:DMC electrolyte and highly ordered graphite, the SEI at the edge plane is thought to be several times thicker (several nm) than that at basal plane. A thicker layer on the edge plane is consistent with the observation of higher reaction current at the edge plane than the basal plane. [58] On the edge sites for this particular case, the SEI is mainly composed of loosely packed inorganic lithium carbonates, organic lithium alkali carbonates and polymeric compounds on the electrolyte side. On the graphite side of the edge sites, the SEI is mainly composed of densely packed LiF, Li 2 O, and Li 2 CO 3 . In between these two phases, there is an intermixed zone forming a trilayer structure. Overall, LiF and Li 2 CO 3 make up more than half of the SEI M A N U S C R I P T layer. [59,60] On the basal sites, the SEI is composed of lithium carbonates more than LiF on the electrolyte side.
On the graphite side of the basal sites, the SEI is composed of similar portions of Li 2 O, LiF, and lithium carbonates with small portions of polymeric compounds. LiF in the SEI is typically found in fluorine system electrolytes such as LiAsF 6 , LiPF 6 , and LiBF 4 .
Depending on LIB operating, anode sampling, and analysis conditions, the compositions of the SEI may vary even with the same electrolytes and electrodes used in a cell. LiPF 6 salts are unstable in elevated temperature and may precipitate into LiF during storage or operation. Lithium carbonates can be also decompose and form LiF after reacting with HF. Hence, LiF may be found in the SEI more frequently when there are other reactions before or during the surface analysis. HF formation, particularly observed in the case of LiPF 6 based electrolytes, is considered to dramatically affect the performance of LIBs by attacking the SEI layer. HF production during the SEI formation process is due to the reaction between decomposition products of the LiPF 6 salt and traces of water in the liquid electrolyte phase and/or adsorbed on the graphite surfaces. In recent studies, the amount of LiF found in the SEI still varies considerably from one study to the next. In particular, computational simulations rarely show LiF formation because they generally do not consider impurities like water causing HF production or self-decomposition from a salt, a poor assumption.
For the case of soft carbon in the same electrolyte, polymer and solvent reduction products are more prevalent than salt reduction products. [61] For other salts such as LiBF 4 , LiTFSI, or LiBETI, the percentage of LiF is small and other carbonated species comprise most of the SEI layer. [62] In general, SEI layers are composed of densely packed inorganic compounds such as Li 2 O, Li 2 CO 3 and LiF on the graphite side and loosely packed inorganic and organic species like Li 2 CO 3 , lithium alkyl carbonate (ROCO 2 Li) and polymer on the electrolyte side. [63] There are also other studies that argue large portions of inorganic Li compounds such as LiF are also found on electrolyte side [64][65][66]. These components, formed by solvent, lithium salt, and electrolyte additive decomposition, are neither uniformly distributed nor well-ordered within the SEI layer.
13 These semi-quantitative concepts about SEI compositions are much less debated than those hypotheses with respect to exact composition, morphology, structure, and formation. The reasons for the uncertainty and inconsistency among different studies arise from analysis difficulty, different electrolyte compositions, different types of Table 1. List of known chemical compounds formed on the surface of carbon/graphite SEI layers ("Present" denotes that the compound was identified in the references given, and "Not Present" denotes that the compound was not identified). [ It may also appear as a reaction product of semicarbonates with HF, water, or CO 2 .
ROLi [73,[75][76][77][78] Most commonly found in the SEI formed in ether electrolytes like tetrahydrofuran (THF), but may also appear as DMC or ethyl methyl carbonate (EMC) reduction product. [72] It is soluble and may undergo further reactions. [79] LiF [72,74,80] Mostly found in electrolytes comprising of fluorinated salts like LiAsF 6 , LiPF 6 , LiBF 4 . It is a major salt reduction product. HF contaminant also reacts with semicarbonates to give LiF byproduct. Amount of LiF increases during storage. [ It is mainly formed due to water contamination. [89,90] It may also result from reaction of Li2O with water or with ageing. [75] Li 2 C 2 O 4 [75,78] It is found to be present in 18650 cells assembled in Argonne National Laboratory containing 1.2 M LiPF 6 in EC:EMC (3:7) electrolyte. Li carboxylate and Li methoxide were also found in their SEI. [75] HF [91,92] It is formed from decomposition LiPF 6 and the water in the solvents. It is highly toxic and can attack components of the cell.

M A N U S C R I P T
A C C E P T E D carbon/graphite, various SEI formation processes, and other physical and environmental conditions (i.e. temperature).

Formation Mechanism of SEI Layer
There are various reduction processes that compete with each other on the carbon/graphite surface during charging. The reactants are solvents, salts, additives, and trace air impurities (such as water). Electrochemical reaction rates differ depending on their intrinsic properties such as reductive potential, reduction activation energy, and 15 exchange current density. They also depend on reaction sites (basal or edge), predecomposed precipitate sites, and many other different anode surface conditions. [58,[93][94][95][96] Temperature, concentration of electrolyte salt, and reduction current rate significantly affect SEI formation as well. [97] When all of these variable conditions are mixed, it is difficult to analyze the SEI formation mechanism. However, it is generally accepted that  16 lithium ions and solvent co-intercalate into the graphene planes. If this co-intercalation is excessive, the SEI layer may not fully develop because of continuous exfoliation. Figure   4 illustrates these processes graphically.
In general it has been accepted that the SEI formation is a two-step process. During the first step when the graphite electrode is polarized, the components in the organic electrolyte undergo reductive decomposition to form new chemical species. In the second step, these decomposition products undergo a precipitation process and begin forming the 19 scanning electron microscopy (SEM) [38,55,64,112,116,[120][121][122][123], transmission electron microscopy (TEM) [65,118,121,124,125], scanning tunneling microscopy (STM) [119,126], atomic force microscopy (AFM) [64,85,126], and Ellipsometry [115] have been implemented to image the surface features and morphology of the SEI. TEM can also show surface crystallinity, in-situ interface formation, and lithiation/delithiation in operando. [127][128][129] AFM is a useful tool for studying SEI morphology and thickness because it can measure differences in depth at Angstrom resolution. Ellipsometry is a non-destructive optical tool that measures thickness and roughness of thin films by using reflectance ratios, but its weakness is that the measured signal depends not only on thickness, but also material properties.

Effects of Carbon/Graphite Properties on SEI Formation
Carbons are widely used as LIB anodes because of their stability and low working potential. Graphite is a crystallite and the most stable allotrope of carbon. It has perfect stacking of graphene layers in AB form and in some cases ABC form. In general, aggregates of perfectly stacked graphite crystallites exist with different orientations in an electrode. Graphite has a redox potential very close to Li/Li + , is safe, is the most stable form of carbon, is environmentally benign, and has low (pre-processed) cost. A lithium atom is intercalated between the graphite layers to form an intercalation compound (i.e. LiC 6 ) during LIB operation. [19,20] The intercalation reaction prevents the deposition of metallic lithium on the graphite surface and avoids dendritic growth making these types 20 of LIBs safe. The lithium-ion charge is also maintained, essentially eliminating the activation energy associated with the formation of a chemical bond. The carbon is reduced to maintain charge balance. Figure 7 shows an aggregate graphite particle and the graphite layers within that particle.
In a crystallite of graphite, the two characteristic surfaces are referred to as basal and edge planes. The surfaces parallel to the graphene layers are called basal planes, and the surfaces normal to the graphene layers are edge planes. Lithium predominantly intercalates into the graphene layers through the edge planes in the direction parallel to 21 well as the chemical and physical properties of the SEI layer. In turn, these property differences will affect electrode performance during early life and the shape of the longterm capacity fade curve.
The SEI layer forms differently depending on composition and structure of carbon/graphite surface. The key factors for SEI formation are particle size, basal-toedge-plane ratio, pore size, degree of crystallinity, and surface chemical composition (adsorbed species). [132,133] The surface area of small particles is greater than that of large ones for the same weight. Smaller particle size generally causes more edge sites, as Few studies have been reported on the effect of the graphite particle size distribution, porosity, surface roughness, surface chemistry and crystallinity. [120,132,[134][135][136][137] Graphite with low specific surface area tends to exfoliate more. As the surface area is decreased by heat treatment, the number of surface defects increases and the exfoliation tendency of the graphite further increases. Above a critical specific surface area of 0.2 m 2 /g, the exfoliation tendency is suppressed in the absence of surface defects. [120] Graphite particle sizes with surface areas of 1-5 m 2 /g may mitigate exfoliation without introducing excessive irreversible capacity loss. The degree of graphite crystallinity is also an important factor in SEI formation, and highly ordered graphite is preferred for high-energy LIB anodes. When the particle size is small, irreversible capacity loss increases due to larger surface area for lithium inventory loss during SEI formation (a side benefit, though, is that the power density increases with smaller anode particle size). 22 Increasing graphite size is one way for lower irreversible capacity loss and higher energy density, but large particle sizes increase the probability of exfoliation.
Chemical composition on carbon/graphite surfaces may also affect the exchange current density and potential for SEI formation, as well as wettability of electrolyte, chemical adhesion between carbon and the SEI layer after decomposition, and between carbon and electrolyte before decomposition. [64,93,[138][139][140][141][142][143][144] The presence of oxygen species on graphite surfaces increases the reduction potential vs Li/Li + and helps early SEI formation before lithium intercalation. The presence of these species has been shown to be vital in SEI formation by serving as nucleation sites for electrolyte decomposition, and a lack of them can hinder the decomposition reactions and increase the exfoliation of the graphene layers. [145] Oxygen species on graphite surfaces can be attached by heat treatment in various environments. Natural graphite usually has oxygen-containing species on its surface, although the degree of crystallinity may not be that high. To increase it, calcination in air produces more crystallites and increases the adsorbed oxygen-containing species. EMI-BF 4 also has a higher oxidation potential, but its ionic conductivity is lower due to the high solvent viscosity. Lithium bis(oxalato)borate (LiBOB) also has less thermal reactivity. It is used as a salt by itself or an additive in an electrolyte. LiBOB stabilizes the graphite structure effectively even in pure propylene carbonate (PC) and facilitates SEI formation on the surface of electrode materials. On the other hand, its solubility and conductivity in other common solvents such as EC and PC are inferior. However, these limitations are improved by using a more appropriate solvent such as dimethyl sulfite (DMS) with γ-butyrolactone (γBL). [148,149] M A N U S C R I P T  25 reaction rates. Considering reduction potential, reaction enthalpy and activation energy, VC is reduced before EC and PC do during a reduction cycle (charge) although the reduction products of VC may not be as stable as those of EC and PC. [153] In industry, many different proprietary additives are used in even more combinations together with variations on the formation protocol, and the resulting SEI structures are closely guarded.

Relationship between Electrolyte Decomposition Reactions and LIB Formation Protocol.
In 1979, Peled first used the term solid electrolyte interphase (SEI) for the LIB anode passivation layer [140], and, in 1990, Dahn et al. discovered the advantage of using EC in the electrolyte for forming the SEI. [157] Before the use of EC was commonplace, PC was the most widely used LIB electrolyte solvent, which was highly compatible with lithium metal anodes. Early LIBs implemented lithium anodes, but dendrite growth was a problem in terms of safety and long-term performance. [158][159][160][161] Once the discovery of graphite as a safe, high-performing anode intercalation material was made, its major drawback was also quickly discovered. PC easily co-intercalates with lithium ions and exfoliates the graphene layers during electrolyte decomposition, while also releasing propylene gas. Attention shifted to using amorphous carbons having little crystallinity because they tended to exfoliate much less in PC based electrolyte solutions and showed good reversible capacity. The problem with these materials, though, was high initial capacity loss due to thick SEI layer formation. When Dahn et al. found that EC reduced the first-cycle capacity loss (due to a much thinner SEI layer formation) and increased the stability of the SEI by mitigating exfoliation of graphite, solvent mixtures high in EC concentration were mainly used. As the appreciation of having a stable, durable SEI has grown since the early 1990s, much effort has been dedicated to: 1) improving its formation by using additives that result in better SEI-layer architectures; 2) modifying the anode surface for improving exchange current density and charge-discharge reaction kinetics; 3) implementing charge-discharge cycles that enhance layer formation; and 4) developing alternative electrolytes that result in less lithium inventory loss during formation. Few studies have been reported on electrolyte wetting of electrodes, although it is an important factor for reducing SEI formation time and manufacturing resources that directly affect LIB pack cost. In fact, wetting takes the majority of cell production time and involves many of the latter manufacturing steps such as addition of insulators, seals, and safety devices [7]. It takes many hours for electrolyte to completely wet the separator and reach the smallest pores of the electrodes [15], and studying wetting transport phenomena without a complete cell assembly facility is difficult.

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
Wetting electrodes (at low vacuum pressures during electrolyte filling and subsequently at elevated temperature after cell sealing) with electrolyte and forming SEI layers requires ~0.5-2 weeks for the entire process. 27 layer is first formed. The initial wetting process is slow because the electrolyte has to permeate into all pores of the separator and electrodes in a near fully assembled cell.
Evacuating gases out of the pores under high vacuum during cell assembly can accelerate infusion of the electrolyte and enable uniform distribution, although it requires more equipment and processing expense. Even under an evacuated condition, the smallest pores of the electrodes and separator may not fully wet unless they have a higher surface energy than the electrolyte. This situation is due to the competition between hydrodynamic forces at low pressure and non-wetting surface forces (the smaller the pore size, the lower the vacuum pressure needed to make a non-wetting liquid enter a pore).
[55] PF 5 is known as a strong Lewis acid that can also react with traces of Figure 10. Linear carbonate (LC) reduction process (reference groups in parentheses; details are shown in Table 2).

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
An et al. (2015) 32 water to produce HF, eliminate alkyl carbonate from lithium alkyl carbonate, and react with solvent carbonyl groups to produce insoluble ether-containing species. [17,92,145] At higher anode potentials, decomposition products like LiF are usually generated on graphite edge plane sites, which are preferable for nucleation due to a lower energy requirement than on basal plane sites. Fortunately LiF generation is kinetically slow and the amount of decomposition is small above 1 V vs. Li/Li + .
It is well accepted that most of the SEI layer formation takes place within the potential range of 0.2-1.0 V vs. Li/Li + . However, the formation mechanism(s) is highly debated.
There have been two different concepts on the SEI formation process hypothesized, although their final structures are mostly alike. One concept follows a one-electron transfer to the electrolyte at high potential and a multi-electron transfer at low potential. [55,145] In other words, "bulk" lithium compounds (precipitates) are first partially reduced at high potential by a one-electron process due to insufficient electrons Figure 11. Electrolyte salt reduction process (reference groups in parentheses; details are shown in Table 2).

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
An et al. (2015) 33 with low electron transfer resistance from electrode to electrolyte for complete reduction.
At low potential, "compact" lithium compounds are generated from further reduction of preexisting bulk lithium-containing precipitates on the anode surface, or directly by complete reduction of lithium compounds in electrolyte driven by high energy at low potential. The second concept supposes little or no precipitates preoccupying the anode graphite surface at high potential. Hence, electrons can transfer to electrolyte constituents without interference from resistive interfacial compounds. This high electron transfer rate induces compact lithium compound formation at higher potentials, and, as the potential is lowered and the decomposition layer thickens, electrolyte solvent molecules are gradually reduced. Subsequently, partially reduced bulk lithium compounds precipitate on top of the resistive compact layer to a greater extent as the potential reaches ~0.2 V vs.
Proponents of both hypotheses agree that major SEI formation begins around 0.8V vs.
Li/Li + , but it can be higher for certain highly porous carbonaceous active materials [131].
In an ideal situation, the SEI formation will occur prior to lithium intercalation, which prevents co-intercalation of electrolyte constituents. For highly oriented graphite, lithium intercalation occurs at 0.2V vs. Li/Li + or less, but it can start at slightly higher potentials in disordered or porous carbons [64,116,131]. Within the potential range of 0.6-0.8V vs.
Li/Li + , electrons at the graphite surface transfer to the liquid electrolyte containing solvated lithium ions. These cations diffuse towards the graphite particles with an average of four strongly coordinated solvent molecules and an uncoordinated PF 6 anion. [118] Uncoordinated solvent molecules are less likely to accept an electron because they are more stable than those associated with lithium ions. Most decomposition products from electro-reduction precipitate on the anode surfaces, while minor amounts diffuse back into the liquid phase and eventually re-precipitate elsewhere. Decomposition reactions prefer graphite edges because these sites usually include defects that are unstable and tend to be highly reactive. As the reaction rate on edge sites slows due to an 34 lithium ions (cathode lithium inventory), solvent molecules, and salt anions (see Table 2 and Figures 8-11). [145] Below 0.6V vs. Li/Li + , a much greater extent of electrolyte decomposition takes place [55]. Dense inorganic lithium compound formation (i.e. LiO 2 , Li 2 CO 3 and LiF) is highly favorable at edge sites because of low electronic resistance to reducing solventcoordinated lithium ions. Some less dense inorganic species such as lithium alkali carbonates and other organic species generally decompose on basal planes where reactions are less favorable. Since electronic resistance increases when a dense inorganic film is present on the graphite surfaces, subsequent decomposition onto the inorganic compounds will involve only partial reduction resulting in precipitation of loosely aggregated compounds such as lithium alkyl carbonates, or polymers. A portion of the lithium alkyl carbonate can be further reduced to form LiO 2 or Li 2 CO 3 and release ethylene (from EC), propylene (from PC), or CO 2 gases. [131] The release of these gases may cause cracks in the existing SEI layer or even expose new anode graphite surface to electrolyte for further SEI decomposition reactions (further consuming lithium inventory).
A second or even third formation charging is usually needed to completely form a stable SEI layer for long LIB lifetimes. These subsequent charging half-cycles are often at progressively faster C rates.

Prospects for Improving SEI Properties and Reducing Formation Time
Reducing LIB formation protocol time is necessary to lower production cost of cells (and ultimately packs) and manufacturing capital costs. The process currently lasts from about 4-5 days up to ~2 weeks depending on the cell chemistry, and it consumes a great deal of process energy (low-grade heat and electricity). In addition, it is also a substantial process Anode active material surface coatings and chemical modifications can also improve SEI layer properties. [144,176] The volume change of carbonaceous materials during lithium intercalation is much lower when compared to other anode materials such as Al, Si, Sn, and Sb. [43] It has been found that an increase in disordered carbon anode surface oxygen resulted in low graphite exfoliation and stable SEI formation. [135,143,177,178] Using this property, a thin carbon coating on graphite can be implemented for improving capacity retention. Non-graphitic carbons do not undergo exfoliation to a great extent, but their first-cycle irreversible capacity loss is much higher due to greater surface area.
When graphite is used without a carbon coating, reversible capacity is lower; however, once coated by a high-surface-area disordered carbon, reversible capacity increases because the majority of the SEI layer forms within the thin coating. This surface modification tends to prevent extensive graphite exfoliation. [96,136,179] It is plausible that SEI compounds differ not only depending on the reaction sites, such as edges vs. basal planes, but also depending on the initial surface elemental composition.
Initial precipitates on the non-oxidized graphite surfaces are likely different from those on oxidized surfaces, which could affect subsequent precipitation. Other surface modifications of various types of anode graphite have shown significant effects on SEI formation. [93,94,[136][137][138][179][180][181] M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
An et al. (2015) 36 Generally, a high charging rate during the first cycle results in a porous and highly resistive SEI layer, while a low charging rate results in the opposite SEI characteristics. It has been found that for a 0.5C charging rate during SEI formation, capacity retention at room temperature operation was negatively affected. [113] Also, when the cell temperature was held above 40°C, capacity retention was even more negatively affected for a 0.5C SEI formation charging rate. [114] Hence, a first charging rate between 0.05C and 0.2C is preferred for stable SEI formation. In some cases, though, high charging rate can be beneficial to SEI formation. For example, when TIMREX ® SFG44 graphite was heat-treated in an inert gas at 3000°C, a high charge current of 320mA/g (~1C), showed better reversible capacity in 1M LiPF 6 EC/DMC than a much lower charge current of 10mA/g (~0.03C). [122] In this case, high current decomposed the electrolyte faster than solvents could intercalate into graphene sheets and cause exfoliation. Low charging rates may be beneficial for SEI formation, but they slow cell production rates and increase production cost and plant capital expense. Building a stable SEI with a charging rate greater than 0.5C may require a great deal of further effort on developing proper additives, optimizing cell temperature, and modifying the anode surface chemistry.
At higher temperatures, SEI formation may also be accelerated. SEI layers formed at temperatures around 40°C tend to have more compact lithium precipitates, such as Li 2 CO 3 and Li 2 O, rather than softer, organic precipitates like ROCO 2 Li. However, high temperature may induce LiF precipitation from fluorine containing salts. energies. [189][190][191] Density functional theory (DFT) is more rigorous than MD, and the former is another computational approach in quantum mechanics that solves Schrödinger equation. It estimates the electronic structures in atomic and molecular systems, but it is limited to smaller simulation sizes than MD because of the associated computational intensity. One of the issues in using DFT is weak van-der-Waals-like forces of graphene layers, which can affect calculations for lithium-ion/solvent-molecule co-intercalation into graphite.

Recent Progress in SEI Layer Studies and Prospects for
Computed graphene interlayer binding energy significantly varies depending on DFT functionals. [192] Local-density approximation (LDA), a well-known and simple functional, underestimates the binding energy of graphite interlayer as shown in Figure   12. The binding energies from experiments were 31-52meV/atom. [193][194][195] Another issue in using quantum simulations is that the simulations are typically not suitable for estimating competing reactions.
DFT and MD mainly deal with Angstrom and nanometer length scales, respectively.
Because of the small length scales and heavy calculation load, it is extremely difficult to  and Li + (EC) n could cause graphite disintegration by widening the C-C layers (which experience only weak van der Waals forces), and since Li + (PC) n is more massive than Li + (EC) n , the probability of this disintegration process could be higher for Li + (PC) n than Li + (EC) n .

Reduction reactions
Most reduction processes take place between 0.9V and 0.2V vs Li/Li + on highly ordered graphite, but salt products may decompose at higher voltages according to one DFT study. LiF formation, deposition, and radical recombination were found to occur near 2V vs Li/Li + before the main solvent reduction reactions below 0.9V vs. Li/Li + for FEC or below 0.6V vs. Li/Li + for EC. [203] In the case of EC and PC based electrolytes, reduction energies of Li + (EC) 4 were 8 kcal/mol lower than Li + (PC) 4 , [199] which implies that Li + (EC) n is more prone to reduction than Li + (PC) n . Statistical and surface analyses also showed that smaller molecular-weight compounds and salt decomposed on the anode surfaces in the presence of electrolyte, followed by long-chain oligomer compounds. [204][205][206] For the case of 1M LiPF 6   cation was subsequently reduced on the anode and produced CO 2 , aldehyde, and oligomers of alkyl carbonates. CO 2 was generated to a greater extent than CO due to the high activation energy for CO.
Regarding the common VC additive, DFT based MD simulations from Ushirogata et al.
showed that VC reacted with an EC anion radical causing a one-electron reduction of EC, [182] which implies that VC assists the EC reduction process. Although exactly how VC decomposes is not yet clear, experimental and computational agreement has been found that VC increases solvent reduction reaction rates. SEI layers were found to form at about 1 V vs. Li/Li + , and they were composed primarily of inorganic components close to the anode surface and primarily of organic ones close to the liquid electrolyte. In these simulations, however, lithium metal was considered as the anode instead of graphite. It should also be noted that the presence of electrolyte salts were not considered, and reaction gas products from the inner portion of the SEI were not allowed to diffuse outward into the liquid phase (these gases would be removed under real-world formation conditions). Hence, actual SEI layers are composed of somewhat different compounds than these particular calculated ones.

SEI layer composition and ion diffusion
Lithium ion transport in Li 2 EDC, a common component of SEI layers, was studied by EIS experiments and MD simulations with APPLE&P force field and G4MP2 and MP2 levels. [189] The conductivity of Li 2 EDC at room temperature was found to be 10 − 41 activation energy ranged from 64-84 kJ/mol at 393K, indicating that lithium ion transport exhibited a hopping mechanism at high temperature.

SEI on anode metalloid or metal oxide.
Metalloid or metal oxides, especially SiO x and SnO x , are regarded as prime candidate materials for high-energy batteries due to high theoretical capacity and earth abundance, [209][210][211] despite Si and Sn having higher operating potentials than graphite (around 0.4V and 0.6V vs. Li/Li + , respectively). One of main problems, however, is that these two materials suffer from extreme volume changes during lithium intercalation and deintercalation, leading to: 1) extensive, unstable, and thick SEI formation; 2) poor longterm mechanical properties; and 3) severe capacity fade.
Volume changes on these composited and blended materials also become smaller, making the formed SEI layers more chemically stable and longer lasting. Coating Al 2 O 3 on these materials also mitigates volume changes. [248,249] Lithium metal is also being reconsidered as a possible anode candidate because it has the highest specific capacity, but it suffers from significant safety concerns related to lithium dendrite growth into and through the separator (electrical shorting). One reason this degradation and potential failure mechanism occurs is because the SEI layer is not uniform and tough enough to prevent dendrite growth. Some of the proposed ways to prevent lithium dendrite growth are hollow carbon nanospheres covering surface of lithium anode, adopting an alternative salt, and electrolyte additives. [250][251][252] M A N U S C R I P T They found 4-(perfluorooctyl)-1,3-dioxolan-2-one improved capacity retention and lowered impedance in high voltage lithium ion batteries. These pre-formed SEI layers were found to protect the cathode from electrolyte decomposition as well as the anode.

A C C E P T E D ACCEPTED MANUSCRIPT
They also tested lithium difluorooxalatoborate, triphenylamine, and 1,4-benzodiozane-6,7-diol as a combined additive and obtained improved capacity retention and lowered impedance for a Li 1

M A N U S C R I P T
A C C E P T E D

Prospects for future understanding.
LIBs for high energy or high power demands like vehicle application require high capacity, high capacity retention, high voltage operation, low cost, low weight and volume, etc. All of these requirements are deeply related to SEI layers. Regarding capacity and capacity retention during the first few charge cycles, batteries lose approximately 10 to 15% of their capacity due to initial anode SEI formation. Afterword, irreversible capacity loss continues due to gradual SEI deposition during long-term cycling, although it is far less than the loss during initial SEI formation. The capacity loss percentage and initial SEI properties vary depending on formation conditions such as anode material surface area, surface properties of the material, anode-to-cathode capacity ratio, temperature, charge rate, charge depth, surface properties of the anode materials, salt/solvent properties, additives, and impurities. Detailed information about the effects of these properties and conditions can be found in previous sections of this paper.
High-voltage operation of LIBs does not significantly affect anode SEI formation directly, but it does have indirect effects. Cathode materials designed for operation at high cell potential (i.e. overcharged NMC or LMR-NMC) release cathode constituents (mostly Mn, some Ni, and a minor amount of Co) into the electrolyte that diffuse through the separator to the anode side, and in turn, induce more SEI formation by increasing the electron conductivity of SEI layer. High-voltage operation of cells also causes electrolyte instability (oxidation) on the cathode surfaces, and a SEI-like layer forms at the cathode that is chemically less stable than its anode counterpart. Hence, under these cell operating conditions, cell impedance increases due to both changes in the anode SEI layer and excessive growth of the cathode SEI layer.

Summary
This paper comprehensively reviews the science of SEI layer formation on carbon/graphite anode surfaces in the LIBs, including structure, morphology, chemical composition, electrochemistry, formation mechanism, and formation cycling. In order to develop shorter, more robust LIB formation protocols, which are needed to reduce cell 44 needed. Furthermore, new formation protocols which develop ideal SEI layers (those that consume minimal lithium inventory during formation and reduce capacity fade during long-term cell operation) in shorter time periods will require an understanding of SEI layer evolution over the LIB life, a subject which is currently not well understood.
It is understood and accepted, however, that the SEI is formed by the decomposition products of the electrolyte solvent molecules and lithium salt, and it critically affects the short-term and long-term performance of the cell. The importance of the SEI layer was given in terms of first-cycle efficiency, capacity retention, and cell cost, as well as the state of understanding of the SEI formation mechanism and methods of analysis and characterization. Various factors that affect SEI formation were also discussed such as anode materials, surface properties, formation current density, electrolyte additives, and cell temperature.
The anode SEI layer covers the graphite surfaces and shields lithium ions from the electrolyte solution, which prevents further electrolyte decomposition. This shielding property enables reversible capacity during extended charge-discharge cycling. However, the SEI formation process consumes lithium ions and electrolyte when generated, resulting in first-cycle irreversible capacity and lithium inventory losses. This irreversible capacity loss may continue if the SEI is not well formed by hindering electrolyte diffusion or allowing unwanted electron transfer from the graphite to the liquid phase.
SEI compositions and morphologies are complicated and differ depending on graphite surface properties, electrolyte, and formation conditions. Several modeling and experimental efforts are underway to address the correct pathways of these electroreduction reactions and elucidate the debate within the LIB research community. SEI analysis is a challenging task due to its thickness being only ~3-100 nanometers and its