Identiﬁcation of Degradation Mechanisms by Post-Mortem Analysis for High Power and High Energy Commercial Li-Ion Cells after Electric Vehicle Aging

: Driven by the rise of the electric automotive industry, the Li-ion battery market is in strong expansion. This technology does not only fulﬁll the requirements of electric mobility, but is also found in most portable electric devices. Even though Li-ion batteries are known for their numerous advantages, they undergo serious performance degradation during their aging, and more particularly when used in speciﬁc conditions such as at low temperature or high charging current rates. Depending on the operational conditions, different aging mechanisms are favored and can induce physical and chemical modiﬁcations of the internal components, leading to performance decay. In this article, the identiﬁcation of the degradation mechanisms was carried out thanks to an in-depth ante- and post mortem study on three high power and high energy commercial 18,650 cells. Li-ion cells were aged using a battery electric vehicle (BEV) aging proﬁle at − 20 ◦ C, 0 ◦ C, 25 ◦ C, and 45 ◦ C in accordance with the international standard IEC 62-660, and in calendar aging mode at 45 ◦ C and SOC 100%. Internal components recovered from fresh and aged cells were investigated through different electrochemical (half-coin cell), chemical (EDX, GD-OES, NMR), and topological (SEM) characterization techniques. The inﬂuence of power and energy cells’ internal design and Si content in the negative electrode on cell aging has been highlighted vis- à -vis the capacity and power fade. (Glow Discharge–Optical 7 Li MAS NMR induces a signiﬁcant phase shift for large chemical displacements (264 ppm). This is why here the peak associated with lithium metal is downwards (Figure 6a). The phase can then be corrected so that the peak rotates upwards Figure 6b,c, but this produces baseline distortions. This is of no consequence given that the objective of this analysis is simply to detect the presence of metallic lithium. Metallic Li is observed after aging at − 20 ◦ C for 30Q, 32E, and 35E cells, and also after aging at 0 ◦ C for 32E cell.


Introduction
Li-ion batteries with graphite-based negative electrodes are today largely widespread in electric mobility applications due to their higher energy density and durability than other storage systems. Currently the main developments in Li-ion batteries concerns the increase of their power density and lifetime by introducing new positive electrode materials or adding silicon in the negative electrode. Lifetime is more particularly studied because certain environmental and operational conditions favor some degradation mechanisms inside the cell that can accelerate the capacity fade [1,2]. Indeed, depending on the cycling and storage conditions, different aging mechanisms can be activated and degrade the electrode components (loss of active materials) or their interfaces (loss of cyclable lithium).
The representative BEV cycling test was carried out at various temperatures (−20 • C, 0 • C, 25 • C, 45 • C) from SOC 0-20% to 100%. For the calendar test, the storage was carried out at SOC 100% and at 45 • C. The same aging protocols were applied to the three cell references in order to obtain easily comparable results. Ante mortem and post mortem analyses at the electrode scale, including electrochemical tests in half-coin cell, SEM, EDX, GD-OES, XRD, and Li NMR characterizations, were performed in order to clearly identify the main aging mechanism(s) that led to the cell performance fading.

Material and Methods
The three chosen commercial cells are from SAMSUNG SDI: INR18650-30Q (noted 30Q thereafter), INR18650-32E (noted 32E thereafter) and INR18650-35E (noted 35E thereafter) and are issued from their same respective production batches. The choice of the same manufacturer was made to ensure that the mechanical design, the separator and the electrolyte composition of the three cell references are similar. To access these internal characteristics, all three reference cells were analyzed. The obtained design characteristics are detailed in the ante mortem section.
The cells were aged using a BEV (battery electric vehicle) representative cycling test according to the international standard IEC 62-660 at various temperatures (−20 • C, 0 • C, 25 • C, 45 • C). The calendar aging was carried out at 45 • C, either at constant voltage (CV) at 4.2 V, or at 100% SOC in open circuit voltage (OCV, meaning that the voltage is not maintained constantly). The voltage difference between CV and OCV calendar aging is between 0.02 V and 0.1 V maximum after 6 weeks aging. The cycling is representative of a typical driving cycle including a CC-CV charge and a discharge profile with current pulses that simulate acceleration (discharge) and braking (charge) phases. As described in the standard IEC 62-660, the cycling test is stopped once the cell has lost at least 20% of its initial performance in terms of capacity, energy or power density, or after 6 months of cycling. The storage test is stopped in calendar conditions after 18 weeks. State of health assessment via electrical performance measurements were realized at 25 • C every 28 days of cycling and every 6 weeks for calendar aging. Such electrical measurements allow tracking of the evolution of some relevant cell characteristics: capacity, internal resistance, and nominal voltage. Every aging profile was performed on 15 cells for each condition. The cycling test and periodical electrical performance measurements were made on a PEC ® SBT 05250 test bench (6 V, 50 A).
After aging, the cells were opened in a glove box under Ar atmosphere (H 2 O and O 2 ≤ 10 ppm). Samples were extracted from each cell electrodes and washed two times in DMC to remove the lithium salt. Electrochemical characterizations were performed on both positive and negative electrodes extracted from fresh and aged cells in coin cell vs. metallic lithium. All electrode disks have been sampled in the middle of the electrode length to avoid being impacted by very local heterogeneities that may more particularly appear in the core area of the winding. The laboratory test bench used was a modular potentiostat/galvanostat/EIS VMP3 from Bio-Logic ® ([−5 V, +5 V], 400 mA).
SEM and EDX analysis were performed using a LEO 1530 Gemini Zeiss microscope. Images of the negative electrodes were taken with a 5 kV acceleration beam voltage and a working distance of 6.0 mm. Complementary characterizations were done on negative electrodes to clearly identify aging mechanism: 7 Li Magic Angle Spinning (MAS) NMR analysis were performed using a Bruker AVANCE III 500 spectrometer with a magnetic field of 11.7 T and a larmor precession frequency of ω 0 = 194.37 MHz and elemental depth profiling has been obtained by GD-OES (Spectruma GDA750).

Ante Mortem Analysis
Such study requires a thorough knowledge of the initial electrochemical performance and the internal design of the cell. For this reason, we started with a detailed ante mortem analysis. Two representative cells of each reference were taken to undergo electrochemical tests in the aim of determining their initial performances in terms of capacity, internal resistance, and nominal voltage. One cell of each reference was dismantled to recover internal components samples (positive and negative electrodes, separator, and electrolyte). By performing dimensional measurements and physical and chemical analyses, the exact design and chemical composition of the three cells was obtained. All results of ante mortem studies are given in Table 1. Positive and negative active materials were determined by SEM-EDX and XRD analysis. Electrolyte composition was obtained by GC-MS analysis. The nature of the polymer matrix and coated layer were identified by FTIR analysis. It is noteworthy that the three batches of cells (equal to 100 units for each batch) showed a very low standard deviation (σ) of cell's characteristics such as weight, capacity, nominal voltage, internal resistance, and energy density, as reported in Table 1. The chemistry of three cells showed similarities regarding the nature of the positive material consisting of a blend in which the principal material was NCA (LiNi 0.88 Co 0.10 Al 0.02 O 2 ). The minority compound of the positive electrode blend was NC (LiNi 0.8 Co 0.2 O 2 ) for cells 30Q and NCA (≈LiNi 0.78 Co 0.20 Al 0.02 O 2 ) for cells 32E and 35E. The nature of separator (AlOOH coated PE film) was the same for the three cell references. Some more significant differences were noted concerning the negative electrode that was composed of graphite with silicon oxide content of 4.5%wt for 30Q cell, pure graphite for 32E cell, and graphite with silicon oxide content of 1.5%wt for 35E cell. The electrolyte composition, analyzed by GC-MS, was not exactly the same but two main solvents (EC:DMC) were common to the three cells. FEC was present when silicon oxide was present at the negative electrode (30Q and 35E). All of them contained succinonitrile in low quantities. This is an additive known to improve the thermal stability and broaden the oxidation electrochemical window of electrolyte [35]. A low amount of methyl propionate as an additive to improve the low temperature performance [36] was detected only in the 32E cell. The cell design was very different for the three references. The 30Q cell had the thinnest and therefore the longest electrodes compared to 35E and 32E cells, which gave it a power design with the ability to accept high currents. On the other hand, 35E and 32E cells had an energy design as the electrodes were thicker and consequently shorter in length. Such a power design was obtained because the thickness of the negative electrode could be reduced by the presence of a silicon compound whose high specific capacity made it possible to obtain a high surface capacity. In the case of the 35E cell, the lower quantity of silicon in the negative electrode led to an increase in its surface capacity, a decrease in its thickness, and consequently allowed association of it to a more loaded positive electrode. Such design is representative of an energy design more advanced than the 32E design where the negative electrode is pure graphite and consequently thicker that the 35E negative electrode.

Aging and Cell Performance Monitoring
Aging effects on cell performance were followed all along the tests. The cell performance evolution is presented in Figure 1. It is noteworthy that each mark on the curves is the average value of the parameter considered measured on 15 cells for each temperature condition. The standard deviation is indicated by an error bar and reached a maximum value of ±3.3% for 35E cell values at −20 • C, but is mostly less than ±1%.
The behavior in cycling is given in Figure 1 (left curves). Cycling at very low temperature (−20 • C) is very harmful. At high temperature cycling (45 • C) the cell performance decreases quite quickly: the effective capacity decreases by about 20% after 300 equivalent cycles for cell 30Q. When cycling at 0 • C and 25 • C, the evolution of 30Q cell capacity is less pronounced: capacity fading is approximately 10% at 0 • C and 15% at 25 • C after 300 equivalent cycles. All those results are in agreement with the manufacturer data sheets [37] which specify a lifetime equal to 250 cycles with charging conditions between 0 • C and 50 • C. After 18 weeks of calendar aging at 45 • C ( Figure 1, right curves) and for the three cell references, the loss of capacity is approximately 10%. As expected, capacity decay during calendar aging is less pronounced than for cycling aging, but not negligible. The performance fading is generally slower during calendar aging than during cycling at the same temperature and the same test time. Calendar aging in CV is more damaging than in OCV. Considering the individual behavior of the three cell references, 32E and 35E cells show a quite different cycling behavior compared to the 30Q cell. They present a fast decay of capacity at 0 • C and, in contrast, very slow performance losses at 25 • C and 45 • C. The calendar aging of 35E cell is very weak compared to the others.  The internal resistance and the nominal voltage were also monitored during aging and their evolution are accessible in the Supplementary Materials, respectively in Figures S1 and S2. Regarding the internal resistance measurements, we observe that aging impacts the internal resistance with the same trend for the three cell references. The internal re sistance increases more significantly during cycling especially at high temperatures (25 °C and 45 °C). For example, the 30Q cell internal resistance increases approximately by abou 35% after cycling at 0 °C and −20 °C and by about 160% after cycling at 25 °C and 45 °C Calendar aging is also more particularly stressful under CV condition. For example, the 30Q cell internal resistance increases by 60% in OCV and by 175% in CV. In terms of the The internal resistance and the nominal voltage were also monitored during aging and their evolution are accessible in the Supplementary Materials, respectively in Figures S1 and S2. Regarding the internal resistance measurements, we observe that aging impacts the internal resistance with the same trend for the three cell references. The internal resistance increases more significantly during cycling especially at high temperatures (25 • C and 45 • C). For example, the 30Q cell internal resistance increases approximately by about 35% after cycling at 0 • C and −20 • C and by about 160% after cycling at 25 • C and 45 • C. Calendar aging is also more particularly stressful under CV condition. For example, the 30Q cell internal resistance increases by 60% in OCV and by 175% in CV. In terms of the internal resistance variation, 18 weeks of calendar aging at 4.2 V and 45 • C is equivalent to 320 cycles at the same temperature. The evolution of nominal voltage shows similarities for the three references. At high temperatures (25 •  Cycling at −20 • C induces faster energy losses of 2% per EFC (Equivalent Full Cycle) for 32E cell than for 35E cell and also for 30Q cell that exhibit a similar loss of 0.2% per EFC. At 0 • C both energy cells, the 32E and the 35E, have significant losses whereas the 30Q power cell is not damaged. Whatever the cell reference is, cycling at 25 • C and 45 • C do not induce huge energy losses; 0.02% per EFC and 0.08% per EFC respectively. According to Figure 2, we can state that the power cell has an optimal working temperature range between 0 • C and 25 • C while the energy cell operates best between 25 • C and 45  Lower temperatures are more detrimental and it is crucial to evaluate the importan of the degradation mechanisms leading to the performance decay. Post mortem analy have been thus performed.

Post Mortem Analysis
Aged cells are disassembled at 0% SOC and samples from negative and positive el trodes are cut out in order to achieve all further electrochemical, physical, and chemi analyses. Figure 3 compares the potential profiles vs. metallic Li in half cells of the fre Lower temperatures are more detrimental and it is crucial to evaluate the importance of the degradation mechanisms leading to the performance decay. Post mortem analysis have been thus performed.

Post Mortem Analysis
Aged cells are disassembled at 0% SOC and samples from negative and positive electrodes are cut out in order to achieve all further electrochemical, physical, and chemical analyses. Figure 3   Regarding the positive electrode, similar behaviors are shown after aging for the three cell types. The aging at low temperature (−20 °C) did not change significantly the potential profile of the positive electrode, which means that cycling at low temperature did not induce any degradation on this level. However, aging at high temperature (45 °C impacted the potential profile of the positive electrode with a significant increase in the electrode polarization observed after aging at 25 °C and 45 °C. The voltage drop was ap proximately equal to 0.1 V in lithiation for the three references cycled at 45 °C. Such posi tive electrode evolution has been reported in the literature [38]. This voltage drop was Regarding the positive electrode, similar behaviors are shown after aging for the three cell types. The aging at low temperature (−20 • C) did not change significantly the potential profile of the positive electrode, which means that cycling at low temperature did not induce any degradation on this level. However, aging at high temperature (45 • C) impacted the potential profile of the positive electrode with a significant increase in the electrode polarization observed after aging at 25 • C and 45 • C. The voltage drop was approximately equal to 0.1 V in lithiation for the three references cycled at 45 • C. Such positive electrode evolution has been reported in the literature [38]. This voltage drop was smaller for the electrode from the cell aged in calendar mode (CV) at 45 • C. Therefore, we can assume that on the positive electrode level, aging at high temperature (25 • C, 45 • C cycling and 45 • C calendar CV) causes damages to active material, whereas aging at low temperature did not cause significant damages. This damaging could be attributed to interfacial phenomena leading to changes in the kinetic properties of the active material.
Regarding the negative electrode, the aging impacted slightly more significantly the negative electrode profiles for 30Q and 35E cells after cycling at low temperature (−20 • C) and calendar aging under CV (45 • C), while the impact on 32E negative electrode seemed to be more negligible. That impact resulted in a small polarization causing voltage drop (0.03 V) in the last LiC 12 /LiC 6 potential plateau (~60-80% SOC) [39][40][41] in lithiation and in the Si feature (~20% SOC) in delithiation. This feature is specifically indicative of the crystalline Li 15 Si 4 phase, which was formed (at 0.053 V) during lithiation of amorphous α-Li x Si and dissipates (at 0.4 V) on delithiation [42]. It is noteworthy that the 0.4 V feature during delithiation was more affected for the 30Q cell that contains a higher amount of silicon. The difference in potential profile characteristics after aging affected how the limiting lower cutoff potential intervened, impacting differently the lithiation condition of Si particles. Considering that, we can assume that the cycling at low temperature (−20 • C) and calendar aging at high temperature (45 • C) damaged the negative active material.
In a first approach, SEM, EDX, and XRD analysis reveal (not presented here) that the positive electrode did not have significant morphological and structural changes after aging, as already noted in the literature for similar aging conditions [43]. As mentioned above, the suspected aging mechanisms on the positive electrode responsible for the potential profile modification (Figure 3) are a crystalline phase transformation on particle surface through transition metal (Ni, Co, Al) dissolution [15,44,45] and/or CEI (cathode-electrolyte interface) growth [46] during high temperature aging. One and/or the other phenomenon could be confirmed by implementing characterization techniques at the atomic scale (highresolution TEM) or at the surface chemistry scale of the material (XPS).
Because the negative electrode focuses the major degradation mechanisms at an electrode scale [15,47], post mortem analysis was focused on that electrode and highlighted some morphological and chemical changes within the electrode. Figure 4 represents the EDX cartography performed on the three fresh negative electrodes extracted from the cell 30Q, 35E, and 32E. EDX cartography and spot analysis allow identification of graphite and silicon particles. The graphite particles have a smooth and angular aspect in contrast to the silicon particles, which are rather rough and spherical. Once the particles are identified, their evolution after aging can be observed. Figure 5 shows the SEM images of fresh and aged negative electrode surface from the three cells (30Q, 32E and 35E). For better readability, only four key aging conditions are represented (cycled at −20 • C, 25 • C, 45 • C, and 45 • C calendar CV).
After cycling at −20 • C, a deposit covering the particle surface was visible for the three cell types. This deposit appeared very important because it partially masked the surface particles and the empty space between particles. Regarding the 30Q and 35E electrodes, the Si particles did not seem as visible as in the fresh electrode but appear fragmented. After cycling aging at 0 • C, a more scattered surface deposition was observable. After aging at 45 • C (cycle and calendar), the particle's surface states looked puffy due to a film surface deposition associated with SEI growth as already reported in the literature [8][9][10][11]48]. This surface film deposition was observable for all aged electrodes. Although the electrodes were washed before being analyzed, 32E electrodes cycled at 25 • C, and to a lesser extent after calendar aging under OCV at 45 • C, showed extra deposition attributed to salt residue.    On the negative electrodes, morphological surface modifications are thus observed for all aging conditions but are possibly the consequence of two different degradation mechanisms, i.e., lithium metal deposition partially or totally oxidized by contact reaction with electrolytes at low temperature and SEI growth at high temperature as largely reported in the literature [4]. XRD analysis revealed (not presented here) that the graphite particles of the negative electrode did not have significant structural changes after aging. However, it is not excluded that graphite cracks or exfoliation of the graphene edges had occurred [27,28]. To detect the presence of metallic "dead" lithium, 7 Li MAS NMR was On the negative electrodes, morphological surface modifications are thus observed for all aging conditions but are possibly the consequence of two different degradation mechanisms, i.e., lithium metal deposition partially or totally oxidized by contact reaction with electrolytes at low temperature and SEI growth at high temperature as largely reported in the literature [4]. XRD analysis revealed (not presented here) that the graphite particles of the negative electrode did not have significant structural changes after aging. However, it is not excluded that graphite cracks or exfoliation of the graphene edges had occurred [27,28]. To detect the presence of metallic "dead" lithium, 7 Li MAS NMR was performed. Furthermore, to gain information on the nature and the thickness of this covering film, elemental depth profiling of the electrode surface were obtained by GD-OES (Glow Discharge-Optical Emission Spectroscopy). GD-OES is thus very information rich. However, dure to the rare availability of this technique, only one reference cell was investigated. The choice was made to characterize in priority the 30Q cell electrode containing the greater amount silicon in weight.
The direct detection signal acquisition mode for 7 Li MAS NMR measurements induces a significant phase shift for large chemical displacements (264 ppm). This is why here the peak associated with lithium metal is downwards (Figure 6a). The phase can then be corrected so that the peak rotates upwards Figure 6b,c, but this produces baseline distortions. This is of no consequence given that the objective of this analysis is simply to detect the presence of metallic lithium. Metallic Li is observed after aging at −20 • C for 30Q, 32E, and 35E cells, and also after aging at 0 • C for 32E cell.
performed. Furthermore, to gain information on the nature and the thickn ering film, elemental depth profiling of the electrode surface were obtain (Glow Discharge-Optical Emission Spectroscopy). GD-OES is thus very in However, dure to the rare availability of this technique, only one reference tigated. The choice was made to characterize in priority the 30Q cell electr the greater amount silicon in weight.
The direct detection signal acquisition mode for 7 Li MAS NMR me duces a significant phase shift for large chemical displacements (264 ppm here the peak associated with lithium metal is downwards (Figure 6a). The be corrected so that the peak rotates upwards Figure 6b,c, but this produc tortions. This is of no consequence given that the objective of this analys detect the presence of metallic lithium. Metallic Li is observed after agin 30Q, 32E, and 35E cells, and also after aging at 0 °C for 32E cell.  Figure 7 shows the evolution of the surface composition for the elem P of the cell 30Q after different aging conditions measured by GD-OES. Th  After cycling at −20 °C, the amount of Li is higher whereas the quantity of O lower comparing to other cycling aging conditions. This could possibly indicate mation of lithium plating in sufficient quantity so that a part of it, not oxidized by lyte, remains active in metallic form. However, in the case of anodes containin GD-OES method is not capable of identifying Li depositions clearly due to form Li silicates [49]. A method development into this direction is currently on the ZSW's labs.
For 25 °C and 45 °C, the Li content is similar like for −20 °C ; however, the co P is higher. In combination with the observed low O content, this indicates SEI gr salt degradation considering the chemical reaction below [50,51].
At 0 °C, the quantity of Li is similar as for −20 °C and 25 °C, but not the quan and P. Most likely, this temperature is in a range where neither one nor the oth aging mechanism is dominating. This matches with the slow cell performance this temperature. NMR shows that Li is not in a metallic state for aging at 0 °C . tected Li could originate from a low quantity of deposited lithium totally oxidize action with the organic molecules of the electrolyte.
At 45 °C calendar aging, the increase of Li coupled with the increase of O decrease of P could reveal a SEI growth rather by solvent degradation than by deg of conductive salt. This mechanism seems to be more pronounced under the CV der the OCV condition, according to the chemical reactions below [50,51].
EC: 2(CH 2 O) 2 CO + 2e − + 2Li + → (CH 2 OCO 2 Li) 2 ↓ + C 2 H 4 ↑ DMC: CH 3 OCO 2 CH 3 + e − + Li + → CH 3 OCO 2 Li ↓ + CH 3 • DEC: CH 3 CH 2 OCO 2 CH 2 CH 3 + e − + Li + → CH 3 CH 2 OCO 2 Li ↓ + CH 3 CH 2 • The arrow notation in the reaction schemes refers to solid products that beco of the SEI layer (↓), and to gaseous products removed from the surface (↑). The d tion (•) in the reaction schemes refers to free radicals which are not stable and w After cycling at −20 • C, the amount of Li is higher whereas the quantity of O and P is lower comparing to other cycling aging conditions. This could possibly indicate the formation of lithium plating in sufficient quantity so that a part of it, not oxidized by electrolyte, remains active in metallic form. However, in the case of anodes containing Si, the GD-OES method is not capable of identifying Li depositions clearly due to formation of Li silicates [49]. A method development into this direction is currently on the way at ZSW's labs.
For 25 • C and 45 • C, the Li content is similar like for −20 • C; however, the contents of P is higher. In combination with the observed low O content, this indicates SEI growth by salt degradation considering the chemical reaction below [50,51].
At 0 • C, the quantity of Li is similar as for −20 • C and 25 • C, but not the quantity of O and P. Most likely, this temperature is in a range where neither one nor the other main aging mechanism is dominating. This matches with the slow cell performance decay at this temperature. NMR shows that Li is not in a metallic state for aging at 0 • C. The detected Li could originate from a low quantity of deposited lithium totally oxidized by reaction with the organic molecules of the electrolyte.
At 45 • C calendar aging, the increase of Li coupled with the increase of O and the decrease of P could reveal a SEI growth rather by solvent degradation than by degradation of conductive salt. This mechanism seems to be more pronounced under the CV than under the OCV condition, according to the chemical reactions below [50,51].
DMC : CH 3 OCO 2 CH 3 + e − + Li + → CH 3 OCO 2 Li ↓ + CH 3 • DEC : CH 3 CH 2 OCO 2 CH 2 CH 3 + e − + Li + → CH 3 CH 2 OCO 2 Li ↓ + CH 3 CH 2 • (4) The arrow notation in the reaction schemes refers to solid products that become part of the SEI layer (↓), and to gaseous products removed from the surface (↑). The dot notation (•) in the reaction schemes refers to free radicals which are not stable and will react with another compound in their environment. Figure 8 shows the GD-OES depth profiles for silicon of the fresh and aged negative electrodes for the 30Q cell. The Si concentration of the negative electrode surface (near the separator) has changed after aging. After cycling at −20 • C and 0 • C, the Si content has increased relatively to the fresh cell. This could be related to the Si particle cracking observed on SEM images and to the migration of soluble Si species from them to the electrode surface. Whereas after aging at 25 • C and 45 • C (cycling and calendar aging) the Si content decreases on the electrode surface probably due to the growth and dissolution of SEI layers.
electrodes for the 30Q cell. The Si concentration of the negative electrode surface (near th separator) has changed after aging. After cycling at −20 °C and 0 °C, the Si content ha increased relatively to the fresh cell. This could be related to the Si particle cracking ob served on SEM images and to the migration of soluble Si species from them to the elec trode surface. Whereas after aging at 25 °C and 45 °C (cycling and calendar aging) the S content decreases on the electrode surface probably due to the growth and dissolution o SEI layers.
It is noteworthy that the Si content, which is initially well distributed in negativ electrode depth in form of the Si particles [49], was found to be more concentrated on th electrode surface after the formation, as well as the low and high temperature aging. Th migration is most likely related to the formation of Li-silicates [49,52], which consum cyclable Li as described by the chemical reaction below.
5SiO + 6Li → Li 2 O + Li 4 SiO 4 + 4Si (5 In the investigated electrodes, there is the additional effect of Si particle fragmenta tion, which could possibly increase this effect at low temperatures. On the other hand, th Si compounds on the surface are partly soluble [49], which is likely to be increased at hig temperatures. In view of the information obtained, a schematic representation of the main degra dation mechanism on the negative electrode surface according to the aging conditions ca be proposed in Figure 9. At low temperature cycling, lithium plating and silicon particl cracking are the two dominant aging mechanisms. It can be expected that at higher C rates both mechanisms become more pronounced [53,54]. At high aging temperatures, SE growth is the main aging degradation mechanism, but the origin of this mechanism seem to be different for cycling and calendar aging. During cycling, the salt degradation is th origin of the thickening of the SEI layer. However, during calendar aging, it is the solven decomposition that is responsible for the same mechanism. Another observed agin mechanism on silicon particles is the formation of Li silicates, especially at high tempera ture. It is noteworthy that the Si content, which is initially well distributed in negative electrode depth in form of the Si particles [49], was found to be more concentrated on the electrode surface after the formation, as well as the low and high temperature aging. This migration is most likely related to the formation of Li-silicates [49,52], which consume cyclable Li as described by the chemical reaction below.
In the investigated electrodes, there is the additional effect of Si particle fragmentation, which could possibly increase this effect at low temperatures. On the other hand, the Si compounds on the surface are partly soluble [49], which is likely to be increased at high temperatures.
In view of the information obtained, a schematic representation of the main degradation mechanism on the negative electrode surface according to the aging conditions can be proposed in Figure 9. At low temperature cycling, lithium plating and silicon particle cracking are the two dominant aging mechanisms. It can be expected that at higher C-rates both mechanisms become more pronounced [53,54]. At high aging temperatures, SEI growth is the main aging degradation mechanism, but the origin of this mechanism seems to be different for cycling and calendar aging. During cycling, the salt degradation is the origin of the thickening of the SEI layer. However, during calendar aging, it is the solvent decomposition that is responsible for the same mechanism. Another observed aging mechanism on silicon particles is the formation of Li silicates, especially at high temperature.

Conclusions
Ante-and post mortem characterizations have provided essential insight to link aging conditions to degradation mechanisms. Thanks to these results, an identification of the degradation mechanisms was performed on three commercial 18,650 cell types after aging according to a BEV (battery electric vehicle) aging profile in cycling at −20 °C, 0 °C, 25 °C, 45 °C , and in calendar aging conditions at 45 °C and SOC 100%. The complementary information obtained from electrochemical, morphological (SEM) and chemical (EDX, GD-OES, and NMR) characterization techniques allowed to identify the main aging mechanisms for each aging condition.
The main degradations are located at the negative electrode and are manifested by the growth of SEI, either by solvent degradation during high temperature calendar aging, or by the growth of SEI by salt degradation during high temperature cycling. The deposition of lithium metal for low temperature cycling leading to the formation of a secondary SEI is identified. In addition, silicon particles suffer significantly during cycling aging at low temperature through particle cracking and disaggregation.
The internal design of the cell influences the aging mechanisms, as high energy cells are more exposed at low temperature to lithium metal deposition due to their thicker electrodes coating which decrease the electrodes' kinetics.

Conclusions
Ante-and post mortem characterizations have provided essential insight to link aging conditions to degradation mechanisms. Thanks to these results, an identification of the degradation mechanisms was performed on three commercial 18,650 cell types after aging according to a BEV (battery electric vehicle) aging profile in cycling at −20 • C, 0 • C, 25 • C, 45 • C, and in calendar aging conditions at 45 • C and SOC 100%. The complementary information obtained from electrochemical, morphological (SEM) and chemical (EDX, GD-OES, and NMR) characterization techniques allowed to identify the main aging mechanisms for each aging condition.
The main degradations are located at the negative electrode and are manifested by the growth of SEI, either by solvent degradation during high temperature calendar aging, or by the growth of SEI by salt degradation during high temperature cycling. The deposition of lithium metal for low temperature cycling leading to the formation of a secondary SEI is identified. In addition, silicon particles suffer significantly during cycling aging at low temperature through particle cracking and disaggregation.
The internal design of the cell influences the aging mechanisms, as high energy cells are more exposed at low temperature to lithium metal deposition due to their thicker electrodes coating which decrease the electrodes' kinetics.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/batteries7030048/s1, Figure S1: Internal resistance evolution of (a) 30Q, (b) 32E and (c) 35E cells during their cycle (left curves) and calendar (right curves) aging. Internal resistance was measured with the voltage drop at the end of a 10 s current pulse at I max ., Figure S2: Nominal Voltage evolution of (a) 30Q, (b) 32E and (c) 35E cells during their cycle (left curves) and calendar (right curves) aging.