Performance of Commercial Li-Ion Cells for Future NASA Missions and Aerospace Applications

The physical characteristics of the cell components, including anodes, cathodes, separator and can thicknesses are listed in Table I. The height of the electrodes or jelly roll is around 59 ± 1 mm. The lengths of the electrodes, both anodes and cathodes, are generally in the range of 600 ± 20 mm for the high energy cells. For the two high-power cells, Samsung 30Q and LG Chem HG2, the electrodes are longer by about 50%. The electrodes are also thinner for these two high-power cells, measuring about 100 μm for both anodes and cathodes. In contrast, the high-energy electrodes are considerably thicker at 170–200 μm and 150–160 μm for anodes and cathodes, respectively. Electrode thicknesses were not measured for Samsung 35E, 36 G and Sony VC7 cells. The thickness of the separators ranges from 13–20 μm and the cans are about 150–200 μm (6–8 mil) thick.

Figure 2 shows the XRD pattern of the anodes from the first batch of the cells. As labelled in the figure, all the anodes show strong peaks suggesting some graphitic form of carbon. It is interesting to note that the peaks corresponding to Si are absent, which implies that Si is absent, or present in a small quantity undetectable by XRD, or in the amorphous form.

Zoom In Zoom Out Reset image size

Figure 3 shows the XRD pattern of cathodes from these cells. As labelled in the figure, the diffraction patterns contains peaks typical for a LiMO₂ layered structure plus Al foils. Specifically, the cathode compositions include lithium nickel manganese cobalt oxide with different ratios of Ni, Mn and Co (NMC) or, lithium nickel cobalt aluminum oxide (NCA); in some cases the addition of manganese spinel oxide, LiMn₂O₄ (LMO) was also detected.

Zoom In Zoom Out Reset image size

In order to understand the morphology and quantitively identify the composition, SEM examination and EDS analyses were made on the cathodes. Figure 4 shows the SEM images and EDS analysis of the cathodes from LG Chem MJ1 cells, for example.

Zoom In Zoom Out Reset image size

The cathodes typically contain primary particles of ∼1–5 μm, agglomerates of which form secondary particles of 25 μm. Energy-dispersive X-ray spectroscopy (EDS) measurements were made at various locations on the cathodes and anodes to quantitatively identify electrode material compositions. All the cathode EDS spectra are shown in Fig. S1 (supplementary figures) and the results are summarized in Table II.

We inferred the major components of the electrolyte through ¹H, ¹⁹ F and ³¹ P NMR analysis of d₆-DMSO washings obtained from separator samples. ¹H spectra indicated the organic solvent species present, while ¹⁹ F and ³¹ P spectra allowed us to identify the lithium salts (often in combination with their hydrolysis products). A number of minor species were also observed in a number of cases, but these were not able to be assigned unambiguously. No quantitative analysis was attempted due to likely solvent evaporation and salt hydrolysis during the DPA process and storage before sample preparation.

Table II gives the summary of the compositional analysis of the cathodes, anodes, and electrolytes from these cells. All the cells have essentially graphite as the anode material, which in some cases (LG Chem M36 and Murata VC7) is less crystalline. There was also a small amount of Si detected in the Samsung 35E cells. Based on the XRD and EDS data, the cathodes are either.

NMC with 80%–90% of Ni, NCA, or a blend of NCA and LMO (in LG Chem M36). The electrolyte in all these cells is a blend of carbonates (EC and DMC), with LiPF₆ salt as well as LiFSI in small quantities, apparently as an additive, or in some cases in higher concentration as part of a mixed salt system.

Figure 5 shows cycle life performed at +20 °C in a temperature-controlled chamber expressed in terms of capacity, percent capacity retained, specific energy and watt-hour efficiency. The first batch of cells, i.e., LG Chem MJ1, Samsung 35E, LG Chem M36, Panasonic BJ, and Murata VC7 have completed ∼1400 cycles with good capacity retention and are continuing to cycle. One of the two 35E cells failed open after 200 cycles, possibly due to internal pressure build- up causing a current-interrupt device (CID) to open; the other 35E cell continued to cycle. Some of the cells which were added to this test later, e.g., LG Chem HG2 and Samsung 30Q, Samsung 36 G and Panasonic GA have completed about 800 cycles so far. Figure 5A shows the variation of capacity during cycling of these cells, while the capacity retention is displayed in Fig. 5B. Figure 5C provides the variation of the specific energy of the cells during cycling and the energy round-trip efficiency is shown in Fig. 5D.

Zoom In Zoom Out Reset image size

The LG Chem MJ1 cells displayed best cycle life, with the highest initial specific energy (>230 Wh kg⁻¹) as well as the best capacity retention during cycling (90% capacity retention after 1100 cycles). Both Samsung 35E and Panasonic BJ have an initial specific energy of 220 Wh kg⁻¹ and retained about 80% of the capacity after 1100 cycles. Both LG Chem M36 and Murata VC7 start off with ∼210 Wh kg⁻¹ and retain about 85% and 82%, respectively after 1100 cycles. In the recent batch of cells, Samsung 36 G and Panasonic GA cells show ∼220 Wh kg⁻¹ initially and a retention of ∼95% over 500 cycles. The high-power cells, Samsung 30Q and Lg Chem HG2 have 200–210 Wh kg⁻¹ with comparable fade rate during cycling.

The round-trip efficiencies (Fig. 5B) average around 96%, with the exception of LG Chem MJ1 cell with higher (97%) and Panasonic BJ cell with lower efficiencies. The two high-power cells, LG Chem HG2 and Samsung 30Q show even higher energy efficiency >97.5%, as expected from their lower cell overpotentials during charge and discharge.

EIS vs cycling .–EIS was performed on the +20 °C cycle life group cells every 100 cycles to attempt to characterize the impedance growth. Cells were charged to 4.10 V (100% state of charge) before each measurement. Figure 6 shows the Nyquist or complex plane impedance plots of the first batch of cells measured at +20 °C after every 200 cycles.

Zoom In Zoom Out Reset image size

The above Nyquist plots of all the cells show two distinguishable loops. In all these cases, there is a gradual increase in the cell impedance upon cycling, especially at low frequencies. These changes may be understood quantitatively in terms of the changes in the anode and cathode behavior during cycling, as detailed below.

The EIS behavior may be correlated to the electrochemical processes within the cells, using appropriate equivalent circuit model. A simplified representation of an electrochemical process is an equivalent circuit comprising a resistance in series with a parallel resistor-capacitor network representing the charge transfer kinetic process, and a Warburg element characteristic of an ideal semi-infinite linear diffusion in some cases, as described in detail in Ref. 23. Recently, Choi et al., have provided a good review on the modeling and applications of electrochemical impedance spectroscopy for lithium-ion batteries. ²⁴ Each of the electrode responses in a Li-ion is typically represented by a series resistance, a high-frequency loop corresponding to the interphase layer and a low frequency loop corresponding to charge transfer kinetics and a Warburg impedance characteristic of diffusion across the interphase layer or electrode bulk. The solid electrolyte interphase (SEI) layer, which is generated by the decomposition of the electrolyte and/or its reactions with the electrodes has a more profound influence on the electrochemical characteristics of anode than cathode. The individual EIS response at the anode and cathode may be experimentally determined separately in half-cells or in three-electrode cells. We have routinely used the EIS technique with the above equivalent circuit to quantify the interfacial and kinetic effects of graphite anode and metal oxide cathodes, especially with different electrolytes, which affect the SEI characteristics. ^25,26

The EIS response of the overall Li-ion cell is a series combination of the responses from individual electrodes, with surface film and kinetic effects of graphite anode metal oxide cathode, expressed over characteristic frequency ranges depending on relative anode and cathode kinetics.

In the absence of a reference electrode, it is no doubt difficult to assign the semicircles or Warburg impedance to anode or cathode. However, it was shown in our earlier studies with three-electrode cells with electrodes used in prototype cells containing MCMB anode and lithium nickel cobalt oxide cathode, that the high-frequency semicircle relates to the anode SEI and the low-frequency relaxation loop relates to the cathode kinetics. ²⁷ We used similar assignments here for the analysis of the EIS data in Fig. 6, using the equivalent circuit in Fig. 7. For an improved fit, the capacitive elements were replaced with constant-phase elements (CPE), whenever needed. Figure 8 shows the changes of the series ohmic resistance (Fig. 8A), SEI resistance on the anode (Fig. 8B) and the charge transfer or polarization resistance (Fig. 8C) at the cathode at various stages of cycling.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Except for the Panasonic BJ cells, the series ohmic resistance of the cells is ∼25–30 m Ω and shows little change during cycling (Fig. 8A). The LG Chem M36, Sony and Samsung cells have the lowest series resistance of 24 mΩ increasing to ∼25–27.5 mΩ after 1200 cycles. The MJ1 cells on the other hand, have ∼30 mΩ before cycling, which showed no change during cycling. Both the BJ cells have higher series resistance of 43 m Ω even before cycling, which increase to 55 m Ω after 1200 cycles, suggesting that the electrolyte is relatively resistive and becomes even more resistive in the course of cycling compared to the rest of the cells.

Except for the Panasonic BJ cells, the anode SEI resistance values of the cells are ∼5 mΩ for M36, MJ1, Samsung and Sony cells, before cycling (Fig. 8B). These values have increased little, by ∼1 mΩ over 1200 cycles, suggesting a robust and stable SEI in both cases. In contrast, the Samsung and Sony cells have shown slightly higher increase in the anode SEI resistance by 2–3 mΩ to 7 mΩ after 1200 cycles. As with the series resistance, the Panasonic BJ cells have slightly higher initial SEI resistance of ∼7 mΩ, which had decreased initially during cycling from 100–400 cycles but thereafter increased rapidly to 21–22 mΩ, once again suggesting an instability of the anode/electrolyte interphase, which is also reflected in rapid capacity fade during cycling.

The cathode charge transfer resistance estimated from the low frequency semi-circle shows similar trend (Fig. 8C). The M36 cells have the lowest charge transfer resistance of 1.3 mΩ initially, which increased to 5–6 mΩ after cycling. The Samsung and Sony cells have an initial value of 2.0 mΩ increasing to ∼10 mΩ. The MJ1 cells have slightly higher charge transfer resistance of 2.4 mΩ, which increased to 15 mΩ. This is surprising and not expected from the

Lowest capacity fade rate shown by the MJ1 cells during cycling. Probably the stable anode SEI of the MJ1 cells may be compensating for the increase of cathode polarization resistance. Alternately, the effects of cathode impedance (at low frequencies) may be evident only at higher discharge rates, but not at the moderate rate of C/5 used for cycling. The Panasonic BJ cells, are once again out of family, with an initial charge transfer resistance of 6 mΩ increasing sharply to ∼80 mΩ after 1200 cycles, which is consistent with their poor cycle life.

The Warburg impedance, which is related to the diffusion coefficient of Li ⁺ ion, is in the range of 300–400 mΩ for most of the cells; the LG Chem MJ1 and M36 cells have Warburg impedance closer to 400 mΩ, while the Samsung and Sony cells have the values closer to 300 mΩ. In all these case, these values are constant during cycling, suggesting little change in the cathode bulk. The Warburg impedance of the Panasonic BJ cells is also ∼300 mΩ initially, but is quickly reduced to ∼100 mΩ, which may be due to a fragmentation of the cathode particles during cycling.

Often, Li-ion batteries are expected to operate at lower temperatures in planetary missions. Even though the delivered capacities are lower at low temperatures, the cycle life is generally found to be better, unless the temperature is low enough to cause complications from lithium plating. Cells were cycled at 0 °C, with both charge and discharge at C/5 between 3.00 and 4.10 V and 15-min rests between steps. Figure 9A shows the cycle life of LG Chem MJ1, LG Chem M36, Samsung 35E, Murata VC7 and Panasonic BJ cells. Before beginning the cycle life test and every 100 cycles thereafter, a characterization test was performed at +20°C to measure capacity and DC resistance by current interrupt.

Zoom In Zoom Out Reset image size

All the cells have shown high specific energy of ≥190 Wh kg⁻¹ at 0 °C, with LG Chem MJ1 cells offering the highest specific energy of ∼205 Wh/kg. Once again, the LG Chem MJ1 and LG Chem M36 cells display the highest capacity retention during cycling, with 97% of initial capacity after 250 cycles, followed by Samsung 35E with 94%. Both Samsung 36 G and Panasonic GA cells show lower retention of ∼89% and some recovery after each 100 cycles, when the +20 °C characterization was performed.

Table IV summarizes the collected data on capacity fade due to cycle life at both +20 °C and 0 °C, expressed as percent capacity lost after 500 cycles compared to the first cycle at that temperature. The MJ1 cell has a fade rate of ∼7% at 20 °C, which is reduced to ∼4% at 0 °C. Both LG M36 and Samsung 35 E cells also have their fade rate reduced from ∼10% to 4% and 6%, respectively. In general, the fade rate is lower during cycling at 0 °C, compared to cycling at 20 °C.

DC resistance vs cycling at 0 °C.–Performing EIS on this group of cells was not practical due to the particular circumstances and location of the test. In lieu of this, every 100 cycles the cells' DC resistance was measured at 0 °C, and then the cells were warmed to 20 °C where capacity and DC resistance were measured. DC resistance was measured by current interrupt with a discharge pulse of 1 A at 100%–20% SOC in 20% increments. Figure 10 shows these resistance values. Some cells showed a decrease in measured resistance between the initial measurement and after the first 100 cycles. LG M36 and MJ1 showed very little change in the resistance values over the course of 800 cycles; Samsung 35E showed an initial decrease of resistance followed by growth of roughly 12%–15%, while the Panasonic GA displayed a growth of roughly 20%–40% over 500 cycles (Panasonic GA cells have completed fewer cycles than the other cells in this comparison). Samsung 36 G cells nearly tripled in resistance over the course of 800 cycles (missing data points for these cells are a result of the DC pulse triggering a low voltage limit and not completing the measurement). These resistance growth trends are consistent with the relative cycling performance of these cells shown in Fig. 9.

Zoom In Zoom Out Reset image size

For applications requiring long cycle life, batteries are designed such that the depth-of- discharge (DOD) in each cycle is low. Cycle life of aqueous batteries, especially nickel-cadmium and nickel-hydrogen batteries, improves exponentially as the depth of discharge is reduced, e.g., reducing the depth of discharge by half from 80% to 40% can improve the cycle life from 2000 to 20,000. This may be related to the mechanical or even thermal stresses that the electrodes experience at deeper discharges, contributing to faster degradation. Similar effects are observed for lithium-ion batteries also, with the cycle life increasing by several fold at low depths of discharge. ²⁸ Additionally, it is preferred to adopt a lower charge voltage together with low DOD; for example, for a DOD of 40%, it is better to cycle from 90% state-of-charge (SOC).

to 50% SOC, compared 100%–60% SOC. During cycling of Li-ion cells with high charge voltages, there are electrolyte-induced degradation processes occurring both at the anode and cathode, and there is also the risk of lithium plating at the anode, especially at high charge currents. The higher the depth of discharge, the higher the utilizable energy, but the cycle life will be shorter, which may be due to the increased mechanical stresses during deep lithiation-delithiation, and thermal effects from increased heat dissipation rates from the cells at deep discharges. Selecting appropriate depth of discharge for the required mission life is a strategy used in electric vehicles, as well in space applications, including satellites, orbiters and long-duration rovers. An electric vehicle (EV) battery, for example, cycles between 80%–20% SOC with the 20% capacity at either end acting as a reserve. Typically sized for a range of 250 miles per cycle, the battery will need to survive over 4000 cycles at 60% depth of discharge, to provide one million miles. Some of the cells tested here display impressive cycle life even at 100% depth of discharge, retaining 80%–90% of their initial capacity, suggesting that at 60% depth of discharge, the cycle life may be expected to be ∼ 10,000 cycles, which, combined with the higher specific energy of these cells compared to current EV battery technologies, may translate to be ∼2.5 million miles.

In contrast to the electric vehicles that typically undergo one discharge-charge cycle per day, the Earth-orbiting satellite and planetary orbiters have more challenging cycling profiles. The batteries are required to provide power for a short time during eclipse (loss of solar exposure for photovoltaic power) and must be quickly recharged back to the original state-of-charge. One typical scenario of this kind is a 90-min cycle comprising 30 min of discharge followed by 60 min of charge, accumulating 16 cycles per day or 5840 cycles per year. The battery is expected to support the mission for several years, with an anticipated cycle life of tens of thousands of such cycles. With this fixed duty cycle, higher depth of discharge implies higher discharge current as well as charge current. Cycle life in these instances is defined as the number cycles until the discharge voltage drops below the minimum cell voltage (typically 3.0 V).

In order to assess suitability for satellite and orbiter applications, the cells were evaluated in a 90-minute cycle regime at two different depths of discharge: 20 and 40%. The discharge current was selected based on a cell's nameplate capacity to achieve the desired DOD in 30 min, and the charge current was two-thirds of the discharge current (with a constant voltage taper) to ensure complete recharging of the cells in 60 min. Figure 11 shows the performance (two cells of each type) during LEO cycling at 20% depth of discharge, each segment comprising 498 lEO cycles and two 100% DOD cycles, with the 500th cycle determining the cell capacity. The discharge rate corresponds to C/2.5 and the charge rate to C/3.75. Figure 11A (left) shows the variation of the End-of-Discharge Voltage (EODV, after 30 min), Fig. 11B shows the capacity the cells during LEO cycling, measured at 100% DOD every 500 cycles, and Fig. 11C shows the capacity retention (%) during cycling.

Zoom In Zoom Out Reset image size

The EODV increases marginally after each 500 cycles, due to the cells being recovered during the (100%) capacity measurement at a slower discharge rate of C/5. In the mix, there is one power cell (Samsung 30Q) for comparison. So far, the cells have completed 6500 cycles, and the cells exhibit stable EODVs, except in the case of Samsung 35E cell, which showed an initial decrease but has recovered subsequently. As expected, the power cell Samsung 30Q has higher EODV, indicative of its ability to support high discharge and charge currents with less polarization. All the cells retain impressive capacities of >90% of initial values after 6000 cycles.

Figure 12 shows the performance of cells cycled under LEO conditions at 40% depth of discharge (two cells in each case), with the cell capacities determined after each 500 lEO cycles. The discharge and charge rates were selected to conform to the 90-minute cycle described earlier and correspond to C/1.25 and C/1.875, respectively. Figure 12A (left) shows the variation of the EODV, Fig. 12B shows the capacity of the cells during LEO cycling, measured at 100% DOD every 500 cycles and Fig. 12C shows the capacity retention (%) during cycling.

Zoom In Zoom Out Reset image size

In all the cases except the Panasonic GA cell, the EODV is stable through 6500 lEO cycles at 40% DOD so far. The Panasonic GA cell sustains larger capacity fade during cycling, and further studies are needed to understand its poor performance; one of the two GA cells under test at 40% DOD failed open after 4500 cycles, while the other is still ongoing. Samsung 30Q displays higher EODV, as expected from its ability to support high discharge and charge currents, followed by the LG Chem cells, MJ1 and M36, which retain an impressive 85%–89% of their initial capacities after 6500 cycles. Panasonic GA cell and to some extent Samsung 35E display rapid capacity fade with the capacity reducing to 75% and 85%, respectively. Extrapolating this behavior from Fig. 13, which illustrates the cycle life at different depths of discharge and is termed as the Wohler curve, the cycle life at 100% DOD is estimated to be 3000 cycles (to 80% capacity), cycle life at 40% extrapolated to be 25,000 (to 40% SOC) and the cycle life at 20% depths of discharge extrapolated to the 60,000 cycle (to 20% SOC). This is similar to the Wohler's curve for Li-ion batteries,

Zoom In Zoom Out Reset image size

which is a plot of depth of discharge (%) vs the number of cycles realized from the battery, developed through modeling. ²⁹ Based on this data, it is clear many of these cells, e.g., LG Chem MJ1 and M36 and Samsung 35E are expected to provide more than 4000 cycles at 60% depth of discharge, with impressive useable specific energies of 120 to 140 Wh kg⁻¹.

Many papers in literature describe the performance degradation of Li-ion cells during cycling, which are briefly discussed here. In a lithium-ion cell, there are several processes that contribute to the capacity fade during cycling under normal conditions of operation, i.e., SEI formation due to electrolyte decomposition, ³⁰ SEI re-formation caused by cracking of the layer, a decrease in accessible surface area and porosity due to SEI growth, contact loss of active material particles due to volume changes during cycling, cathode electrolyte interphase (CEI) growth, cathode surface restructuring ³¹ and transition-metal dissolution from the cathode that can cross over and affect the anode SEI, ³² phase changes in the insertion materials ³³ and crack propagation and fracturing of electrode particles, ^34,35 increasing Li consumption for film formation. In the non-optimum operating conditions, e.g., during overcharge and charging at high rates, especially at low temperatures, ³⁶ there may be issues related to lithium deposition on the anode and electrolyte oxidation at the cathode releasing gaseous products. All these processes cause loss of lithium inventory and loss of both anode and cathode active materials.

With several processes occurring due to different causes and at different rates, either independently or through an interdependence, it is difficult to model the capacity degradation during cycling. ³⁷ Each of these degradation processes have been modeled separately. ^38,39 Recent models of capacity degradation ⁴⁰ suggest three phases of capacity fade: Phase-1 corresponding to a rapid initial capacity drop, as Li is consumed during SEI formation/ reformation, Phase-2 reflecting a moderate drop in capacity following an almost a linear degradation, associated with loss of Li inventory in side reactions, possibly bulk effects in the electrodes, and Phase-3 showing rapid capacity fade leading a quick cell failure, attributed to lithium plating and subsequent impedance increase. ⁴¹ The recent comprehensive data on various Li-ion cells reported by Preger et al. ¹⁴ is consistent with this trend.

We have attempted to elucidate analytical expressions for the cycle life behavior presented in Fig. 5. In almost all cases, except LG chem MJ1 cells, the cycle life curves show different fade rates, indicative of different fade mechanisms in operation, maybe even concurrently (Fig. 14). In all the cases, fade rate is high initially and levels off, which may be attributed to the SEI formation. Subsequently, the fade slows down, showing an almost linear fade rate in all the cases. Only in the case of Panasonic BJ cells, the fade rate becomes steeper, implying the onset of the third mechanism dealing with possibly lithium plating. Analytically, the capacity life follows a simple exponential decay in the case of LG Chem MJ1 cells, which may be fortuitous. In all the other cells, it is difficult to establish a simple correlation. Instead, the cycle life trends as third or even fourth-order polynomial equations, highlighting the complexity of describing the capacity fade using simple kinetics equations for a single degradation mechanism, and the need to invoke multiple degradation processes within the cells (Fig. 14).

Zoom In Zoom Out Reset image size

As evident from the above discharge characterization tests, these cells provide impressive power densities in addition having high specific energies. Even though most of the applications do not require higher discharge than 1 C rate, there are applications, e.g., planetary ascent vehicles, planetary aerial vehicles (helicopters) and electric aircraft, which require much higher power densities. The first batch of cells, i.e., LG Chem MJ1, M36, Samsung 35E, Panasonic BJ and Murata VC7 cells were evaluated at a high continuous discharge current of 9.6 A (3.3 C rate) at temperatures from 20 °C to 50 °C (Fig. 16). These tests were performed in a convective thermal chamber, and also with the chamber turned off following a 4 h soak at the relevant temperature. The cell temperatures were monitored during discharge.

Zoom In Zoom Out Reset image size

Both LG Chem MJ1 and M36 cells show impressive performance providing more than 2 Ah capacity even at room temperature at 9.6 A, which continues to improve at warmer temperatures. In contrast, Samsung 35E and Murata VC7 cells provide ∼1.5 Ah capacity at 20 °C, 2 Ah at 30 °C and >2 Ah capacity only at warm temperatures of 40 °C–50 °C. The Panasonic BJ show reduced performance of ∼1 Ah at the high rate even at warm temperatures. In all cases, there is considerable self-heating of the cells, with the cell temperatures increases by ∼16 °C–18 °C at all start temperatures. The steady state heat dissipation at the 3 C discharge rate is estimated to be 4 W, estimated using the open circuit values as thermo-neutral potentials, which was shown to be a valid assumption. ⁴² The heat rate of 4 W is substantial and is about 6% of the energy released during thermal runaway. ^43,44 In short, the cells can indeed provide high power densities of ≥700 W kg⁻¹ in a continuous mode with appreciable capacities, especially at warm temperatures, but their high rates of heat dissipation needs to be managed with suitable thermal designs.