Multiple Knee Points in the Degradation of a Commercial Lithium-ion Battery: A Case Study of the NCM/Graphite System

Abstract

Lithium-ion batteries are used on an increasingly large scale, making their lifetime prediction a critical issue. Especially, a rapid decrease in capacity after the mild degradation period, referred to as the knee point, is often observed, and thus understanding the knee point and its mechanism is a critical issue. Although numerous studies have been dedicated to the analysis of this phenomenon, studies on a commercial large-format lithium-ion battery are lacking. Further studies are required to continuously track changes over time and elucidate the source of knee points during operation. Herein, we conduct degradation cycle tests using a large-format commercial lithium-ion battery (> 50 Wh, LiNi_0.5Co_0.2Mn_0.3O₂/graphite) and analyze the knee points during cycling by employing electrochemical impedance spectroscopy (EIS) at different states of charge (SOCs) to track the degradation state over time, in combination with differential analysis and post-mortem methods. As a result, two knee points appear during degradation, caused by increases in resistance that are mainly derived from electrolyte depletion and Li plating at the anode. These observations are described based on the SOC dependency of the EIS results, which can be leveraged to identify the cause of knee points.

1. Introduction

The use of lithium-ion batteries in large-scale applications has grown rapidly, and thus, the importance of predicting their degradation has increased. Lithium-ion batteries are already ubiquitous in our lives, but they are also being used on an increasingly large scale, including transportation and grid applications. In such applications, accurate lifetime predictions are critical for planning maintenance and capital investment.

Despite the increasing demand for lithium-ion batteries, lifetime predictions are still challenging because of the “knee point.” Lithium-ion batteries initially exhibit slow and predictable capacity degradation, following either a square root law^1,2 or linear law.³ However, lithium-ion batteries sometimes undergo a rapid decrease in capacity after the mild degradation period, referred to as the knee point, rollover, or sudden death.⁴ Predicting the appearance of knee points is not simple because the formation mechanisms are diverse and largely depend on the conditions. For example, resistance growth,⁵ Li plating^6,7 including Li plating induced by heterogeneity,⁸ electrolyte depletion,⁹ and mechanical deformation¹⁰ have been reported to cause knee points. In recent years, studies have focused on observing knee points that arise during the operation of small commercial batteries under various conditions,^11–13 including simulation-based studies.^7,14 However, there are few precise analyses of the knee points that appear during the operation of large commercial lithium-ion batteries.^8,9 Although the relationships between knee points and operating conditions have been investigated, it remains necessary to precisely track changes in the battery during degradation, especially in large-format commercial batteries, to fully characterize the knee point phenomenon.

In this study, we performed degradation cycle tests using a commercial large-format lithium-ion battery (> 50 Wh, LiNi_0.5Co_0.2Mn_0.3O₂ [NCM523]/graphite) that exhibits multiple knee points during its life. Electrochemical impedance spectroscopy (EIS) was employed to track the changes in the batteries over time and determine the cause of the knee points. Although EIS is commonly used for investigating the kinetics of batteries,¹⁵ the cause of the knee points can further be identified when multiple states of charge (SOCs) are considered, revealing the changes in kinetics and their SOC dependence. For example, Li plating should exhibit a change in kinetics and a unique SOC dependency because Li plating/stripping tends to proceed at a high SOC and decreases the resistance of anode reactions.¹⁶ However, EIS responses are seldom obtained at multiple SOCs before and after knee points to track the kinetics and SOC dependence. Moreover, a comparison between the EIS responses of the battery and the disassembled electrodes helps identify the cause of the change in EIS response, by attributing the resistive components to the cathode, anode, or electrolyte.

Herein, we tracked the EIS response and their SOC dependence in commercial batteries with multiple knee points throughout their lives. Additionally, we conducted EIS using fresh electrodes and degraded electrodes obtained after cell disassembly. From these EIS responses of the battery and the fresh or degraded electrodes, in combination with differential voltage (dV/dQ) and post-mortem analyses, the causes of the knee points were determined.

2. Experimental

A large-format commercial pouch-type lithium-ion battery (Maxell, Nominal capacity of 15 Ah) consisting of an NCM523 cathode and a graphite anode was cycled, and its degradation behavior was investigated using multiple methods, with a focus on the EIS results.

Two identical cells were used during the degradation cycle test under the same conditions, to confirm reproducibility. The schematic explanation of the flow of the degradation cycle test in combination with capacity checks is displayed in Fig. 1a. Degradation cycle tests were performed using a Charge/discharge Unit (Hokuto Denko, Japan) at 45 °C and 2C (a rate of 1C corresponds to the current that will discharge the full capacity of a battery in one hour), with a SOC range of 9 %–92 % (3.32–4.12 V), over total discharge time of 950 h. Because the objective of this study is to precisely analyze the large-format lithium-ion battery with knee points, the high cycling rate (2C) that has been shown to usually have knee points was adopted.^11,17 Capacity checks were conducted at 25 °C by sequentially using the rates of 0.05C, 0.2C, 0.5C, 1C, and 2C during degradation cycle testing. Three cycles of charge/discharge were conducted at each rate and the capacity value of the third discharge was adopted if not mentioned. Capacity checks were conducted every 61 days for the first 139 days and mostly every 34 days for the period after 140 days. The detailed cycle test periods and the number of cycles before each of the capacity checks in the Cell 1 are shown in Fig. 1b. The differential voltage (dV/dQ) curves were calculated based on the charging voltage curves obtained at 0.05C. To determine the values of the peak positions, each peak was fitted with the Gaussian function or Lorentzian function, as a general peak fit function. After every capacity check, EIS was measured using VSP and VMP-3B-20 instruments (Biologic, France) for the SOCs of 10 % (3.48 V), 50 % (3.68 V), and 90 % (4.02 V), from 20 kHz to 500 µHz, at the amplitude of 0.75 A. Each EIS was carried out after performing a constant-voltage (CV) charge at each defined voltage for 6 h. For the EIS data analysis, fitting calculations were performed using Z-view (Scribner Associates, USA).

Figure 1.

(a) A schematic explanation of the experimental flow of the target battery. (b) The cycle test periods and the number of cycles before each of the capacity checks.

Disassembly of the target batteries, extraction of the electrodes, and EIS using the extracted electrodes were conducted for both the fresh and degraded batteries (i.e., post-mortem analysis), as described below. The fresh battery and degraded battery after testing were disassembled in an argon-filled glove box (Miwa FMG, Japan). The fresh battery was disassembled after complete discharge (SOC0 %). The degraded battery was disassembled after performing a capacity check (25 °C) and 2C charge (until reaching the upper cut-off voltage), to check whether Li metal was present in the anode. For the EIS of the electrodes obtained from the fresh and degraded batteries, symmetric cells adjusted to SOC10 % (3.48 V in battery) were utilized, separating the data of each electrode. Details regarding the fabrication of symmetric cells using fresh or degraded electrodes are provided in the Supporting Information. The symmetric cells contained the disassembled electrodes with diameters of 16 mm, in which the composite electrode materials were removed on one side, as previously described.¹⁸ All symmetric cells were fabricated using a fresh electrolyte, namely 1 M LiPF₆ in ethylene carbonate + dimethyl carbonate (50 : 50 by volume).

In addition to EIS, several other methods were used to investigate the disassembled electrodes. Charge/discharge tests were conducted to evaluate the reversible capacity and dV/dQ of each electrode, and cross-sectional scanning electron microscopy (SEM) and ⁷Li solid-state nuclear magnetic resonance (NMR) were conducted. Half cells using a Li metal disc (19 mm diameter), the same electrodes used in the symmetric cells (16 mm diameter), and the same electrolyte used in the symmetric cells were employed to conduct charge/discharge tests and dV/dQ analyses. To observe changes in the appearance of the electrodes before and after degradation cycling, pieces of the cathodes and anodes from both batteries were removed, and images of cross-sectional scanning electron microscopy (CS-SEM) were acquired using an IM4000 Plus ion milling system (Hitachi High-Tech, Japan) and Quanta 250 scanning electron microscope (FEI™, USA). Note that cross-section polishing and cross-sectional scanning electron microscopy (SEM) observation of the anode were all performed in an inert atmosphere. Then, ⁷Li solid-state NMR of the degraded anode was measured to investigate the presence of Li metal, using an AVANCEIII600 system (Bruker, USA) and Li rotors (4 mm diameter).

All electrochemical measurements of the cells with disassembled electrodes were conducted at 25 °C using a VMP3 system (Biologic, France). The half cells with cathodes and anodes from the fresh and degraded batteries were cycled to measure the reversible capacity at 0.05C. EIS measurements of the symmetric cells using either fresh or degraded electrodes were conducted at SOC10 %, as adjusted using full cells or half cells, from 200 kHz to 500 µHz, at the amplitude of 10 mV.

3. Results and Discussion

3.1 Results of the degradation cycle tests and recognition of the knee points

Figure 2 displays the plots of 2C cycle capacity retention against total discharge time during the degradation cycle test. Under the same condition, the two cells showed similar behavior, confirming the reliability and reproducibility of the devices. In Fig. 2, the sudden acceleration in capacity loss occurred at the capacity retention of approximately 80 % and 40 % (hereafter referred to as knee point 1 and knee point 2, respectively). The appearance of knee points in high-rate cycling was consistent with other sudies.^11,17

Figure 2.

Discharge capacity retention during the degradation cycle tests at 2C. Cell 1 and Cell 2 represent the results of the two identical cells operated under the same conditions to confirm reproducibility.

Furthermore, Fig. S2 shows the capacity at each capacity check during the degradation cycle test against the total discharge time of the degradation cycles. Comparing this result with the capacity retention in Fig. 2, a trend similar to knee point 1 is observed only in the capacity check at high rates. Furthermore, although a gradual but similar trend to knee point 2 is observed at every rate, the trend is more obvious at high rates. This indicates that the appearance of the knee points is related to the resistance of the battery, strengthening the rationale for using EIS as a main method to identify the mechanism of the knee points. Notably, taking a closer look at the capacity check results as shown in Fig. S3, the variation of the capacity values obtained from each of the three cycles at the high rate increased after knee point 2. It is shown that the capacity considerably decreased even during the three cycles of capacity checks at a high rate to cause the variety of the capacity values, implying that the capacity check conducted at a high rate at 25 °C at least partially caused the non-linear capacity fade at knee point 2.

In the following, each mechanism of the knee points is discussed, and the main degradation mode of each region separated by knee points (from the beginning to knee point 1, from knee point 1 to knee point 2, and after knee point 2) is analyzed using the combination of electrochemical and post-mortem methods. Electrochemical methods focus on EIS as well as dV/dQ, measured at each capacity check during the degradation test to track the battery state over time. Post-mortem methods include disassembly of the degraded battery and subsequent electrochemical or non-electrochemical analyses to reveal the state of the degraded state.

3.2 dV/dQ analyses: Estimation of the initial state and quantification of the capacity loss caused by conventional degradation

Before analyzing the EIS results and discussing the change in kinetics during degradation, the thermodynamic (quasi-statically measured) capacity loss was evaluated by performing charge/discharge tests and dV/dQ analyses. Charge/discharge tests are often conducted at a very low rate, such as 0.05C, to quantify the total thermodynamic loss, and dV/dQ analysis is performed using such charge/discharge data to quantify the thermodynamic loss of active materials for each electrode and the thermodynamic loss of active lithium ions by tracking shifts of the peak positions or fitting the characteristic shapes using dV/dQ curves of the anode and cathode as Refs. 19–21. The loss of active materials and loss of active lithium ions are conventional thermodynamic degradation modes observed in the early stages of lithium-ion battery operation.^18,22 In the following, the loss of active material and loss of active lithium ions are collectively called “conventional degradation modes.” Moreover, the capacity losses measured at a very low rate are considered “thermodynamic” because they are obtained under almost static conditions without the influence of resistance causing energy loss,²³ which should be separately evaluated from the change in the kinetics.

Figures S4 and S5 show the dV/dQ curves obtained by charging (a capacity check) before the degradation cycle test and during the degradation cycle test, respectively. To interpret the obtained dV/dQ data of the battery and separate the contributions of the cathode and anode, the dV/dQ curves of fresh and degraded cathodes and anodes were obtained separately using half cells of the disassembled electrodes, as shown in Fig. S4. Combining the dV/dQ curves of the battery and the half cells, the initial shift of the operation window was obtained in the fresh battery, and the change in the shift due to degradation was tracked in the degradation cycle test. Based on the dV/dQ of the half cells consisting of disassembled degraded electrodes (shown in Fig. S7), the anode in this battery exhibited little capacity loss, whereas the cathode exhibited approximately 10 % capacity loss, leading to the loss of active materials only in the cathode. Moreover, the dV/dQ curve of the cathode has only one peak, which makes it difficult to distinguish the contributions between the loss of cathode active material and loss of active lithium ions; thus, their contributions to the capacity loss were combined and evaluated as a total value in this study. In this case, the conventional capacity loss refers to the total loss of the cathode active material and the active lithium ions. The detailed procedure used to analyze dV/dQ curves is presented in the Supporting Information. Before discussing the capacity losses, we note that the anode/cathode capacity ratio (N/P ratio) in the fresh battery was calculated to be 1.30 based on the results displayed in Fig. S6, from dV/dQ analyses of the fresh battery and fresh electrodes. Considering that the typical N/P ratio in lithium-ion batteries is 1.1–1.2,^24,25 our target battery had a high N/P ratio, which is thought to be a safe design against Li metal deposition.

The separation of the total thermodynamic capacity loss and the capacity loss obtained by conventional degradation from the total capacity loss that occurred during the degradation cycles at 2C is shown in Fig. 3. These losses are described as a function of the total discharge time and displayed on a scale of capacity retention. As a result, the total thermodynamic capacity loss (i.e., capacity loss measured at 0.05C) almost completely coincides with the loss caused by conventional degradation modes, demonstrating that the reduction in thermodynamic capacity is mainly due to the total loss of cathode active material and active lithium ions. This indicates that the other factors contributing to the loss of thermodynamic capacity such as the change of potential curves of the cathode or anode due to the increase of the surface area of active materials²⁶ or due to the heterogeneity of the electrodes^8,27 are limited. Furthermore, the thermodynamic capacity loss increases almost at the same rate as the total discharge time, regardless of the presence of knee points. Therefore, the appearance of knee points was attributed to the change in kinetics, which is the other factor of capacity loss.

Figure 3.

The capacity retention of degradation cycles performed at 2C, the capacity measured at 0.05C (thermodynamic capacity), and the capacity loss derived from conventional degradation modes (cathode active material loss + active lithium-ion loss), and the capacity loss by other than conventional degradation modes, as a function of the total discharge time. For the calculation of the capacity loss by other than conventional degradation modes, obtained by subtracting the capacity loss by conventional degradation modes from the overall capacity loss of degradation cycles, the capacity loss by conventional degradation modes obtained by dV/dQ analyses was linearly interpolated to have the values between capacity checks.

3.3 EIS analyses: Quantification and tracking of the resistance increase and influence of resistive components

The resistance and resistive components were quantified by EIS analyses to investigate the change in the kinetics of the battery during degradation, which was shown to cause the appearance of knee points. EIS was carried out in the capacity checks performed before and during degradation cycle tests to clarify the change in resistance.

First, the Nyquist plots of the fresh battery at SOC10 %, SOC50 %, and SOC90 % were obtained, as shown in Fig. 4a. The enlarged view without the inductive effect of the cable is shown in Fig. 4b; the details of the cable impedance are provided later. For every SOC, two semicircles and a subsequent tail were observed in the Nyquist plots. Because SOC10 % had the best arc separability, the interpretation of the resistive components was discussed by comparing the EIS results of the battery (SOC10 %) to the EIS results of the symmetric cells fabricated by disassembled fresh electrodes. The Nyquist plots of the symmetric cells are displayed in Figs. 4c and 4d, showing one semicircle and a subsequent tail for both the anode and the cathode. The top frequency of the semicircle in each symmetric cell was approximately 200 and 3 Hz for the anode and cathode, respectively. These semicircles are attributed to charge-transfer reactions at the electrodes, according to the frequencies reported in the literature.^28,29 Because the top frequencies of the semicircles observed for the symmetric cells are approximately similar to those of the battery, the first semicircle from the high frequency in the battery Nyquist plot is attributed to anode charge transfer, and the second semicircle from the high frequency in the battery Nyquist plot is attributed to cathode charge transfer. The subsequent tail of lower frequency is attributed to the sum of Li diffusion in solid active materials and Li-ion diffusion inside pores of electrodes soaked with electrolyte.^30,31 The frequency of the tail part is almost equal for the anode and cathode, and thus the EIS behavior contains both contributions. To separate these different components in the EIS responses and quantify them, the EIS results were analyzed in the manner shown in Fig. 4e. Namely, in the high to medium frequency region, semicircles were fitted using the equivalent circuit shown in Fig. 4e, and then the resistance corresponding to diffusion (tail part) was obtained by subtracting the resistance of the semicircles from the total resistance. This total resistance was defined as the real part of the impedance at 500 µHz, and only its variation was discussed. Note that the EIS response of the cable was included in the equivalent circuit.³² Hereafter, R_sol, R_anode, R_cathode, R_dif, and R_all represent the bulk electrolyte solution resistance, charge transfer of the anode, charge transfer of the cathode, diffusion resistance, and total resistance, respectively. The EIS results of the battery at SOC50 % and SOC90 % were analyzed in the manner established for SOC10 %, as depicted in Fig. 4e, and the resulting quantities of the resistive components were obtained.

Figure 4.

(a) Nyquist plot of the fresh battery measured at SOC10 %, SOC50 %, and SOC90 %. (b) Enlarged view of the Nyquist plots displayed in (a), including the approximate top frequency values of the semicircles. (c) Nyquist plot of the disassembled fresh anode at SOC10 %, obtained by halving the EIS result of a symmetric cell. (d) Nyquist plot of the disassembled fresh cathode at SOC10 %, obtained by halving the EIS result of a symmetric cell. (e) Interpretation of the battery EIS results, including the equivalent circuit corresponding to the semicircles appearing in the high- to mid-frequency regions.

Then, using the results of the above analysis, we investigated the trend of increased resistance during battery degradation. To begin with, the total resistance R_all of the battery against total discharge time was obtained from the EIS data at 500 µHz, as displayed in Fig. 5. Figure 5a shows the average increase in the total resistance R_all at the SOCs of 10 %, 50 %, and 90 %, along with the change in capacity loss caused by factors other than conventional degradation (same values as Fig. 3). Overall, the two trends of R_all and capacity loss matched well. Therefore, knee point 1 and knee point 2 were derived from the phenomena that caused the increase in resistance. Notably, the capacity loss by other than conventional degradation modes seems not to have grown between the capacity checks after knee point 2, the reason for which is to be suggested later in Section 3.5.

Figure 5.

(a) Comparison of the trends in total resistance increase for the battery and capacity loss by factors other than conventional degradation (loss of active materials and active lithium ions), as a function of total discharge time. The total resistance increase is the average of the increased value of total resistance R_all obtained at SOC10 %, SOC50 %, and SOC90 %. (b) Increase in total resistance R_all as a function of the total discharge time for each SOC, obtained by EIS.

To further discuss the cause of the knee points, the increase in R_all measured at each SOC during the degradation test is shown in Fig. 5b. Although both knee point 1 and knee point 2 were derived from the resistance increase, the increasing trend of the total resistance R_all between knee point 1 and knee point 2 was almost independent of SOCs, whereas a significant SOC dependency emerged after knee point 2. After knee point 2, the resistance increase for SOC90 % is much smaller than the sharp increases observed for SOC10 % and SOC50 %. Then, the increase in each resistive component at each SOC during the degradation cycle test was investigated, as shown in Fig. 6. In the region between knee point 1 and knee point 2, the values of the resistive components R_sol, R_anode, and R_cathode increased by up to 2 mΩ, showing gradual increasing trends, whereas R_dif increased by approximately 4 mΩ, showing a rapid increase, at any SOC. Thus, the increasing trend and the increased values of every resistive component are similar regardless of SOC in the range between knee point 1 and knee point 2, and knee point 1 likely emerged because of the relatively large increase in R_dif. Moreover, after knee point 2, only the diffusion resistance R_dif exhibited a large SOC dependence and rapid growth. This indicates that knee point 2 also emerged because of the large increase in R_dif, but it was dependent on the SOC. As previously mentioned, the diffusion resistance R_dif contains contributions from both the Li diffusion in solid active materials and Li-ion diffusion inside the pores of electrodes soaked with electrolytes. Knee point 1 and knee point 2 are triggered by the hindrance of Li diffusion either in the solid active material or the electrolyte inside the pores. We discuss the difference in SOC dependence between the two knee points and the direct cause of the rapid growth in the diffusion resistance R_dif later in Section 3.5.

Figure 6.

The resistance increase of each resistive component for each SOC, as a function of the total discharge time.

3.4 Post-mortem analyses: Battery disassembly and investigation of the disassembled electrode

To determine the direct cause of the increase in R_dif leading to knee points, the battery (fresh and degraded) was disassembled. As Fuhrmann et al. mentioned,⁸ gray powder was observed on the anode surface after disassembly, implying the deposition of Li metal-related materials. Herein, the disassembled electrodes were analyzed by CS-SEM, ⁷Li solid-state NMR, EIS using symmetric cells operated at SOC10 %, and charge/discharge testing in half cells. The charge/discharge results in half cells are discussed above (Fig. S7), but the other results are described in this section.

The cross-sectional SEM images obtained before and after degradation cycling are shown in Fig. 7. For the cross-sectional views of the cathodes in Figs. 7a and 7b, severe particle cracks are observed. In this degradation cycle test, the charge/discharge rate appeared to be sufficiently high to cause cracks.³³ Particle cracks are one of the causes of capacity loss, resulting from the partial isolation of some active materials. Fractures also generate fresh surfaces in the active materials and may accelerate the consumption of electrolyte.³³ Figures 7c and 7e displays the cross-sectional views of the anodes, showing light-colored structures, as opposed to the dark graphite particles, on the anode surface of the separator side and inside the pores between graphite particles in the degraded anode (see Figs. 7d and 7f for enlarged views). Notably, the light-colored structures inside the pores between graphite particles are also observed in the fresh anode. The elemental composition was examined by energy-dispersive X-ray spectroscopy (EDX), as shown in Fig. S9. Accordingly, the light-colored structures contain much more oxygen and fluorine and far less carbon than the graphite structures. Although EDX cannot detect Li, Li metal can easily react with oxygen and fluorine to form various compounds such as LiF and Li₂O and so on, suggesting that these light-colored structures contain Li. Furthermore, Fig. S10 shows secondary electron and backscattered electron images of the degraded anode, including the light-colored structures. In backscattered electron images, contrast depends on the composition, and a higher average value of the atomic number of the components gives a brighter contrast. In Fig. S10, the light-colored structures inside the pores were brighter than graphite, whereas the inner part of the light-colored structures on the anode surface was darker. Therefore, although the light-colored structures inside the pores are mainly attributed to the lithium salts derived from electrolytes that originally contained more oxygen and fluorine, considering that they are also seen in the fresh anode, the light-colored structures on the separator side of the anode surface appeared to be attributed to Li metal and its compounds derived from Li plating.

Figure 7.

CS-SEM views of the disassembled electrodes obtained from the fresh and degraded batteries. All views are secondary electron images, obtained using acceleration voltages of 20 and 5 kV for the cathodes and anodes, respectively. (a) Cathode of the fresh battery. (b) Cathode of the degraded battery. (c) Anode of the fresh battery. (d) Enlarged view of the anode from the fresh battery (a central part of (c), dotted lines correspond to those in (c)). (e) Anode of the degraded battery. (f) Enlarged view of the anode from the degraded battery (a central part of (e), dotted lines correspond to those in (e)).

Based on the SEM-EDX results, ⁷Li solid-state NMR of the degraded anode was conducted to distinguish the chemical states of lithium. The resulting spectrum is shown in Fig. 8, and a characteristic peak of Li metal was confirmed at the chemical shift of 264 ppm.³⁴ Therefore, the light-colored structures contain Li metal and its reacted compounds, and Li metal deposition occurred during the degradation test.

Figure 8.

⁷Li solid-state NMR spectrum of the degraded anode.

The EIS results of the symmetric cells using degraded electrodes obtained at SOC10 % are shown in Fig. 9. Compared with the symmetric cells using fresh electrodes (Figs. 4c and 4d), the total resistance R_all of the degraded cathode was almost unchanged. Considering the dV/dQ results showing that the thermodynamic capacity loss was mainly attributed to the loss of cathode active materials and active lithium ions, the observed particle fracture did not directly affect the resistance increase, whereas it may have affected the thermodynamic capacity loss. Thus, the cathode behavior does not fully explain the increase in the diffusion resistance R_dif of the battery after knee point 1. In Fig. 9b, the EIS response of the degraded anode was smaller than that of the fresh anode, which also does not fully explain the rapid increase in the diffusion resistance R_dif after knee point 1. However, the EIS results of the disassembled electrodes can be accounted for by considering the refreshed electrolytes in the reassembly of the symmetric cells. Namely, the rapid increase of the diffusion resistance R_dif may be caused by the depletion of the electrolyte, which hinders Li-ion diffusion inside the electrode pores. In this study, the term “depletion” includes both the drying up of the electrolyte and the decreasing of “healthy” electrolytes not denatured by the reaction at the electrodes (i.e., dilution of the electrolyte).⁹ The increased resistance in Li-ion transfer in the electrolyte can cause polarization at the anode, which gives rise to Li metal plating. Moreover, although the degraded anodes showing a smaller EIS response than the fresh ones need further reason and investigation as future work, it can be described based on the report that graphite anode can reduce its resistance after cycling by enlarged d-spacing and larger surface area.³⁵ While the reduced resistance of the anode may not be reflected in the overall battery EIS response during the degradation cycle test due to the depletion of electrolytes, the influence of such changes in graphite might be observed as a smaller EIS response in the symmetric cell with the refreshed electrolytes. More details of the overall degradation are described in the next section.

Figure 9.

Nyquist plots of the EIS results of the fresh and degraded cathodes and anodes, obtained by halving the EIS results of the symmetric cell at SOC10 %. Two samples were evaluated for each condition to confirm reproducibility. (a) Nyquist plots of the EIS results obtained for the cathodes. (b) Nyquist plots of the EIS results obtained for the anodes.

3.5 Degradation scenario

One explanation for the knee points is that electrolyte depletion increased the diffusion resistance R_dif, leading to Li plating at the anode. This scenario is illustrated in Fig. 10. Before knee point 1, conventional degradation occurred (i.e., loss of cathode active materials and active lithium ions). Some cathode active materials may have been lost because of particle cracking, as observed by SEM. The loss of active lithium ions may have been caused by side reactions and lithium-ion consumption at the cathode and anode. The electrolyte is gradually consumed, along with the loss of active lithium ions,³⁶ thereby increasing the resistance. As a result, capacity retention gradually and constantly decreased, similar to the “standard” case of degradation.¹⁸ Then, after knee point 1, the electrolyte started to deplete in the pores of the electrodes in addition to the degradation modes before knee point 1. The electrolyte consumption should be faster in the pores than in the bulk electrolyte because of the high surface area. This caused a rapid increase in the diffusion resistance R_dif (independent of the SOC) because the diffusional transfer of lithium ions inside the electrode composite layer was hindered, which can occur at any SOC. Note that this mechanism of electrolyte depletion in the pores can occur in both the anode and the cathode. After that, the polarization at the anode continuously increased to lower the anode reaction potential until it became comparable to that of Li plating, and then Li plating was initiated at knee point 2. The Li plating/stripping reaction occurred at higher SOCs, whereas the graphite intercalation reaction continued at moderate SOCs and below. This increased the diffusion resistance R_dif in a SOC-dependent manner because Li ions must be transferred into the inner part of the electrode for graphite intercalation, whereas Li plating/stripping can occur without Li-ion transfer inside the composite electrode, shortening the diffusion length in the high-resistance electrolyte and suppressing the resistance increase. Considering that the capacity substantially decreased even during the three cycles of capacity check at 2C after knee point 2, Li plating might have been initiated in high-rate cycles of the capacity check because the temperature for the capacity checks (25 °C) was low enough to increase the resistance and polarization of the anode to start Li plating at the state of knee point 2. Furthermore, once Li plating started, the rate of electrolyte consumption rapidly increased owing to film formation on the Li surface. Thus, after knee point 2, the diffusion resistance R_dif significantly increased for low-to-medium SOCs and gradually increased for high SOCs. It should be noted here that the capacity loss by other than conventional degradation modes in Figs. 3 and 5 corresponding to resistance increase did not grow between the capacity checks after knee point 2. We speculated that this is because the behavior of the deposition/stripping reaction of Li metal is dependent on operating temperatures and thus the efficiency of the reaction gets better and less inactive “dead Li” is generated in high-temperature conditions.³⁷ Therefore, high-temperature degradation cycles are thought to have smaller resistance caused by dead Li and reactive products compared with the experiments at 25 °C. Finally, the diffusion resistance R_dif measured for the symmetric cells using the degraded anode and cathode was much lower than that observed in the degraded battery; this is because the electrolyte was refreshed during reassembly, thereby recovering the lower resistance of the electrolyte.

Figure 10.

Illustration of the estimated degradation in the tested battery and the change in degradation modes before and after the knee points.

4. Conclusion

Degradation cycle testing was performed using a large-format commercial lithium-ion battery (> 50 Wh, NCM523/graphite) to determine the knee point formation mechanism during battery operation. EIS was conducted at multiple SOCs, as well as dV/dQ analysis, to track the degradation state during cycling, in addition to using post-mortem methods. As a result, two knee points were observed, and different trends were revealed for the resistance increase after each knee point. A SOC-independent increase in the diffusion resistance occurred after the first knee point, whereas a SOC-dependent increase in the diffusion resistance occurred after the second knee point. These observations were explained by gradual electrolyte depletion starting at the first knee point and Li plating/stripping at the anode after the second knee point.

Acknowledgments

We thank Seiki Komiya (Central Research Institute of Electric Power Industry) for fruitful discussions.

The authors were waived from the APC with the support of The Committee of Battery Technology, ECSJ.

Data Availability Statement

The data that support the findings of this study are openly available under the terms of the designated Creative Commons License in J-STAGE Data at https://doi.org/10.50892/data.electrochemistry.28491317.

• 1) Lithium-ion batteries are used on an increasingly large scale, making their lifetime prediction a critical issue. Especially, a rapid decrease in capacity after the mild degradation period, referred to as the knee point, is often observed, and thus understanding the knee point and its mechanism is a critical issue. Although numerous studies have been dedicated to the analysis of this phenomenon, studies on a commercial large-format lithium-ion battery are lacking. Further studies are required to continuously track changes over time and elucidate the source of knee points during operation. Herein, we conduct degradation cycle tests using a large-format commercial lithium-ion battery (> 50 Wh, LiNi_0.5Co_0.2Mn_0.3O₂/graphite) and analyze the knee points during cycling by employing electrochemical impedance spectroscopy (EIS) at different states of charge (SOCs) to track the degradation state over time, in combination with differential analysis and post-mortem methods. As a result, two knee points appear during degradation, caused by increases in resistance that are mainly derived from electrolyte depletion and Li plating at the anode. These observations are described based on the SOC dependency of the EIS results, which can be leveraged to identify the cause of knee points.

CRediT Authorship Contribution Statement

Yui Fujihara: Conceptualization (Lead), Data curation (Lead), Formal analysis (Lead), Investigation (Lead), Writing – original draft (Lead)

Takeshi Kobayashi: Conceptualization (Equal), Funding acquisition (Lead), Writing – review & editing (Supporting)

Conflict of Interest

The authors declare no conflict of interest in the manuscript.

Footnotes

This paper is based on the content of the presentation given at the 65th Battery Symposium in Japan, November 20–22 in Kyoto (No. 1MH21).

Y. Fujihara and T. Kobayashi: ECSJ Active Members

References

• 1)  H. Yoshida, N. Imamura, T. Inoue, and K. Komada, Electrochemistry, 71, 1018 (2003).
• 2)  M. Broussely, S. Herreyre, P. Biensan, P. Kasztejna, K. Nechev, and R. Staniewicz, J. Power Sources, 97–98, 13 (2001).
• 3)  S. Atalay, M. Sheikh, A. Mariani, Y. Merla, E. Bower, and W. D. Widanage, J. Power Sources, 478, 229026 (2020).
• 4)  P. M. Attia, A. Bills, F. Brosa Planella, P. Dechent, G. dos Reis, M. Dubarry, P. Gasper, R. Gilchrist, S. Greenbank, D. Howey, O. Liu, E. Khoo, Y. Preger, A. Soni, S. Sripad, A. G. Stefanopoulou, and V. Sulzer, J. Electrochem. Soc., 169, 060517 (2022).
• 5)  A. R. Mandli, A. Kaushik, R. S. Patil, A. Naha, K. S. Hariharan, S. M. Kolake, S. Han, and W. Choi, Int. J. Energy Res., 43, 2044 (2019).
• 6)  D. Anseán, M. Dubarry, A. Devie, B. Y. Liaw, V. M. García, J. C. Viera, and M. González, J. Power Sources, 356, 36 (2017).
• 7)  X. Yang, Y. Leng, G. Zhang, S. Ge, and C. Wang, J. Power Sources, 360, 28 (2017).
• 8)  M. Fuhrmann, L. Torcheux, and Y. Kobayashi, J. Power Sources, 621, 235210 (2024).
• 9)  L. Hartmann, L. Reuter, L. Wallisch, A. Beiersdorfer, A. Adam, D. Goldbach, T. Teufl, P. Lamp, H. A. Gasteiger, and J. Wandt, J. Electrochem. Soc., 171, 060506 (2024).
• 10)  A. Pfrang, A. Kersys, A. Kriston, D. U. Sauer, C. Rahe, S. Käbitz, and E. Figgemeier, J. Power Sources, 392, 168 (2018).
• 11)  X. Ma, J. E. Harlow, J. Li, L. Ma, D. S. Hall, S. Buteau, M. Genovese, M. Cormier, and J. R. Dahn, J. Electrochem. Soc., 166, A711 (2019).
• 12)  Y. Preger, H. M. Barkholtz, A. Fresquez, D. L. Campbell, B. W. Juba, J. Romàn-Kustas, S. R. Ferreira, and B. Chalamala, J. Electrochem. Soc., 167, 120532 (2020).
• 13)  T. C. Bach, S. F. Schuster, E. Fleder, J. Müller, M. J. Brand, H. Lorrmann, A. Jossen, and G. Sextl, J. Energy Storage, 5, 212 (2016).
• 14)  J. Keil and A. Jossen, J. Electrochem. Soc., 167, 110535 (2020).
• 15)  J. Zhu, M. S. Dewi Darma, M. Knapp, D. R. Sørensen, M. Heere, Q. Fang, X. Wang, H. Dai, L. Mereacre, A. Senyshyn, X. Wei, and H. Ehrenberg, J. Power Sources, 448, 227575 (2020).
• 16)  M. Koseoglou, E. Tsioumas, D. Ferentinou, N. Jabbour, D. Papagiannis, and C. Mademlis, J. Power Sources, 512, 230508 (2021).
• 17)  P. Keil, S. Simon, C. Lüders, H. Hesse, R. Arunachala, and A. Jossen, 3rd Electromobility Challenging Issues Conference (2015).
• 18)  Y. Kobayashi, T. Kobayashi, K. Shono, Y. Ohno, Y. Mita, and H. Miyashiro, J. Electrochem. Soc., 160, A1181 (2013).
• 19)  I. Bloom, A. N. Jansen, D. P. Abraham, J. Knuth, S. A. Jones, V. S. Battaglia, and G. L. Henriksen, J. Power Sources, 139, 295 (2005).
• 20)  K. Honkura, H. Honbo, Y. Koishikawa, and T. Horiba, ECS Trans., 13, 61 (2008).
• 21)  H. M. Dahn, A. J. Smith, J. C. Burns, D. A. Stevens, and J. R. Dahn, J. Electrochem. Soc., 159, A1405 (2012).
• 22)  M. Klett, J. A. Gilbert, S. E. Trask, B. J. Polzin, A. N. Jansen, D. W. Dees, and D. P. Abraham, J. Electrochem. Soc., 163, A875 (2016).
• 23)  Y. Hu, J. Liang, X. Chen, G. Chen, Y. Peng, S. Tang, Z. He, D. Li, Z. Zhang, Z. Gong, Y. Wei, and Y. Yang, J. Power Sources, 606, 234590 (2024).
• 24)  M. Weiss, R. Ruess, J. Kasnatscheew, Y. Levartovsky, N. R. Levy, P. Minnmann, L. Stolz, T. Waldmann, M. Wohlfahrt-Mehrens, D. Aurbach, M. Winter, Y. Ein-Eli, and J. Janek, Adv. Energy Mater., 11, 2101126 (2021).
• 25)  J. Teng, X. Tang, M. Tang, Q. Wu, and J. Li, Energy Storage, 51, 104588 (2022).
• 26)  K. Betsuyaku and T. Kobayashi, J. Power Sources, 565, 232893 (2023).
• 27)  Y. Xie, S. Wang, R. Li, D. Ren, M. Yi, C. Xu, X. Han, L. Lu, B. Friess, G. Offer, and M. Ouyang, J. Power Sources, 542, 231753 (2022).
• 28)  M. Umeda, K. Dokko, Y. Fujita, M. Mohamedi, I. Uchida, and J. R. Selman, Electrochim. Acta, 47, 885 (2001).
• 29)  W. Kim, D. Jang, and H. Kim, J. Power Sources, 510, 230338 (2021).
• 30)  J. Huang and J. Zhang, J. Electrochem. Soc., 163, A1983 (2016).
• 31)  J. Moškon, J. Žuntar, S. Drvarič Talian, R. Dominko, and M. Gaberšček, J. Electrochem. Soc., 167, 140539 (2020).
• 32)  T. Osaka, T. Momma, D. Mukoyama, and H. Nara, J. Power Sources, 205, 483 (2012).
• 33)  P. Li, Y. Zhao, Y. Shen, and S.-H. Bo, J. Phys. Energy, 2, 022002 (2020).
• 34)  J. Conard and P. Lauginie, TANSO, 2000, 62 (2000).
• 35)  J. Guo, Y. Li, J. Meng, K. Pedersen, L. Gurevich, and D. Stroe, J. Energy Chem., 74, 34 (2022).
• 36)  P. Arora, R. E. White, and M. Doyle, J. Electrochem. Soc., 145, 3647 (1998).
• 37)  M. Tao, X. Chen, H. Lin, Y. Jin, P. Shan, D. Zhao, M. Gao, Z. Liang, and Y. Yang, ACS Nano, 17, 24104 (2023).