Improving performance of Li-Ion cells demands high precision, high accuracy testing
Lithium-ion batteries have become staples of modern electrical and electronic products; in fact, they are now used in a wide range of applications from smartphones and power tools to medical equipment and electric vehicles. Well-made Li-ion cells can easily meet the needs of smartphone, laptop and power tool applications because of their high energy density, long calendar/cycle life, and relatively low cost in comparison with other rechargeable battery technologies, such as nickel-metal hydride. However, developing cells capable of cycling with minimal capacity and energy loss for ten years as required by automotive applications or even longer for grid energy storage is far from trivial.
Designing such high quality cells can be complicated, as can the process of evaluating their performance. Cell manufacturers constantly make small changes to the cell design that can impact the cell chemistry as part of their efforts to extend the cycle life or reduce the cost without sacrificing performance. However, testing all of these different experimental cell chemistries under real-world conditions would be impractical. Consider, for example, a battery for an electric vehicle that is only cycled once per day over its ten-year lifetime; obviously, it would take far too long to determine whether the experimental changes made were beneficial and therefore should be implemented in commercial cells. No one can afford to wait years to complete the R&D feedback loop. Cell chemistry researchers need tools and techniques that allow them to perform reliable accelerated lifetime testing to determine if proposed changes extend cell lifetimes.
Accelerated lifetime testing
The most common form of accelerated lifetime testing involves high rate cycling; cells are charged and discharged at rates up to one cycle per hour (up to around 20 cycles per day, many more than would occur in actual use) in order to acquire data for hundreds or thousands of cycles over an experiment of reasonable length (several months). Over these cycles, the experimental cell’s loss of capacity and energy are measured and compared with those of a control cell or a cell already in production to decide whether the experimental chemistry offers any benefits.
The degradation of Li-ion cells is not only cycle-dependent but has a strong time dependency. Smith et al. [Ref. 1] showed that cells being cycled at different low rates (either one cycle per two, four or eight days) all failed after the same amount of time, despite differing in the number of cycles completed by factors of two and four. This means that if a cell is measured to only lose 10% of its initial capacity after 1,000 cycles in two months, it does not mean that the same cell would only lose only 10% of its capacity after 1,000 cycles over three years.
Work in Dr. Jeffery Dahn’s research group in the Department of Physics and Atmospheric Sciences at Dalhousie University has suggested a new method for distinguishing between the lifetimes of different experimental cells within just a few weeks: High Precision Coulometry. Given that the side reactions within cells (that is, solid electrolyte interphase or SEI growth, electrolyte oxidation, transition metal dissolution, etc.) all involve transferring charge that is not associated with the intercalation/deintercalation of lithium from the electrodes, those reactions can be detected coulometrically. If all of the lithium stored in the negative electrode on charge was returned during the subsequent discharge, then the charge (QC) and discharge (QD) capacity would be equal, so the coulombic efficiency (CE = QD/QC) would be exactly unity and the cell should be able to cycle indefinitely. However, due to these parasitic reactions occurring within a cell, the coulombic efficiency is less than the ideal value of 1.0000 and the cell degrades. This causes a typical voltage versus capacity curve to “slip” to high absolute capacities with subsequent cycles because the discharge capacity is always less than the previous charge capacity.
Figure 1. A typical voltage vs. capacity (V-Q) plot for 13 cycles. The insets show the top of charge and bottom of discharge endpoints shifting to the right with continual cycling.
Figure 1 is a typical V-Q curve showing this type of behaviour, with the insets showing the top of charge and bottom of discharge endpoints to better illustrate the rate at which the curve slips to the right. Because the coulombic efficiency is defined as the discharge divided by previous charge capacity, it is directly related to the rate of motion of the bottom of the discharge endpoint, referred to as discharge endpoint slippage – ΔD (CE = QD/QC = 1 – ΔD/QC). The top of charge endpoint slips to higher capacity with subsequent cycling as well, referred to as charge endpoint slippage. This can be measured independently of the coulombic efficiency because it primarily relates to reactions that occur at the positive electrode [Ref. 2]. Given that parasitic reactions cause the voltage curve to slip to the right (decreasing coulombic efficiencies and increasing charge endpoint slippage), then cells with higher coulombic efficiencies and lower charge endpoint slippage rates must have lower rates of parasitic reactions and therefore should have longer cycle lives. This idea is the underlying premise for using High Precision Coulometry as a way to compare cell lifetimes in short-term experiments.
Cell coulombic efficiency
The coulombic efficiency of cells is highly rate-dependent [Ref. 3] because these parasitic reactions occur as a function of time, not cycle (the rate of reaction is also highly dependent on cell voltage and temperature). Because these reactions are occurring all the time, with decreasing cycling rate (longer cycles), the coulombic efficiency further departs from 1.0000 because there is a long time in each cycle during which these parasitic reactions can lead to cell degradation. In contrast, when cycling cells at high rates, the coulombic efficiency becomes much closer to unity because there is very little time for parasitic reactions to occur during the span of a single cycle.
The coulombic efficiency of well-made cells in the early cycles at moderate temperature and relatively low cycling rate (10–20 hours per charge or discharge) is very high and can vary from 0.9980–0.9999, depending on cell chemistry. This poses the problem of accurately differentiating between the coulombic efficiencies of different experimental cells because small improvements to the coulombic efficiency (such as 0.0005) could lead to significant increases in cell lifetime.
Therefore, it is crucial to verify that the error in the coulombic efficiency is less than 0.0001 to ensure that measured changes between different experimental cells are truly measures of differences in the rates of parasitic reactions within the cells and not due to variability in the cycling equipment.
Development of the Ultra High Precision Charger
Commercially available battery testing equipment can’t differentiate between the coulombic efficiencies of cells at this high level of accuracy and precision [Ref. 4, 5]. Therefore, Dr. Dahn’s group has configured a battery cycler using high precision current sources and multimeters called the Ultra High Precision Charger (UHPC). Figure 2 shows a 100-channel UHPC built in 2013 using Keithley Model 2602B System SourceMeter Source Measure Unit (SMU) instruments (Figure 3) as current sources and Keithley Model 2002 8-1/2-digit multimeters (DMMs) for voltage measurements (Figure 4). These instruments were selected because their specifications met the requirements to achieve the desired accuracy and precision in capacity, and therefore coulombic efficiency, measurements [Ref. 4]. Figure 2 also includes a simple schematic of the system components.
Figure 2. The 100-channel Ultra High Precision Charger in the Department of Physics and Atmospheric Sciences at Dalhousie University (Halifax, Nova Scotia, Canada) and a simple schematic of the charger setup.
Figure 3. In the UHPC, Keithley’s Dual-channel Model 2602B System SourceMeter Source Measure Unit (SMU) instruments function as current sources. Each channel supports sourcing DC currents from the 100 nA range (with 2 pA resolution) up to 3A (with 20 μA resolution). It can also source pulsed currents up to 10A.
Figure 4. Keithley’s Model 2002 Digital Multimeter can measure voltages with up to 8-1/2 digits of resolution because its 28-bit A/D converter provides the resolution needed to discern smaller changes in voltage output. On the lowest DC voltage measurement range (200 mV), it can resolve voltages as low as one nanovolt in enhanced accuracy mode.
Each channel includes a dedicated current source, a set of precision “sense” resistors in line with the current output (switched to different resistance values based on the set current), and connections from the multimeters to measure the voltage drop on the sense resistor, which is necessary to calculate current flow as a function of time and the voltage of the cell under test. This system can source currents from micro-amps to 3A with very high accuracy (around 0.03% within each decade of current) and low noise (1–20ppm depending on the range). This very precise current is also monitored to track any fluctuations or small drifts with time to further improve the accuracy of the capacity calculations and the subsequent coulombic efficiency calculations.
The very high precision voltage measurement ensures that cells are always cycled between exactly the same voltage limits for each cycle so that capacities can be properly compared. The high level of accuracy in the voltage and capacity measurements produces voltage vs. capacity curves with virtually no noise, which offers additional advantages, such as the ability to perform differential capacity (dQ/dV) [Ref. 6] and differential voltage (dV/dQ) [Ref. 7] analysis to understand the degradation of cells over time. At this level of precision, it’s critical that the temperature of the cell under test is well controlled because the voltage of a Li-ion cell depends on temperature; under typical testing procedures, the charge and discharge endpoints are defined by the cell voltage.
Earlier charging system implementation
A 60-channel test system built in 2009 did not originally have the sense resistor and second multimeter connection in each channel that the UHPC does. This system, known as the High Precision Charger (HPC), incorporates Keithley Model 220, 224 and 6220 current sources and Model 2000 6-1/2-digit DMMs. Although these current sources offer accuracy and noise specifications similar to the Model 2602Bs, they are limited to a maximum output current of 100 mA. This system was upgraded in 2012 to incorporate precision sense resistors and additional DMMs to monitor the sense resistor voltage and now offers performance comparable to the UHPC.
Figure 5. The precision (left) and accuracy (right) of coulombic efficiency measurements made on the UHPC, the HPC, and a Maccor Series 4000 charger, expressed in parts per million (ppm).
Figure 5 illustrates the differences in accuracy and precision in the coulombic efficiency data collected using the HPC (when the current was assumed to be exactly that of the set-current from the Keithley current source), the UHPC, and a Maccor 4000 Series battery testing system. Machine-made identical cells were cycled on all three systems under identical conditions (temperature, cycling rate and voltage range) to collect the coulombic efficiency vs. cycle number data. The accuracy and precision were calculated using the difference between sets of identical cells and the noise in the data, respectively, for all three systems. The cells were cycled at low rates (a full cycle in 40 hours) to improve data quality; at low cycling rates, it is easier to ensure that the cell is only charged or discharged to precisely the correct cut-off voltage. At higher rates, due to limitations in the voltage sampling rate of different systems, the cells can be charged slightly beyond the cut-off voltages, which adds noise to the capacity and coulombic efficiency data.
The data in Figure 5 makes it clear that commercially available systems, even those of high quality, do not operate at the level necessary to distinguish between the differences in coulombic efficiency of high quality cells. The UHPC is almost two orders of magnitude better in terms of both precision and accuracy than the Maccor 4000 Series system. This figure also demonstrates the benefit of monitoring the current flow as a function of time; the precision and accuracy of the UHPC improves upon that of the original HPC by about a factor of four.
Figures 6 and 7 show cell data collected with the HPC on the coulombic efficiency in the early cycles, along with the long-term performance of those cells. This data supports the idea that cells with coulombic efficiencies closer to unity in the early cycles will have better capacity retention during long-term cycling and therefore longer cycle lives. All of these cells are machine-made by commercial manufacturers, then sent to Dalhousie University for cycling on the HPC (before upgrading the UHPC). Cells are identical except for the use of small amounts of different electrolyte additives believed to affect the performance.
Figure 6. High precision measurements of the coulombic efficiency in the early cycles (left) and long-term cycling performance (right) of prismatic, commercially made Li-ion cells (pair cells shown as crosses and squares where available).
Figure 6 shows the performance of LiCoO2/graphite energy cells, which were cycled at 40.0°C ±0.1°C on the HPC at a rate of C/20 (a cycle in 40 hours) between 3.4 and 4.075 V for about 400 hours. After this cycling, the same cells were moved to a different system for long-term cycling, where they were cycled at 55°C ±0.5ºC at a rate of C/10 (a cycle in 20 hours) between the same voltage limits. After only 200 hours of cycling on the HPC, clear differences between the coulombic efficiencies of these cells became apparent, so they could be ranked from highest to lowest in terms of performance. Cell C is clearly the best performer, followed closely by cells B and D (which are hard to differentiate, even at this level of precision), then a large decrease in performance for cell E, and finally cell A. The long-term cycling data shows exactly the same trends in terms of capacity retention as were predicted by the short-term measurements of coulombic efficiency. Without advanced knowledge about the performance of some cells during long-term cycling, it is difficult to use the coulombic efficiency as a quantitative predictor; however, the qualitative predictions are still highly valuable.
Figure 7. High precision measurements of the coulombic efficiency in the early cycles (left) and long-term cycling performance (right) of 18650-style, commercially made Li-ion cells (pair cells shown as crosses and squares where available).
The cell cycles plotted in Figure 7 show almost no capacity loss until the onset of rapid capacity loss, leading to cell failure [Ref. 8]. These are 18650-style Li[NiMnCo]O2/graphite power cells made to show this type of phenomenon. They were cycled on the HPC at 30°C ±0.1°C at a rate of C/20 between 2.8V and 4.2V for thirteen cycles to collect the coulombic efficiency data shown on the left. They were then returned to the manufacturer for long-term cycling at room temperature on a standard battery cycling system at a rate of 1C (a charge in one hour) with a constant voltage hold at the top of charge. Around cycle 10 of HPC cycling, a glitch in the temperature box caused one point in each data set to be out of line with the others. This serves as a reminder of the importance of temperature control in these experiments.
Once again, clear differences can be seen in the early cycle coulombic efficiencies of these cells; those with higher columbic efficiencies achieve more cycles before showing this rapid capacity loss. These cells were designed to have relatively short cycle lives (only a few hundred cycles) so that the experiment could be run to completion, but the same concept would hold if the cells were made to have cycle lives in the range of thousands of cycles. With cells that show this type of failure, it is impossible to identify cells that are performing better than others using only standard cycling techniques because all cells have identical capacity loss rates until they show rapid failure at different cycle counts. This emphasises the need for precision measurements of the coulombic efficiency in order to identify better performing cells accurately in a short amount of time to accelerate the research and development process.
Conclusion
Our work makes it clear that characterising coulombic efficiency of Li-ion cells in the early cycles can be highly useful in predicting their long-term performance. However, battery testing systems currently on the market do not offer the level of precision necessary to make such measurements on high quality cells with relevant lifetimes for applications like electric vehicles and grid energy storage. The high accuracy and precision of the sources and meters used was essential to building a custom battery testing system capable of making such measurements. Making measurements at such a high level of accuracy requires highly reproducible cells so that changes in measured values can be attributed to changes in the chemistry within the cell rather than cell-to-cell variations, good temperature control of the cells, and proper data collection and analysis methods. Other research groups in both industry and academia are beginning to appreciate the value of these types of measurements and the use of High Precision Coulometry is growing within the Li-ion research community.
References
[1] A.J. Smith, H.M. Dahn, J.C. Burns, and J.R. Dahn, “Long-Term Low-Rate Cycling of LiCoO2/Graphite Li-Ion Cells at 55°C,” J. Electrochem. Soc., vol. 159, pp. A705–A710, 2012.
[2] A.J. Smith, J.C. Burns, D. Xiong, and J.R. Dahn, “Interpreting High Precision Coulometry Results on Li-ion Cells,” J. Electrochem. Soc., vol. 158, pp. A1136–A1142, 2011.
[3] A.J. Smith, J.C. Burns, and J.R. Dahn, “A High Precision Study of the Coulombic Efficiency of Li-Ion Batteries,” Electrochem. Solid-State Lett., vol. 13, pp. A177–A179, 2010.
[4] A.J. Smith, J.C. Burns, S. Trussler, and J.R. Dahn, “Precision Measurements of the Coulombic Efficiency of Lithium-Ion Batteries and of Electrode Materials for Lithium-Ion Batteries,” J. Electrochem. Soc., vol. 157, pp. A196–A202, 2010.
[5] T.M. Bond, J.C. Burns, D.A. Stevens, H.M. Dahn, and J.R. Dahn, “Improving Precision and Accuracy in Coulombic Efficiency Measurements of Li-ion Batteries,” J. Electrochem. Soc., vol. 160, pp. A521–A527, 2013.
[6] A.J. Smith, J.C. Burns, and J.R. Dahn, “High-Precision Differential Capacity Analysis of LiMn2O4/graphite Cells,” Electrochem. and Solid-State Lett., vol. 14, pp. A39–A41, 2011.
[7] Hannah M. Dahn, A.J. Smith, J.C. Burns, D.A. Stevens, and J.R. Dahn, “User-Friendly Differential Voltage Analysis Freeware for the Analysis of Degradation Mechanisms in Li-Ion Batteries,” J. Electrochem. Soc., vol. 159, pp. A1405–A1409, 2012.
[8] J.C. Burns, Adil Kassam, N.N. Sinha, L.E. Downie, Lucie Solnickova, B.M. Way, and J.R. Dahn, “Predicting and Extending the Lifetime of Li-Ion Batteries,” J. Electrochem. Soc., vol. 160, pp. A1451–A1456, 2013.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
