Rate-Dependent Aging Resulting from Fast Charging of Li-Ion Cells

Electrochemical studies were performed using Li_1.03(Ni_0.5Co_0.2Mn_0.3)_0.97O₂ (NCM523) cathodes and graphite (Gr) anodes. The positive electrode contained 90 wt% active oxide (Toda), 5 wt% conductive additive (C45, Timcal) and 5% poly(vinylidene difluoride) binder (PVdF, Solvay 5130), coated on a 20-μm-thick battery-grade Al foil. The cathode was coated to a total loading of 18.6 mg cm⁻² and calendered to a final thickness of 71 μm (35.4% porosity). The negative electrode contained 91.8 wt% graphite (Superior Graphite SLC1506T), 2 wt% C45 and 6 wt% PVdF binder (KF-9300, Kureha); 0.2 wt% oxalic acid was added to the slurry to improve adhesion to the 10-μm-thick battery-grade Cu current collector. The total anode loading was 9.9 mg cm⁻², corresponding to a calendered thickness of 70 μm at 34.5% porosity. Full-cells assembled with these electrodes have a ratio of negative and positive electrode capacities (N/P ratio) of ∼1.3 at 4.1 V. Electrodes were fabricated at the Cell Analysis, Modeling and Prototyping (CAMP) Facility at Argonne National Laboratory (ANL). Electrodes were cut into 1.58 cm² discs and dried overnight at 120 ^oC under dynamic vacuum before use. The electrolyte was a 1.2 mol l⁻¹ solution of LiPF₆ in 3:7 w/w ethylene carbonate (EC) and ethyl methyl carbonate (EMC), procured from Tomiyama. All tests were carried out in 2032-type cells, using 2320 Celgard trilayer separator and 40 μl of electrolyte (4 times the total pore volume in the cell).

NCM523//Gr cells were cycled at 30 ^oC in a Maccor 2300 battery tester. To begin, all cells underwent formation cycles consisting of two cycles at C/10 and one at C/25 (3.0–4.1 V). Subsequently, testing consisted of a series of high-rate pulses, capacity-limited constant-current charges, and slow diagnostic cycles (Fig. 1).

Zoom In Zoom Out Reset image size

The cycles containing the constant-current charge are referred to as aging cycles, while the pulse-current and slow cycles were used to determine cell impedance rise and capacity fade, respectively. Cell discharge before and after the constant-current charge was performed at C/5, with a hold at 3.0 V until the current decreased to C/100. For every constant-current charge, tests included a range of terminal SOCs, as indicated in Fig. 2a. Note that these terminal SOCs are defined in terms of initial capacity rather than cell voltage, leading to accelerated aging: Li⁺ loss and impedance rise will expose the cathode to higher potentials in subsequent cycles to meet the capacity target (Fig. 2b), magnifying aging mechanisms.

Zoom In Zoom Out Reset image size

Pulses and diagnostic cycles are performed every two and every twenty charges, respectively. Impedance was measured from voltage changes following the application of 10 s 3C pulses at 3.65 V and 3.8 V; only the data collected during discharge pulses at 3.8 V are shown here, as similar trends were observed at 3.65 V and during the charge pulses. Diagnostic cycles were performed in the 3.0–4.1 V range at a C/25 rate.

This series of current-pulse cycles, fast-charge cycles, and slow cycles was repeated until one of the following conditions were met: (i) the area-specific impedance (ASI) of the full-cell reached 40 Ω cm² (this is the ASI of NCM523//Gr cells after 400 h of potentiostatic hold at 4.39 V,²⁶ and corresponds to an average ASI increase of ∼75%; (ii) cell voltage exceeded 4.8 V during the constant-current charge to meet the target capacity-limit; or (iii) 100 partial cycles were completed. Partial cycles correspond to the number of aging cycles applied to the cell, as opposed to equivalent full cycles, which are counted in terms of the number of 100% ΔSOC swings that are cumulatively completed by the cells.²⁷ For simplicity, cycle numbers (partial and equivalent) discussed in this manuscript refer to the number of capacity-limited cycles endured by the cell, and do not include the current-pulse and diagnostic steps. 100% SOC was defined as 3.9 mAh (2.47 mAh cm⁻²), which is the typical initial C/25 cell capacity, and 1C was assumed to be 3.11 mAh (1.97 mAh cm⁻²); at this current, a fully-charged full-cell will reach the end-of-discharge in 1 h. Cells will be referred to by their test conditions, as in 4C/60 to indicate the cell charged at 4C to 60% SOC during its aging cycle.

Selected cells were disassembled at the end of testing and the "harvested" electrodes were used as-extracted to build half-cells; these electrodes were cycled at C/100 rate to quantify permanent losses of electrode capacity. Cathode and anode half-cells were cycled in the 3.0–4.1 V, and 0–1.5 V ranges, respectively. Cells containing the harvested anodes were delithiated, and those containing the harvested cathodes were relithiated, prior to this C/100 cycle.

Experiments with 3-electrode cells used a custom setup that is described in detail elsewhere.^26,28 In brief, the cell contained a 20.3 cm² cathode and anode, spaced by two layers of Celgard 2320 separator. A thin copper wire (25 μm) with a ∼2 mm exposed tip was positioned in between the separators. A Li micro-reference electrode was formed in situ by extracting Li⁺ from the cathode and plating it on the Cu wire; the charge transferred during this lithiation process is negligible compared to the total capacity in the cathode.

Figure 2a represents the terminal SOCs studied at each constant-current charging rate. At the highest rates (8C, 6C and 4C), the maximum state-of-charge to which cells can be safely exposed is limited by the likelihood of Li plating. At high enough currents, there is a threshold SOC above which the anode potentials become negative (vs Li/Li⁺) and deposition of Li metal starts to compete with Li⁺ intercalation into graphite; the higher the SOC, the more favored the plating reaction is Ref. 28. The color code of Fig. 2a indicates the likelihood of lithium plating according to 3-electrode tests using a Li/Cu micro-reference electrode; red indicates that plating is likely to occur at early cycles, blue that it could become possible after aging, and gray that it is unlikely in the duration of our tests given the initial electrode potentials. These testing conditions are expected to establish clear limits for safe cell operation at high rates, and to reveal trends about how fast charging causes cells to age. A peculiarity of our tests is that cell voltage was allowed to float freely to meet the charge capacity targets. Consequently, side reactions leading to the consumption of Li⁺ inventory will increase the effective charge cutoff voltage in the following capacity-limited cycles, as formerly untapped Li⁺ is extracted from the cathode lattice; increased polarization due to ASI rise would cause similar effects. This is clearly seen in Fig. 2b, which shows the C/5 discharge capacities (black) and the charge cutoff voltages (orange) following successive 1C charges to 60% SOC. Cell aging forces the charge cutoff voltage to increase from 3.78 V to 3.83 V to drive a same amount of Li⁺ to the anode. Thus, in our experiments both aging mechanisms—capacity fade and impedance rise—are expected to expose the cathode to conditions that are progressively more aggressive, making aging effects visible after a relatively small number of cycles.

Given the variety of ways through which fast charging can cause aging, test termination conditions varied widely among the cells (Fig. 2c). Cells tested at moderate rates (1C and 2C) and at lower SOCs were able to complete 100 partial cycles; the ones exposed to the highest SOCs at rates ≥ 4C lost Li⁺ quickly due to plating, and were terminated due to excessive cell voltage polarization during charge; finally, the 6C/30 and 4C/60 cells were interrupted due to exhibiting large impedance rise.

Cell capacities measured by the diagnostic cycles (C/25, 3.0–4.1 V) are shown in Fig. 3. In most cases capacity was lost quasi-linearly throughout the test, while certain systems operating at elevated rates and SOCs (8C/30, 6C/40, 4C/80 and 4C/60) transitioned into a precipitous failure. These qualitative traits have been universally reported during aging of Li-ion batteries containing various electrodes and tested at various conditions.^{17,22,23,29–31} The initial region of quasi-linear fade is generally considered to be dominated by Li⁺ inventory losses to graphite's SEI, and also bears the contribution from a slow and steady impedance rise in the cells.^23,31 The sudden "roll-over" failure, on the other hand, has been associated with high levels of impedance rise (Ref. 32) and/or the initiation of additional mechanisms of capacity fade, such as Li plating.^17,19 The possible occurrence of Li deposition in cells tested at the more extreme conditions was anticipated in our experiment design (Fig. 2a), and confirmed by visual inspection of electrodes at the end of the test (Fig. 3f). Although the electrodeposition of lithium during fast charging is a partially reversible phenomenon,²³ capacity fade can occur due to the electronic isolation of Li domains (forming "dead" lithium) and due to chemical reduction of electrolyte at the surface of Li nuclei.³³ As will be discussed below, cells that entered lithium plating conditions exhibited the highest levels of impedance rise, suggesting that the latter could also be contributing to the capacity fade measured by the diagnostic cycles.

Zoom In Zoom Out Reset image size

Interestingly, cells that only exhibited a quasi-linear capacity fade presented similar levels of Li⁺ loss at the end of the test. This lost capacity was also comparable to that measured for a baseline cell cycled at 0.5C to 100% SOC (red lines in Fig. 3), suggesting that, in these cases, the dominant fade mechanisms were nearly independent of the depth of charge. Clear exceptions to this general behavior were cells tested at 1C/100 and 2C/100, especially the latter. We note that these cells were at the threshold of lithium plating conditions at the beginning of the test (Fig. 2a), and that some Li deposition may have contributed to the observed capacity loss. Nevertheless, unequivocal signs of plating were not present in anodes harvested from these cells.

Trends observed for impedance rise are exemplified by Figs. 4a, 4b. In general, a quasi-linear increase in ASI as a function of partial cycles was observed when charging was constrained to low SOCs. Some of the cells charged to high SOCs, however, eventually deviated from this linearity, showing a drastic increase in ASI. Systems that exhibited this latter behavior also displayed accelerated capacity fade (Fig. 3), suggesting that this rapid ASI rise could be correlated with Li plating. In our accelerated aging tests, Li⁺ loss leads to higher cell voltages (and cathode potentials) being accessed in future cycles, as shown in Fig. 2b. With that in mind, the voltage increase of cells undergoing Li plating can be very high, and this sudden spike in cathode potentials can contribute to the observed impedance rise. This becomes clearer in Fig. 5, which contrasts the terminal full-cell voltage during capacity-limited charges (black) with the ASI rise (orange). Figure 5 only shows cells that experienced voltages above 4.3 V during the test, as these were the ones the presented a non-linear impedance variation; in fact, deviations from linearity tended to appear once cells crossed the 4.3 V threshold. The rapid Li⁺ loss of 8C/30, 6C/40 and 4C/80 led to very high cell voltages at the end of the test, and the corresponding high voltages certainly contributed to the observed impedance rise.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In contrast, for cells that remained below 4.3 V throughout the entire test, the increase in ASI continued linearly (Fig. 4c). The rate of this increase was surprisingly very similar for all systems, with the magnitude of ASI rise at the end of the test being similar to the experienced by a baseline cell tested at 0.5C/100 (red line). (6C/30 was a clear outlier, likely due to its high initial impedance rise.) That is, when cell voltage is maintained within "safe" bounds, ASI increases at a similar pace at every partial cycle, regardless of rate or state-of-charge; one cycle at a low rate to a high SOC is just as damaging to the cell as one at a high rate to a low SOC. This observation suggests that higher rates promote a larger impedance rise per 100% SOC variation, which becomes clear when ΔASI is represented as a function of the number of equivalent full cycles (Fig. 6). The ASI rise per equivalent cycle is similar at 8C and 6C (Fig. 6a), which is slightly larger than at 4C (Fig. 6b) and considerably larger than at lower rates (Fig. 6c).

Zoom In Zoom Out Reset image size

An interesting feature of Fig. 6 is that, at the rates assembled in each panel, the ASI increase per equivalent full cycle appears to be independent of the SOC range covered in each partial cycle. In other words, at a given charging rate, ASI rise is independent of the limiting SOC of the partial cycles; the "damage" to the cell is a function of rate and rate alone. As will be discussed below, this is extremely significant to the development of fast-charging protocols, as it suggests that limiting the amount of charge that is injected into the cell at higher rates will only postpone the aging mechanisms that cause impedance rise, rather than preventing them from happening.

Upon completion of the tests described above, electrodes were "harvested" from selected full-cells and re-assembled into half-cells (vs Li metal) to quantify permanent capacity losses. These half-cells were cycled at C/100; at such low rate impedance effects are negligible and most electrode domains still maintaining electronic connectivity should be electrochemically responsive.

Figure 7a shows the C/100 relithiation profiles of the harvested cathode half-cells. All capacities are within 0.1 mAh cm⁻² of the measured value for electrodes that just underwent formation cycles (unaged), indicating that a negligible amount of Li⁺ sites are permanently inactivated during the tests. Half-cells containing electrodes from cells tested under five additional conditions (omitted for brevity) displayed these same trends, in spite of the various levels of ASI rise observed during the experiments. Furthermore, the normalized relithiation profiles (Fig. 7b) indicate that the reaction paths of Li⁺ sites also remain unchanged.

Zoom In Zoom Out Reset image size

Harvested graphite anodes tell a different story. Figure 7c shows that the negative electrode from all analyzed cells exhibit some level of capacity loss, which could be as high as 0.44 mAh cm⁻² (∼14% of the electrode capacity). Furthermore, these losses correlate with the number of equivalent full cycles experienced by the various full-cells (Fig. 8a), suggesting that this behavior is simply a response of Gr to successive cycling. Consequently, larger losses were observed at lower rates (1C and 2C) while the severe Li plating noted in some conditions (e.g., 8C/30) did not appear to cause additional loss of the ability of Gr to respond to charge. Inactivation of Gr domains are typically attributed to mechanical damage due to successive cycles of volume change (particularly at edge sites),^12,17,22 and severe pore clogging due to the deposition of reduced electrolyte species.³⁴

Zoom In Zoom Out Reset image size

Capacity loss of the graphite electrode was also tracked in the full-cells through differential voltage analyses of the C/25 discharge profiles,¹⁹ as exemplified in Fig. 8b for 4C/80. In these dV/dQ plots, peaks indicate single-phase regions of Gr, and hence a decrease in the separation between successive peaks denote losses of anode capacity; that is, it indicates that the capacity stored in the plateau in between these two single-phase regions (peaks) has decreased. The dV/dQ analysis generally agrees with the observations in Fig. 8a, which show that capacity loss scales with the number of equivalent full cycles. However, it also indicates that the onset of measurable capacity loss could occur much earlier for cells cycled at higher C-rates; e.g., after 60 equivalent full cycles at 1C/100 and 2C/100, but after only 16 and 18 for 4C/80 and 8C/30, respectively. Thus, although the extent of "damage" to Gr scales with the overall SOC swing experienced by the electrode, capacity loss consistently had an early start in tests at higher rates.

Another interesting feature observed in harvested Gr electrodes is that obvious changes to the staging behavior occurred in one system (4C/80, Fig. 7d). While the graphite domains that remained active presented the expected behavior in all other electrodes,²⁸ the Gr from 4C/80 exhibited interruptions in the two largest plateaus, with a smoother transition to the single-phase segments of the profile. Very interestingly, this distortion in plateaus was not observed when additional cycles at C/100 were performed with the same cell, showing that staging behavior could be restored (Fig. 7d, shown as dashed dark blue line). Our group has observed similar distortions in the past, and they could always be "healed" by a single slow cycle in half-cells. In nearly all cases these distortions occurred in samples that have undergone significant Li plating, suggesting that this is likely a kinetic effect, rather than structural. Given the high levels of capacity fade observed in 4C/80 (Fig. 4c), it is not clear whether the change in plateau could be affecting full-cell operation, nor if a similar "healing" could be achieved in the full-cell itself.