- Email: [email protected]
Low temperature aging mechanism identification and lithium deposition in a large format lithium iron phosphate battery for different charge profiles
Journal of Power Sources 286 (2015) 309e320
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
Low temperature aging mechanism identification and lithium deposition in a large format lithium iron phosphate battery for different charge profiles Minggao Ouyang a, *, Zhengyu Chu a, Languang Lu a, Jianqiu Li a, Xuebing Han a, Xuning Feng a, b, Guangming Liu a a b
State Key Laboratory of Automotive Safety and Energy, Tsinghua University, Beijing 100084, China Department of Naval Architecture and Marine Engineering, University of Michigan, Ann Arbor, MI 48109, USA
h i g h l i g h t s A turning point is found for the current rate and cut-off voltage limits for degradation when charging at low temperature. The process of lithium deposition is investigated by incremental capacity analysis. The aging mechanism is quantitatively identified through a mechanic model using the PSO algorithm.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 22 January 2015 Received in revised form 7 March 2015 Accepted 30 March 2015 Available online 31 March 2015
Charging procedures at low temperatures severely shorten the cycle life of lithium ion batteries due to lithium deposition on the negative electrode. In this paper, cycle life tests are conducted to reveal the influence of the charging current rate and the cut-off voltage limit on the aging mechanisms of a large format LiFePO4 battery at a low temperature (10 C). The capacity degradation rates accelerate rapidly after the charging current reaches 0.25 C or the cut-off voltage reaches 3.55 V. Therefore the scheduled current and voltage during low-temperature charging should be reconsidered to avoid capacity degradation. Lithium deposition contributes to low-temperature aging mechanisms, as something needle-like which might be deposited lithium is observed on the surface of the negative electrode after disassembling the aged battery cell. To confirm our explanation, incremental capacity analysis (ICA) is performed to identify the characteristics of the lithium deposition induced battery aging mechanisms. Furthermore, the aging mechanism is quantified using a mechanistic model, whose parameters are estimated with the particle swarm optimization algorithm (PSO). The loss of reversible lithium originating from secondary SEI formation and dead lithium is confirmed as the cause of the aging. © 2015 Elsevier B.V. All rights reserved.
Keywords: Lithium-ion battery Low-temperature aging Low-temperature charging Lithium deposition Incremental capacity analysis
1. Introduction Lithium ion batteries have become popular in the automobile industry due to their high energy and power density; however, capacity degradation in practical use restricts their broader application. Capacity degradation can be caused by multiple factors, including material properties, manufacturing techniques and practical operating conditions. The pervasively acknowledged aging mechanisms of lithium batteries are the loss of lithium ion (LLI),
* Corresponding author. E-mail address: [email protected] (M. Ouyang). http://dx.doi.org/10.1016/j.jpowsour.2015.03.178 0378-7753/© 2015 Elsevier B.V. All rights reserved.
the loss of electrode active material (LAM), and an increase in resistance (IIR) [1,2]. LLI occurs mainly on the surface of the anode due to SEI decomposition and regeneration. In terms of LAM, the loss of positive electrode material is largely responsible, although structural damage to the negative electrode of graphite induced by lithium intercalation and deintercalation in Ref. [3], is a second explanation for LAM. The increasing thickness of the SEI film causes IIR [4e6]. In previous studies, a tank model is proposed that illustrates the issue graphically and vividly [1]. Previous studies provide several enlightened approaches to analyzing the capacity degradation of lithium ion batteries. Some ex situ post-mortem technologies are widely utilized to investigate the fading of individual electrodes, such as X-ray diffraction (XRD)
310
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
and scanning electron microscopes (SEM) [7,8]. An in situ noninvasive method, impedance spectroscopy (IS), has also been recommended [9e11]. Additionally, for the semi-quantitative analysis of battery aging mechanisms, ICA and differential voltage analysis (DVA) have been used, as in Refs. [1,12e20]. To quantify the LLI and LAM, a prognostic/mechanistic model is introduced in Refs. [1,2,17e23]. By means of identifying the parameters of the model, the aging mechanism is directly clarified [1]. Batteries age far more at low temperatures than at room temperature [5,24]. It is reported that low-temperature degradation mainly occurs during the charging process due to lithium deposition, the potential for which is more likely to be achieved in the anode due to its elevated resistance at low temperatures [24,25]. S.S Zhang et al. [26] reported that even at a low current rate, the electrode potential at the anode is below 0 V vs. Li/Liþ over a half period of charging. The low potential at the anode leads to lithium deposition, which definitely damages the battery. Jiang Fan et al. studied the effects of different low-temperature voltage profiles on lithium ion batteries and suggested that lithium plating will occur at high-rate charging [25]. Low temperatures are unavoidable in practical use, however, although they are known to damage the battery. Therefore, the charging protocol should adapt to the low temperature environment to avoid possible capacity degradation. For instance, pulse heating can efficiently heat the battery to mitigate the effects of low temperatures [27]. Nevertheless, the polarization on the anode surface will be stronger with a higher rate of charge, which results in increased lithium deposition [25]. In addition, a high current rate can bring about cell capacity decay and impedance increase, whether through a pulse charge or a conventional charge [25]. Furthermore, battery degradation is aggravated by a higher charge cut-off voltage [28e30]. Hence, the low temperature charging must to be modified. Lithium ion batteries with a LiFePO4 cathode are ideal for low temperature capacity fading research because LiFePO4 is more stable than the other cathode materials [31,32]. We employ LiFePO4 as the battery cathode to avoid the degradation inherent in other cathodes and help us focus on studying the capacity degradation caused by lithium deposition at the anode. In this paper, cycle life tests are conducted to reveal the influence of the charging rate and the cut-off voltage limit on the aging mechanisms of a large format LiFePO4 battery at a low temperature (10 C). An experiment matrix was created to conduct a lowtemperature cycling test. Lithium deposition was observed using SEM after disassembling the battery post-aging. ICA is used to investigate the low temperature aging mechanism. A mechanistic/ prognostic model is employed to quantitatively reveal the LLI, LAM and IIR that occurred during low-temperature charging. Parameter changes in the mechanistic model help to quantify the degradation mechanisms, and we confirm that the main reason for low temperature aging is LLI due to lithium deposition at the anode. Based on the test results, when charging a LiFePO4 battery in a low temperature environment, here 10 C, the charge current rate should be restricted to less than 0.25 C and the cut-off voltage to less than 3.55 V.
2.2. Reference performance test A reference performance test (RPT), which comprises a capacity test, micro rate test (close to equilibrium) and hybrid pulse power characterization (HPPC) test, is developed to assess the battery's basic performance. The actual capacity is measured by two charge and discharge cycles, both at a rate of 1/3C with 1 h rest between each cycle. The low current profile of a 1/20 C charge and discharge rate is used to calculate the incremental capacity (IC) curves. The objective of the HPPC test is to measure the resistance of the charge and discharge at different states of charge (SOC) [33]. In consideration of the daily conditions of an electric vehicle, the time increment for discharge and regeneration is 30 s and 10 s, respectively. Derived from the manufacturer's maximum allowable discharge and charge current for 30 s and 10 s, the relative rates are 2 C and 1.5 C, respectively. An entire HPPC test consists of 10 cycles of this profile separated by 10% depth of discharge (DOD) CC 1/3C rate segments, followed by 1 h rest [33]. An entire RPT is shown in Fig. 1. Through such a test, the capacity, equilibrium characteristics and internal resistance can be evaluated. 2.3. Cycle life test at low temperature A cycle life test was performed at 10 C on 13 proposed cells under different conditions such as varied charge current rates, charge cut-off voltages and charge cut-off currents to analyze the aging mechanism when charging a lithium battery at a low temperature, as presented in Table 1. The discharge regime is consistent at 1/3 C current rate and 2.0 V cut-off voltage. The cycle life test consists of a total of 50 cycles for each cell using the constant current e constant voltage (CCeCV) charge protocol as illustrated in Table 1. Prior to the cycling tests, a primary characterization test was conducted to measure the cells' basic features. The cycle life test was separated by the RPT at an ambient temperature to record the cells' feature at 10 cycle intervals. The cycle life test and RPT were repeated after a rest period of 10 h to ensure the experimental temperature. It should be noted that before the final RPT was conducted, the cells remained at the ambient temperature over 10 days due to a breakdown in the environmental chamber. 2.4. Battery disassembling and half battery assembling Cell disassembly aimed at investigating the electrode dynamism through half cells and analysis by means of SEM. Before opening, cells were utterly discharged by CC discharge at a rate of 1/3 C and then by CV discharge to a cut-off voltage of 1/20 C at 2.0 V to reduce
2. Experiment 2.1. Commercial lithium-ion battery and test equipment This paper utilizes a commercial large format LiFePO4/graphite lithium ion battery with a nominal capacity of 11.5 Ah. Isothermal experiments were performed on the batteries in an environmental chamber (DONGGUAN BELL) using a multichannel bench testing system (Neware CT-4008).
Fig. 1. Reference performance test protocol at room temperature.
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320 Table 1 Cycle life test methodology at low temperature. Cell number
Charging cut-off voltage/V
Charging cut-off current/A
Charging rate/A
Cell Cell Cell Cell Cell Cell Cell Cell Cell Cell Cell Cell Cell
3.29 3.55 3.65 3.65 3.65 3.65 3.65 3.65 3.65 3.65 3.5 3.58 3.62
1/10C 1/10C 1/2C (No CV period) 1/20C 1/10C 1/10C 1/10C 1/10C 1/10C 1/10C 1/10C 1/10C 1/10C
1/2C 1/2C 1/2C 1/2C 1/5C 1/3C 1/2C 1/10C 0.25C 0.3C 1/3C 1/3C 1/3C
1 2 3 4 5 6 7 8 9 10 11 12 13
the contact between the lithium in graphite and the air. Fresh cells were opened and then assembled in a glove-box under an argon atmosphere. After a rest period of 12 h, the assembled coin cells were tested on the battery testing system Neware BTS under low current. Cells with different aging phases were opened to observe the changes in the electrodes, the anode in particular. These details will be elaborated upon in 3.1.4. 3. Results and discussion 3.1. Impact of different parameter values of charge protocols on battery characteristics 3.1.1. Capacity fade The impact on capacity fade is compared under different experimental conditions, as represented in Fig. 2. In Fig. 2(a), the effects of different cut-off voltage limits are examined. The capacity has different rates of decrease. The degradation accelerated rapidly as the cut-off voltage increased. Therefore, limiting the cut-off voltage has a mitigating effect on degradation; however, the charge capacity in cycling will diminish as the cut-off voltage is reduced. The cells' RPT indicates that the internal resistance at 50% SOC is 30.49 mU and 10.23 mU at 10 C and 25 C, respectively. This means that at a low temperature, the cut-off voltage is reached in less time at the same current rate. As an example, average charging capacity of Cell 7 maintained only 8.94 Ah, which is 75.54% of the nominal capacity. The capacity evolution is illustrated in Fig. 2(b) for charging at various current rates from 0.1 C to 0.5 C. It is evident that the capacity degradation accelerates as the rate increases. The probable
311
reason is related to the behavior of lithium plating at high rates [34,35]. The maximum designed current rate in this experiment is 0.5 C; however, the capacity fade may decrease as the rate continue to ascend to a higher level, such as 1 C due to elevated temperatures induced by joule heat or a shorter charging period [24]. The regeneration of capacity for the 50th cycle is definitely caused by a certain amount of rest time. The cells degrade non-linearly as the current rate and cut-off voltage in charging increase. There exists a turning point in the capacity retention for both charging parameters, indicating a distinct degradation once the magnitude of the parameters exceeds a certain level. The loss rate of capacity after 40 cycles at various current rates and cut-off voltages of charging is shown in Fig. 3. This indicates an evident turning point for each parameter correlated to capacity degradation, which is approximately at a rate of 0.25 C and 3.55 V. Once the charging parameter exceeds the turning point, capacity fade will be exacerbated significantly. Limited parameters based on the consideration of aging within a secure range are significant for suppressing degradation. The influence of the variation of cut-off currents at 3.65 V on the capacity fade is further investigated; results are shown in Table 2. Through changes in the cut-off current value, the lasting period of the CV stage is controlled. Accelerated degradation occurs at the time of the CV float charging at the same level as the cut-off voltage increases. Compared with Cells 3 and 4, over 85% of the capacity loss originates in CV charging. The charging time and capacity evolution for Cell 4 at 10 cycle intervals are represented in Fig. 4. The CC charging time and proportion simultaneously decrease with less charging time and the capacity to make the CV stage become the major part gradually. In spite of this, the impact of the CV stage depends on the charging condition of the previous CC stage. One recent study suggests that if the charging rate of the CC stage is restricted within a specific harmless range, the subsequent CV charging will not aggravate the situation [36]. Consequently, the degradation initiated within the CC stage. 3.1.2. Increase in internal resistance Battery resistance comprises polarization and ohmic resistance. The battery voltage comes to an abrupt change while undergoing current mutation. The response time of ohmic resistance is very fast, almost instantaneous, followed by a polarization response after a few seconds. In this paper we consider a 1 s response resistance as an ohmic unit and 30 s (discharge) or 10 s (charge) as the total internal resistance of the battery. The polarization resistance comes from the difference between the total resistance and the ohmic resistance. Battery resistance at each SOC was calculated from the RPT test.
Fig. 2. Evolution of standard test discharge capacity at intervals of 10 cycles: (a) charging cut-off voltage; (b) charge current rate.
312
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
Fig. 3. Turning points of capacity loss rate after 40 cycles for different charging parameters: (a) charging rate; (b) cut-off voltage of charging.
Table 2 Effect of CV charge period at 3.65 V on the degradation condition. Cell number
Cut-off current at CV stage
Original capacity/Ah
50 cycle capacity/Ah
Degradation rate
Cell 3 Cell 7 Cell 4
1/2C 1/10C 1/20C
11.80 11.82 11.86
11.32 9.58 8.63
4.07% 18.95% 27.23%
Fig. 5 displays 30 s discharge resistance changes for Cell 7. The resistance shows an apparent increase of approximately 5%e7% with cycles. The increase of SEI thickness across cycles plays an important role, albeit not an exclusive one, in the increase of battery resistance [37]. Fig. 6 displays the evolution of the resistance and capacity. It should be noted that the internal resistance and capacity loss decrease after the 5th ten-cycle group compared with the 4th. After the 50th cycle, the cells were stored at room temperature for sufficient rest, which made the cells recover in performance corresponding with a rise in capacity. This is an indication that the lost capacity is not immediately irreversible; rather, it can be regenerated by sufficient rest. Generally, capacity recovery is involved in the decrease of resistance, implying a correspondence between the two characteristics.
profiles at intervals of 10 cycles are presented to reveal the cells' evolution. Profiles of Cells 4 and 7 show not only a reduction in charge capacity but also evident changes in shape and position. They vary noticeably between the first 20 cycles and immediately
3.1.3. Charge profile evolution Fig. 7 presents the voltage profiles with cycling at 10 C. It is important to note that the profiles do not change continuously with the execution of standard tests; here, each first cycle after the standard test was discharged at room temperature. Therefore, the
Fig. 5. 30 s discharge resistance evolution at different SOC from the RPT for Cell 7.
Fig. 4. Evolution of CC charge time and charge capacity with cycles of the Cell 4.
Fig. 6. Changes in resistance at specific SOC compared with capacity retention across cycles for Cell 7.
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
following 30 cycles. Under 3.5 V, the charge profiles have a reduced voltage with an elevated trend up to 3.5 V. This leads to reaching the cut-off voltage in a very short period of time, resulting in a decrease in the available capacity. An increase in the charge internal resistance may have been a contributing factor. In contrast, cells showing a slight capacity loss exhibited less change in terms of the charge voltage, the shape of which remained stable upon cycling. The charge plateau was also elevated due to the rise in resistance, as in cells 5 and 6. 3.1.4. Cell disassembling examination Cell disassembling is a constructive and direct methodology to observe change in cells whereby three cells of varying degreesea fresh cell, Cells 2 and 7ewere examined. Before the operation, cells are fully discharged by the CC procedure at a 1/3 C current rate with a subsequent CV procedure at a 1/20 C cut-off current rate. After opening the cells, the current collectors are covered with anode and cathode electrodes and the separator samples are cut out. The cathode electrode appears almost unchanged upon a macroscopic check; the anode electrode, in contrast, changes significantly. The anode samples from the different cells after resting in the air for a certain period are shown in Fig. 8. The anode surface of the fresh cell is colored gray, the color of delithiated graphite [38], as shown in Fig. 8(a). The anode surface of Cell 2 is dark blue, but a few white dots are visible in Fig. 8(b). Fig. 8(c) depicts severe degradation in the form of a multitude of white dots on the anode surface of Cell 7; the dots are considered to be oxide products of the active substance in contact with the atmosphere. The color and morphology of the fresh cell and Cell 2 remain relatively steady in the air. In contrast, the anode changes in color and surface morphology due to exposure in the air. The details are
313
shown in Fig. 9. At the instant the cell is opened, denoted as 0 s, a majority of the anode appears dark yellow. In the subsequent 20 s, the yellow portion gradually becomes black. After approximately 80 s, some parts of the surface turn white so that concentrated white dots appear in a bubble formation and escape. The phenomenon suggests lithium plating in the anode. The above discharging procedure is not capable of rendering graphite in a complete state of deintercalation. By inference, the lithium layer on the surface of the graphite remains as the actual part of the anode that oxidizes during exposure. The details are discussed in Section 3.2.2. The above macroscopic changes reveal the strong reactions of the surface layer lithium and graphite intercalation with the atmosphere. In Fig. 10, the lithiated graphite appears as red or gold [38] together with the lithium layer, and the SEI film on the surface is dark yellow. Therefore, the silver metallic luster of the lithium does not appear. Lithium first reacts with N2 and then produces black Li3N, after which LiOH is generated by Li3N in contact with moisture in the air, with NH3 yielding. Finally, LiOH converts to Li2CO3 through the process of slow oxygenation [39], which is the origin of the aforementioned white dots. The microstructure of the anode and separator from the cells are analyzed by means of SEM. An ultrathin conductive coating was deposited on the surface of the samples by an ETD-800 automatic sputter coater from Elaborate Technology Development, Ltd. The SEM is a JSM-7401F of JEOL Co., Ltd. Electron micrographs of the separator and anode are shown in Fig. 10. Separator images of the fresh cell, Cells 2 and 7 are depicted in Fig. 10(a)e(c). The porous separator is gradually blocked by the products of side reactions, rendering Liþ more difficult to transport and leading to an increase in ohmic resistance. In Fig. 10(d)e(f), the
Fig. 7. Comparison of the charging profiles at low temperature among cells with different degrees of degradation: (a) and (b) represent serious degradation; (c) and (d) represent slight degradation; (a) Cell 4; (b) Cell 7; (c) Cell 6; (d) Cell 5.
314
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
Fig. 8. Photographs of the negative electrode from cells at different degrees of degradation: (a) fresh cell; (b) Cell 2 after the cycle test; (c) Cell 7 after the cycle test.
Fig. 9. Changes in the color and morphology of Cell 7's negative electrode surface in contact with the atmosphere from 0 s to 80 s.
SEM images of the anode surface are presented. The particle morphology of the graphite is intensely distinct in Fig. 10(d). As the capacity loss increases, something needle-like which might be metal lithium deposition are visible on the surface of particles, as in Fig. 10(e); images at a higher magnification are shown in Fig. 11. Comparatively speaking, in Fig. 10(f), the particulate graphite disappears. Instead, a film with holes of various sizes covers almost the entire surface area of the negative electrode. It can be deduced that the film consists of the primary SEI film and secondary SEI produced by the deposited lithium [40]. The holes are proved to be created by the bubbles that are the reaction product of the lithium in the air. The lithium deposition is creating a serious situation. At a low temperature, lithium is more likely to be deposited due to a decrease in the solid diffusion coefficient and low anode potential [26,41,42]. The morphology is determined by the current rate. Needle-like lithium deposition at high current rate may break the SEI film due to stress. As a result, the lithium grows out continuously [43]. Needle-like deposition is long and slender with branches and bending as shown in Fig. 11. In Fig. 11(a), deposition needles of different lengths and diameters are indicated by arrows. Needle-like deposition grows on the lithium substrate through the SEI film, as indicated in Fig. 11(b). Therefore, they lie down on the surface and may lack the mechanical strength to perforate the separator and resulting in an internal short-circuit. The dendrite lithium, however, may crack where it is fragile and isolate itself
from the lithium substrate, leading to a transition to the dead lithium [40]. Mechanisms of lithium deposition at room temperature and low temperature are different. Lithium deposition is generally caused by two dominant factors: 1) High rate charging leads to high lithium diffusion rate, and lithium deposition occurs on the surface of anode when the lithium formation rate exceeds its intercalation rate [44]; 2) Lithium deposition is more prone to happen at anode under local high over-potential, which is determined by the product of charge current and resistance at anode [26]. At room temperature, the lithium diffusion rate and the over-potential can be elevated by high current rate during rapid charging. However, at low temperature, the lithium diffusion rate decreases and resistance of anode increases [25,45]. Therefore, the condition of lithium deposition is stricter at low temperature than that at room temperature with same current rate. 3.2. Incremental capacity analysis of the aging mechanism at a low temperature In Section 3.1, the evolution of the capacity, resistance and charging curves are discussed and the SEM images of separators and anodes are analyzed. In addition, quantitative analyses must to be conducted for reliability. It is reported that the IC curve, i.e. dQ/ dV-V, was used to deduce cell aging mechanisms in several studies
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
315
Fig. 10. SEM images of separators and graphite electrodes from cells at different degrees of degradation caused by various cut-off voltages: (a) separator of the fresh cell; (b) separator of Cell 2; (c) separator of Cell 7; (d) graphite electrode of the fresh cell; (e) graphite electrode of Cell 2 (f) graphite electrode of Cell 7.
Fig. 11. SEM images of the needle-like deposition which might be lithium on the surface of the negative electrode Cell 5.
[1,13,37,46,47]. In this section, degradation mechanisms are investigated by the semi-quantitative means of ICA. ICA is a process to reduce the order of the voltage profile, reflecting the electrochemical procedure evolution more specifically during charging [12]. Peaks in the IC (dQ/dV) curve reveal the phase transitions and the magnitude, position, and shape, which reflect electrochemical dynamism [1,12,47]. 3.2.1. IC curves analysis using micro currents at room temperature Each peak in an IC curve is correlated with a specific phase transition, of which the magnitude, shape and position are important indicators [13]. Every standard test at room temperature containing a 1/20 C rate charge process offers close-to-equilibrium conditions to reduce the polarization impact for the purpose of ICA. Here, we derive dQ/dV curve using the probability density function (PDF) method referred to previous studies [16]. The IC curve is then smoothed. Fig. 12 displays the IC curves of the anode (negative electrode) and cathode (positive electrode). Different peaks in the IC curve of the anode manifest corresponding to different phases of the
transition process. The evolution in the concentration of intercalation lithium ultimately accounts for multiple steps of the phase transitions in the graphite negative electrode. It is reported that there are three voltage plateaus at 210 mV, 120 mV, 85 mV in graphite, corresponding to x from 0.08 to 1, where x appears in LixC6 [48]. Further, reactions in this range are divided into four regions in which different stages of lithium-graphite intercalation compounds exist with phase transitions from a higher stage to a lower stage in reaction. It should be noted that the voltage regions of peaks ②, ③ and ⑤ are slightly lower than those previous reported literature due to polarization. In Fig. 13, IC curves of Cells 5 and 7 are compared after different low temperature cycles, respectively. 4 peaks are distinctly observed and denoted as ①, ②, ③ and ⑤, corresponding to four phase transition processes in the anode. Peak ④ cannot be explicitly identified. As for the cathode, it has a FePO4eLiFePO4 two phase transition region offering a long potential plateau as the reference voltage during charge and discharge, which is denoted as Ⅱ. Therefore, four peaks are defined as ①*Ⅱ, ②*Ⅱ, ③*Ⅱ and ⑤*Ⅱ, as suggested in the literature [47].
316
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
Fig. 12. IC curves of the positive and negative electrode: (a) LFP positive electrode; (b) graphite negative electrode.
During 50 cycles at low temperature, the first peak in the IC curve of Cell 7, denoted as ①*Ⅱ, changed in density and position but not in shape, as shown in Fig. 13(a). The ①*Ⅱ peak density of Cell 7 decayed significantly until it disappeared, while the other three peaks remained stable in density in the preceding 40 cycles. Specifically, the density and position of the 50th item returning to the previous state are noticeably reflected. Deviation at a higher voltage occurs in this position, which suggests an increase in the battery resistance, as shown in previous HPPC results in Section 3.1.2. Changes of peaks along with HPPC results and SEM images indicate that the capacity decay originated in LLI from lithium deposition and that the thickness of the SEI film increased due to the reaction between the active deposited lithium and electrolytes, contributing to the raised battery resistance. First, the mechanism of capacity fade is LLI, otherwise, the peak magnitudes in Fig. 13 will change. Second, LLI comes from the lithium deposition on the surface of the negative electrode due to observable lithium dendrites in Fig. 11. Third, insoluble products, which comes from the reaction between the deposited lithium and electrolyte enhance the thickness of the SEI film and the internal resistance inferred from the HPPC test shown in Fig. 5. As Fig. 13(a) shows, peak ①*Ⅱ density of Cell 7 decays to a greater extent because the capacity degradation is more significant. The dominant reaction of the third stage in graphite is as follows [48],
LiC6 $C6 þ Li#2LiC6 Accompanied with the reduction in lithium ion, the phase transition process gradually decreases in number, making the
lithium-graphite compound at the end of discharge exist in the form of the second-stage phase, i.e., LiC12 [38]. 3.2.2. IC curves analysis using charge profiles at low temperature In 3.2.1, peaks featuring the evolution of the IC curve at a low charge rate (1/20 C) were used to elucidate the aging mechanisms of the LFP/graphite battery cycle at a low temperature, which offers a close-to-equilibrium process during charging and therefore renders each peak an accurate and distinct presence. As the rate increases, the polarization for charge and discharge retains a greater influence on the peaks, even forcing some small peaks to merge. However, the analysis is valid within a specific rate range. In our previous study [1], we performed an explicit analysis on a 1/3 C charge profile to identify the on-board aging mechanism in a battery management system (BMS). Here, it is the IC curve that is employed to explain the evolution of the charge curve at a low temperature in different cycles. Fig. 14 shows the difference between the 1st cycle immediately after the room temperature standard test and the subsequent cycles at a low temperature. The difference can be explained as more quantities of charge capacity into the battery in the 1st cycle after a full discharge at room temperature. The shape of the 1st charge curve is close to 2-10th cycles for Cell 3 because of less degradation as shown in Fig. 14(c). Deviation in the shape of the 1st charge curve from that of the 2-10th cycle for Cells 4 and 7 indicates more degradation. The shape of 2-10th charge curves has significant deviation from the 1st charge curve at low temperature cycling, as shown in Fig. 14(a) and (b). The IC curves of the different cycles of Cell 7 compared with the
Fig. 13. IC curves from the micro current test after various cycles of Cells 7 and 5: (a) Cell 7; (b) Cell 5.
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
317
Fig. 14. Low temperature charge curve from the 1st cycle to the 10th cycle: (a) Cell 4; (b) Cell 7; (c) Cell 3; (d) Cell 1.
positive electrode are displayed in Fig. 15. In particular, IC curves are calculated through non dimensional computing, which is convenient for comparing curves among the different capacity scales. In Fig. 15, the red line is the IC curve of the 1st cycle and the cyanic line is the 10th cycle. The green solid and dashed lines are the positive electrode IC curve. Compared with the 1st cycle, peak ③*Ⅱ decreased in cycle 10 and peak ②*Ⅱ became predominant. On the one hand, at a low temperature, the charge and discharge capacity hold at only 70% of their room temperature capacity. The charge capacity of the 1st cycle is 10.81 Ah with 8.95 Ah discharged, whereas a capacity of 9.17 Ah (only 84.83% of the 1st cycle) is charged in the 2nd cycle. The following capacity of discharge is
slightly less than that of the charge with a coulombic efficiency of 98e99%. Therefore, the discrepancy of the SOC range is probably responsible for the charge abnormality in the 1st cycle. The partial discharge in the 1st cycle accounts for the initial offset of the SOC charge to a higher point and the resulting reduction in the magnitude of peak ③*Ⅱ. On the other hand, the shape of the IC curve after the 1st cycle is similar to that of the LFP half-cell after being translated to a higher potential, considering the negative electrode potential and the polarization factor. In other words, lithium deposition on the anode surface after the 1st charge process changes the electrochemical dynamism of the electrodeelectrolyte interface; namely, the lithium ion intercalation into metal lithium increased, but less into the graphite layer to form a graphite compound. This validates the predominant aging mechanism mentioned in 3.2.1. Moreover, the relative magnitude of the characteristic peaks could be a new evolution criterion to detect lithium deposits, which is a potential application of BMS. Herein lies in an additional problem concerning the drastic changes in position in the cycles for Cells 4 and 7. Fig. 16 illustrates the evolution of the IC curves at every 10th cycle of Cell 4. The shape and magnitude of the characteristic peaks do not change significantly, but the positions of peaks 1 and 2 gradually move to a higher potential. This indicates that the rise in battery resistance, which is mainly due to the increase in the thickness of the SEI film, is a key reason behind this change. 3.3. Identification of aging parameters based on prognostic model using particle swarm optimization algorithm
Fig. 15. The evolution of the low temperature cycle of the 1st cycle and the 10th cycle compared with the positive electrode IC curve (solid blue line for the original curve and dashed blue line for the right move curve). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
According to our previous research [1], the method for reconstructing the CC charging voltage curve can be used to quantitatively identify the evolution of the capacity of the two electrodes,
318
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
the quantities of lithium ion and the impedance value. Based on the half electrode charge voltage curve, the LFP/ graphite potential versus Li/Liþ is a uniform function with the number of transferred lithium-ion. Therefore, it is possible to reconstruct the full cell charge voltage curve if the transferred lithium-ion is evaluated. The model is described by Equations (1)e(3):
x ¼ x0 þ
Q Cn
(1)
y ¼ y0
Q Cp
(2)
Vc ¼ Up ðyÞ Un ðxÞ þ IR
(3)
The equations are referred to our previous study [1]. Cn and Cp are estimated to be the capacity of the anode and cathode electrode, respectively. In addition, x0 and y0 represent the initial number value of the lithium ion insertion in the two electrodes, respectively. R is the battery internal resistance as a whole. Here, the parameter Q is the charge capacity at the moment when I is charging current value; both can be measured. It should be noted that the simulated x and y do not represent the real SOC, only the relative values for the evolution of lithium ion amount. Therefore, five total parameters of x0, y0, Cn, Cp and R will be estimated to minimize the cost function as Equation (4):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Ptn 2 t¼t1 ðVm ðtÞ Vc ðtÞÞ RMSE ¼ n
(4)
The cost function is actually the root mean squared error (RMSE) between the measured and calculated voltage. Here, the PSO algorithm is used to determine the optimal parameters. Fig. 17 presents a comparison between the measured and simulated results of Cell 7 from 6 standard tests after every 10 cycles. Only the CC charge section is used for the simulation because CV charge is invalid. The satisfied overlap indicates not only the appropriateness of the method but also the accuracy of the simulation. Based on the above analysis, we can find the effect of the different charge parameters on the cell degradation. It is necessary to further analyze the aging mechanisms at a low temperature. Although the mechanisms of capacity decay and battery aging vary, they can be divided into problems of the cathode electrode, anode
Fig. 16. Evolution of the IC curves at a 10 cycle intervals.
Fig. 17. Charge curve fitting results (green square) vs. experiments profiles (blue line) for Cell 7 after every 10 cycles. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
electrode and electrolytes. A rise in battery impedance may be accounted for by continuous SEI growth due to electrolyte decomposition [37], which may in turn cause capacity fade. In conclusion, changes in the two electrodes and impedance allow a quantitative analysis of aging. Fig. 18 displays the results of the evolution of the identified parameters with cycling. Before the cycle test, the identified original capacity of the negative electrode is approximately 20% beyond the positive, as shown in the figure, which is designed to prevent lithium deposition resulting from overcharging. The capacity of each electrode is higher than the nominal capacity of the entire battery, expressing that only part of the active material is available. After the tests, the relatively unchanged Cp and Cn show no apparent loss of active material (LAM) for both electrodes, and x0 also remains constant. The total lithium number (Cp*y0þCn*x0), however, reveals a clear decrease in Fig. 19, indicating the presence of LLI. The lithium number of Cell 5 at 1/5 C remains comparatively unchanged. It is concluded that the decrease in the rechargeable capacity is the result of increasing irreversible lithium ions. The RMSE between voltages estimated and derived from experiments is within 10 mV. The battery average resistance gradually increased, which verifies the resistance analysis found in the results of the previous HPPC experiment. Therefore, it is deduced from the analysis that the low temperature degradation of LFP/graphite is mainly due to LLI induced by lithium deposition on the surface of anode electrode. The solid diffusion coefficient in the anode, which is universally considered as a rate-limit factor, drops more severely than electrolyte diffusion. As a result, it becomes more difficult for the lithium ion to intercalate into the anode, accumulating on the surface of the graphite. Once it reaches lithium deposition potential level, the lithium metal is plated. It should be noted that the shape/dimension of the large format cell will influence the lithium deposition at low temperature during cycling. The lithium deposition at specific positions within the cell depends on the local current density and temperature. The distribution of current density and temperature depends on shape/ dimension of the large format cell [49]. Therefore the lithium deposition phenomenon within large format cells with different shapes/dimensions can be different, which needs further investigation. However, the main mechanism of the capacity degradation during low temperature cycling is similar as discussed in this paper. It should be also noted that lithium deposition process proceeds
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
319
Fig. 18. Evolution of battery parameters fitting results after 50 cycles: (a) x0; (b) y0; (c) capacity of the negative electrode; (d) capacity of the positive electrode; (e) average resistance; (f) RMSE.
inside the SEI due to its insulation property. As the metallic state, once the lithium makes contact with the electrolyte due to the collapse of the primary SEI, the secondary SEI film will form. Therefore, lithium is prevented from reacting further with the electrolyte by the compact SEI film. If the potential increases again, oxidation and electron transfer will occur on the surface, resulting in lithium dissolution and capacity recovery [43]. This explains the phenomenon that the battery capacity increases post-degradation after resting at room temperature for 10 days. If it is deposited as dendritic lithium on the surface, however, the SEI film may be damaged such that active metal lithium forms undissolved inorganics by reacting with the electrolyte, leading to irreversible loss [30,41,43,50]. The inorganics will also increase the thickness of SEI, causing an increase in the battery resistance, as Fig. 20 represents [51]. Dendritic lithium will also affect the security of the battery by perforating the separator if sufficient mechanical strength is achieved [40].
Fig. 19. Evolution of lithium numbers in the battery with cycles at different charge rate.
4. Conclusion Low temperature cycle life experiments were performed at 10 C, and quantitative methods were used to identify the LFP battery aging mechanism. Capacity fade was more severe with a higher charging rate and higher cut-off voltage, and the turning points for the current rate and cut-off voltage after which degradation accelerated were determined to be approximately 0.25 C and 3.55 V. After dissembling the batteries with different cut-off
Fig. 20. Schematic of lithium deposition on the surface of the graphite negative electrode with cycles at a low temperature. The permanent loss of lithium stems from the secondary SEI formation and dead lithium.
320
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
voltages, something needle-like which might be deposited lithium was observed and the morphologies of the anode with different degrees of degradation were compared with the SEM images. Subsequently, the degradation was examined by ICA using low current charging profiles at room temperature and cycle charging profiles at a low temperature. This revealed that the mechanism is the loss of lithium rather than loss of active electrode material. The loss of lithium stems from the lithium deposition on the surface at the anode. The electrochemical behavior of the anode was altered by the process of lithium deposition, rendering the graphite anode similar to the lithium metal anode. Finally, a prognostic and mechanic model was used to quantitatively analyze the aging mechanisms. The decline in numbers of reversible lithium was quantitatively illustrated, proving that the loss of capacity results from the loss of lithium. The deposited lithium in contact with the electrolyte to form a secondary SEI film and dead lithium from the lithium dendrite are supposed to be the two origins of the LLI. Acknowledgment This research is funded by the MOST (Ministry of Science and Technology) of China under the contract of No. 2014DFG71590 and Beijing science and technology plan No. Z121100007912001 and the National Support Plan 2013BAG16B01 and MOE (Ministry of Education) of China under the contract of No. 2012DFA81190. References [1] X. Han, M. Ouyang, L. Lu, J. Li, Y. Zheng, Z. Li, J. Power Sources 251 (2014) 38e54. [2] M. Dubarry, C. Truchot, B.Y. Liaw, J. Power Sources 219 (2012) 204e216. [3] V.A. Sethuraman, L.J. Hardwick, V. Srinivasan, R. Kostecki, J. Power Sources 195 (2010) 3655e3660. [4] K. Amine, J. Liu, I. Belharouak, Electrochem Commun. 7 (2005) 669e673. [5] N. Omar, M.A. Monem, Y. Firouz, J. Salminen, J. Smekens, O. Hegazy, H. Gaulous, G. Mulder, P. Van den Bossche, T. Coosemans, Appl. Energ. 113 (2014) 1575e1585. [6] H. Lin, D. Chua, M. Salomon, H.C. Shiao, M. Hendrickson, E. Plichta, S. Slane, Electrochem. Solid-State Lett. 4 (2001) A71eA73. [7] J. Li, E. Murphy, J. Winnick, P.A. Kohl, J. Power Sources 102 (2001) 294e301. [8] L. Yang, X. Cheng, Y. Ma, S. Lou, Y. Cui, T. Guan, G. Yin, J. Electrochem Soc. 160 (2013) A2093eA2099. [9] A.N. Jansen, D.W. Dees, D.P. Abraham, K. Amine, G.L. Henriksen, J. Power Sources 174 (2007) 373e379. €ltzsch, O. Kanoun, H. Tr€ [10] U. Tro ankler, Electrochim. Acta 51 (2006) 1664e1672. [11] S.S. Zhang, K. Xu, T.R. Jow, Electrochim. Acta 49 (2004) 1057e1061. [12] M. Dubarry, V. Svoboda, R. Hwu, B.Y. Liaw, Electrochem. solid-state Lett. 9 (2006) A454eA457. [13] M. Dubarry, B.Y. Liaw, J. Power Sources 194 (2009) 541e549. [14] A.H. Thompson, J. Electrochem Soc. 126 (1979) 608e616. [15] H.M. Dahn, A.J. Smith, J.C. Burns, D.A. Stevens, J.R. Dahn, J. Electrochem Soc.
159 (2012) A1405eA1409. [16] X. Feng, J. Li, M. Ouyang, L. Lu, J. Li, X. He, J. Power Sources 232 (2013) 209e218. [17] J. Remmlinger, M. Buchholz, T. Soczka-Guth, K. Dietmayer, J. Power Sources 239 (2013) 689e695. [18] K. Honkura, K. Takahashi, T. Horiba, J. Power Sources 196 (2011) 10141e10147. [19] I. Bloom, A.N. Jansen, D.P. Abraham, J. Knuth, S.A. Jones, V.S. Battaglia, G.L. Henriksen, J. Power Sources 139 (2005) 295e303. [20] J. Kim, N.P. Pieczonka, Z. Li, Y. Wu, S. Harris, B.R. Powell, Electrochim. Acta 90 (2013) 556e562. [21] J. Christensen, J. Newman, J. Electrochem Soc. 150 (2003) A1416eA1420. [22] J. Christensen, J. Newman, J. Electrochem Soc. 152 (2005) A818eA829. [23] X. Feng, J. Sun, M. Ouyang, X. He, L. Lu, X. Han, M. Fang, H. Peng, J. Power Sources 272 (2014) 457e467. [24] S. Tippmann, D. Walper, L. Balboa, B. Spier, W.G. Bessler, J. Power Sources 252 (2014) 305e316. [25] J. Fan, S. Tan, J. Electrochem Soc. 153 (2006) A1081eA1092. [26] S.S. Zhang, K. Xu, T.R. Jow, J. Power Sources 160 (2006) 1349e1354. [27] X. Zhao, G. Zhang, L. Yang, J. Qiang, Z. Chen, J. Therm. Analysis Calorim. 104 (2011) 561e567. [28] O.J. Kwon, W. Fang, C. Wang, Meeting Abstracts, MA2010-03, 2010, p. 279. [29] O.J. Kwon, C. Wang, Meeting Abstracts, MA2009-02, 2009, p. 544. [30] S. Tippmann, D. Walper, L. Balboa, B. Spier, W.G. Bessler, J. Power Sources 252 (2014) 305e316. [31] W. Zhang, J. Power Sources 196 (2011) 2962e2970. [32] F. Gao, Z. Tang, Electrochim. Acta 53 (2008) 5071e5075. [33] Freedom C.S. Department of Energy (October 2003) DOE/ID-11069, (2003). [34] N. Omar, M.A. Monem, Y. Firouz, J. Salminen, J. Smekens, O. Hegazy, H. Gaulous, G. Mulder, P. Van den Bossche, T. Coosemans, Appl. Energ. 113 (2014) 1575e1585. [35] M. Broussely, P. Biensan, F. Bonhomme, P. Blanchard, S. Herreyre, K. Nechev, R.J. Staniewicz, J. Power Sources 146 (2005) 90e96. [36] M. Wu, P.J. Chiang, J. Lin, J. Electrochem Soc. 152 (2005) A47eA52. € ller, J.O. Besenhard, M. Winter, [37] J. Vetter, P. Nov ak, M.R. Wagner, C. Veit, K.C. Mo M. Wohlfahrt-Mehrens, C. Vogler, A. Hammouche, J. Power Sources 147 (2005) 269e281. [38] S.J. Harris, A. Timmons, D.R. Baker, C. Monroe, Chem. Phys. Lett. 485 (2010) 265e274. [39] H. Osborg, Trans. Electrochem. Soc. 66 (1934) 91e105. pez, J.T. Vaughey, D.W. Dees, J. Electrochem Soc. 159 (2012) [40] C.M. Lo A873eA886. [41] Z. Li, J. Huang, B.Y. Liaw, V. Metzler, J. Zhang, J. Power Sources 254 (2014) 168e182. [42] C.K. Huang, J.S. Sakamoto, J. Wolfenstine, S. Surampudi, J. Electrochem Soc. 147 (2000) 2893e2896. [43] J. Yamaki, S. Tobishima, Rechargeable Lithium Anodes, Wiley-VCH Verlag GmbH & Co. KGaA, 2011, pp. 377e404. [44] B.K. Purushothaman, U. Landau, J. Electrochem Soc. 153 (2006) A533eA542. [45] A.N. Jansen, D.W. Dees, D.P. Abraham, K. Amine, G.L. Henriksen, J. Power Sources 174 (2007) 373e379. [46] M. Dubarry, V. Svoboda, R. Hwu, B.Y. Liaw, J. Power Sources 165 (2007) 566e572. [47] M. Dubarry, B.Y. Liaw, M. Chen, S. Chyan, K. Han, W. Sie, S. Wu, J. Power Sources 196 (2011) 3420e3425. [48] T. Ohzuku, Y. Iwakoshi, K. Sawai, J. Electrochem Soc. 140 (1993) 2490e2498. [49] U.S. Kim, C.B. Shin, C. Kim, J. Power Sources 189 (2009) 841e846. [50] G. Zheng, S.W. Lee, Z. Liang, H. Lee, K. Yan, H. Yao, H. Wang, W. Li, S. Chu, Y. Cui, Nat. Nanotechnol. 9 (2014) 618e623. [51] J. Christensen, J. Newman, J. Electrochem Soc. 150 (2003) A1416eA1420.
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
Low temperature aging mechanism identification and lithium deposition in a large format lithium iron phosphate battery for different charge profiles Minggao Ouyang a, *, Zhengyu Chu a, Languang Lu a, Jianqiu Li a, Xuebing Han a, Xuning Feng a, b, Guangming Liu a a b
State Key Laboratory of Automotive Safety and Energy, Tsinghua University, Beijing 100084, China Department of Naval Architecture and Marine Engineering, University of Michigan, Ann Arbor, MI 48109, USA
h i g h l i g h t s A turning point is found for the current rate and cut-off voltage limits for degradation when charging at low temperature. The process of lithium deposition is investigated by incremental capacity analysis. The aging mechanism is quantitatively identified through a mechanic model using the PSO algorithm.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 22 January 2015 Received in revised form 7 March 2015 Accepted 30 March 2015 Available online 31 March 2015
Charging procedures at low temperatures severely shorten the cycle life of lithium ion batteries due to lithium deposition on the negative electrode. In this paper, cycle life tests are conducted to reveal the influence of the charging current rate and the cut-off voltage limit on the aging mechanisms of a large format LiFePO4 battery at a low temperature (10 C). The capacity degradation rates accelerate rapidly after the charging current reaches 0.25 C or the cut-off voltage reaches 3.55 V. Therefore the scheduled current and voltage during low-temperature charging should be reconsidered to avoid capacity degradation. Lithium deposition contributes to low-temperature aging mechanisms, as something needle-like which might be deposited lithium is observed on the surface of the negative electrode after disassembling the aged battery cell. To confirm our explanation, incremental capacity analysis (ICA) is performed to identify the characteristics of the lithium deposition induced battery aging mechanisms. Furthermore, the aging mechanism is quantified using a mechanistic model, whose parameters are estimated with the particle swarm optimization algorithm (PSO). The loss of reversible lithium originating from secondary SEI formation and dead lithium is confirmed as the cause of the aging. © 2015 Elsevier B.V. All rights reserved.
Keywords: Lithium-ion battery Low-temperature aging Low-temperature charging Lithium deposition Incremental capacity analysis
1. Introduction Lithium ion batteries have become popular in the automobile industry due to their high energy and power density; however, capacity degradation in practical use restricts their broader application. Capacity degradation can be caused by multiple factors, including material properties, manufacturing techniques and practical operating conditions. The pervasively acknowledged aging mechanisms of lithium batteries are the loss of lithium ion (LLI),
* Corresponding author. E-mail address: [email protected] (M. Ouyang). http://dx.doi.org/10.1016/j.jpowsour.2015.03.178 0378-7753/© 2015 Elsevier B.V. All rights reserved.
the loss of electrode active material (LAM), and an increase in resistance (IIR) [1,2]. LLI occurs mainly on the surface of the anode due to SEI decomposition and regeneration. In terms of LAM, the loss of positive electrode material is largely responsible, although structural damage to the negative electrode of graphite induced by lithium intercalation and deintercalation in Ref. [3], is a second explanation for LAM. The increasing thickness of the SEI film causes IIR [4e6]. In previous studies, a tank model is proposed that illustrates the issue graphically and vividly [1]. Previous studies provide several enlightened approaches to analyzing the capacity degradation of lithium ion batteries. Some ex situ post-mortem technologies are widely utilized to investigate the fading of individual electrodes, such as X-ray diffraction (XRD)
310
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
and scanning electron microscopes (SEM) [7,8]. An in situ noninvasive method, impedance spectroscopy (IS), has also been recommended [9e11]. Additionally, for the semi-quantitative analysis of battery aging mechanisms, ICA and differential voltage analysis (DVA) have been used, as in Refs. [1,12e20]. To quantify the LLI and LAM, a prognostic/mechanistic model is introduced in Refs. [1,2,17e23]. By means of identifying the parameters of the model, the aging mechanism is directly clarified [1]. Batteries age far more at low temperatures than at room temperature [5,24]. It is reported that low-temperature degradation mainly occurs during the charging process due to lithium deposition, the potential for which is more likely to be achieved in the anode due to its elevated resistance at low temperatures [24,25]. S.S Zhang et al. [26] reported that even at a low current rate, the electrode potential at the anode is below 0 V vs. Li/Liþ over a half period of charging. The low potential at the anode leads to lithium deposition, which definitely damages the battery. Jiang Fan et al. studied the effects of different low-temperature voltage profiles on lithium ion batteries and suggested that lithium plating will occur at high-rate charging [25]. Low temperatures are unavoidable in practical use, however, although they are known to damage the battery. Therefore, the charging protocol should adapt to the low temperature environment to avoid possible capacity degradation. For instance, pulse heating can efficiently heat the battery to mitigate the effects of low temperatures [27]. Nevertheless, the polarization on the anode surface will be stronger with a higher rate of charge, which results in increased lithium deposition [25]. In addition, a high current rate can bring about cell capacity decay and impedance increase, whether through a pulse charge or a conventional charge [25]. Furthermore, battery degradation is aggravated by a higher charge cut-off voltage [28e30]. Hence, the low temperature charging must to be modified. Lithium ion batteries with a LiFePO4 cathode are ideal for low temperature capacity fading research because LiFePO4 is more stable than the other cathode materials [31,32]. We employ LiFePO4 as the battery cathode to avoid the degradation inherent in other cathodes and help us focus on studying the capacity degradation caused by lithium deposition at the anode. In this paper, cycle life tests are conducted to reveal the influence of the charging rate and the cut-off voltage limit on the aging mechanisms of a large format LiFePO4 battery at a low temperature (10 C). An experiment matrix was created to conduct a lowtemperature cycling test. Lithium deposition was observed using SEM after disassembling the battery post-aging. ICA is used to investigate the low temperature aging mechanism. A mechanistic/ prognostic model is employed to quantitatively reveal the LLI, LAM and IIR that occurred during low-temperature charging. Parameter changes in the mechanistic model help to quantify the degradation mechanisms, and we confirm that the main reason for low temperature aging is LLI due to lithium deposition at the anode. Based on the test results, when charging a LiFePO4 battery in a low temperature environment, here 10 C, the charge current rate should be restricted to less than 0.25 C and the cut-off voltage to less than 3.55 V.
2.2. Reference performance test A reference performance test (RPT), which comprises a capacity test, micro rate test (close to equilibrium) and hybrid pulse power characterization (HPPC) test, is developed to assess the battery's basic performance. The actual capacity is measured by two charge and discharge cycles, both at a rate of 1/3C with 1 h rest between each cycle. The low current profile of a 1/20 C charge and discharge rate is used to calculate the incremental capacity (IC) curves. The objective of the HPPC test is to measure the resistance of the charge and discharge at different states of charge (SOC) [33]. In consideration of the daily conditions of an electric vehicle, the time increment for discharge and regeneration is 30 s and 10 s, respectively. Derived from the manufacturer's maximum allowable discharge and charge current for 30 s and 10 s, the relative rates are 2 C and 1.5 C, respectively. An entire HPPC test consists of 10 cycles of this profile separated by 10% depth of discharge (DOD) CC 1/3C rate segments, followed by 1 h rest [33]. An entire RPT is shown in Fig. 1. Through such a test, the capacity, equilibrium characteristics and internal resistance can be evaluated. 2.3. Cycle life test at low temperature A cycle life test was performed at 10 C on 13 proposed cells under different conditions such as varied charge current rates, charge cut-off voltages and charge cut-off currents to analyze the aging mechanism when charging a lithium battery at a low temperature, as presented in Table 1. The discharge regime is consistent at 1/3 C current rate and 2.0 V cut-off voltage. The cycle life test consists of a total of 50 cycles for each cell using the constant current e constant voltage (CCeCV) charge protocol as illustrated in Table 1. Prior to the cycling tests, a primary characterization test was conducted to measure the cells' basic features. The cycle life test was separated by the RPT at an ambient temperature to record the cells' feature at 10 cycle intervals. The cycle life test and RPT were repeated after a rest period of 10 h to ensure the experimental temperature. It should be noted that before the final RPT was conducted, the cells remained at the ambient temperature over 10 days due to a breakdown in the environmental chamber. 2.4. Battery disassembling and half battery assembling Cell disassembly aimed at investigating the electrode dynamism through half cells and analysis by means of SEM. Before opening, cells were utterly discharged by CC discharge at a rate of 1/3 C and then by CV discharge to a cut-off voltage of 1/20 C at 2.0 V to reduce
2. Experiment 2.1. Commercial lithium-ion battery and test equipment This paper utilizes a commercial large format LiFePO4/graphite lithium ion battery with a nominal capacity of 11.5 Ah. Isothermal experiments were performed on the batteries in an environmental chamber (DONGGUAN BELL) using a multichannel bench testing system (Neware CT-4008).
Fig. 1. Reference performance test protocol at room temperature.
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320 Table 1 Cycle life test methodology at low temperature. Cell number
Charging cut-off voltage/V
Charging cut-off current/A
Charging rate/A
Cell Cell Cell Cell Cell Cell Cell Cell Cell Cell Cell Cell Cell
3.29 3.55 3.65 3.65 3.65 3.65 3.65 3.65 3.65 3.65 3.5 3.58 3.62
1/10C 1/10C 1/2C (No CV period) 1/20C 1/10C 1/10C 1/10C 1/10C 1/10C 1/10C 1/10C 1/10C 1/10C
1/2C 1/2C 1/2C 1/2C 1/5C 1/3C 1/2C 1/10C 0.25C 0.3C 1/3C 1/3C 1/3C
1 2 3 4 5 6 7 8 9 10 11 12 13
the contact between the lithium in graphite and the air. Fresh cells were opened and then assembled in a glove-box under an argon atmosphere. After a rest period of 12 h, the assembled coin cells were tested on the battery testing system Neware BTS under low current. Cells with different aging phases were opened to observe the changes in the electrodes, the anode in particular. These details will be elaborated upon in 3.1.4. 3. Results and discussion 3.1. Impact of different parameter values of charge protocols on battery characteristics 3.1.1. Capacity fade The impact on capacity fade is compared under different experimental conditions, as represented in Fig. 2. In Fig. 2(a), the effects of different cut-off voltage limits are examined. The capacity has different rates of decrease. The degradation accelerated rapidly as the cut-off voltage increased. Therefore, limiting the cut-off voltage has a mitigating effect on degradation; however, the charge capacity in cycling will diminish as the cut-off voltage is reduced. The cells' RPT indicates that the internal resistance at 50% SOC is 30.49 mU and 10.23 mU at 10 C and 25 C, respectively. This means that at a low temperature, the cut-off voltage is reached in less time at the same current rate. As an example, average charging capacity of Cell 7 maintained only 8.94 Ah, which is 75.54% of the nominal capacity. The capacity evolution is illustrated in Fig. 2(b) for charging at various current rates from 0.1 C to 0.5 C. It is evident that the capacity degradation accelerates as the rate increases. The probable
311
reason is related to the behavior of lithium plating at high rates [34,35]. The maximum designed current rate in this experiment is 0.5 C; however, the capacity fade may decrease as the rate continue to ascend to a higher level, such as 1 C due to elevated temperatures induced by joule heat or a shorter charging period [24]. The regeneration of capacity for the 50th cycle is definitely caused by a certain amount of rest time. The cells degrade non-linearly as the current rate and cut-off voltage in charging increase. There exists a turning point in the capacity retention for both charging parameters, indicating a distinct degradation once the magnitude of the parameters exceeds a certain level. The loss rate of capacity after 40 cycles at various current rates and cut-off voltages of charging is shown in Fig. 3. This indicates an evident turning point for each parameter correlated to capacity degradation, which is approximately at a rate of 0.25 C and 3.55 V. Once the charging parameter exceeds the turning point, capacity fade will be exacerbated significantly. Limited parameters based on the consideration of aging within a secure range are significant for suppressing degradation. The influence of the variation of cut-off currents at 3.65 V on the capacity fade is further investigated; results are shown in Table 2. Through changes in the cut-off current value, the lasting period of the CV stage is controlled. Accelerated degradation occurs at the time of the CV float charging at the same level as the cut-off voltage increases. Compared with Cells 3 and 4, over 85% of the capacity loss originates in CV charging. The charging time and capacity evolution for Cell 4 at 10 cycle intervals are represented in Fig. 4. The CC charging time and proportion simultaneously decrease with less charging time and the capacity to make the CV stage become the major part gradually. In spite of this, the impact of the CV stage depends on the charging condition of the previous CC stage. One recent study suggests that if the charging rate of the CC stage is restricted within a specific harmless range, the subsequent CV charging will not aggravate the situation [36]. Consequently, the degradation initiated within the CC stage. 3.1.2. Increase in internal resistance Battery resistance comprises polarization and ohmic resistance. The battery voltage comes to an abrupt change while undergoing current mutation. The response time of ohmic resistance is very fast, almost instantaneous, followed by a polarization response after a few seconds. In this paper we consider a 1 s response resistance as an ohmic unit and 30 s (discharge) or 10 s (charge) as the total internal resistance of the battery. The polarization resistance comes from the difference between the total resistance and the ohmic resistance. Battery resistance at each SOC was calculated from the RPT test.
Fig. 2. Evolution of standard test discharge capacity at intervals of 10 cycles: (a) charging cut-off voltage; (b) charge current rate.
312
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
Fig. 3. Turning points of capacity loss rate after 40 cycles for different charging parameters: (a) charging rate; (b) cut-off voltage of charging.
Table 2 Effect of CV charge period at 3.65 V on the degradation condition. Cell number
Cut-off current at CV stage
Original capacity/Ah
50 cycle capacity/Ah
Degradation rate
Cell 3 Cell 7 Cell 4
1/2C 1/10C 1/20C
11.80 11.82 11.86
11.32 9.58 8.63
4.07% 18.95% 27.23%
Fig. 5 displays 30 s discharge resistance changes for Cell 7. The resistance shows an apparent increase of approximately 5%e7% with cycles. The increase of SEI thickness across cycles plays an important role, albeit not an exclusive one, in the increase of battery resistance [37]. Fig. 6 displays the evolution of the resistance and capacity. It should be noted that the internal resistance and capacity loss decrease after the 5th ten-cycle group compared with the 4th. After the 50th cycle, the cells were stored at room temperature for sufficient rest, which made the cells recover in performance corresponding with a rise in capacity. This is an indication that the lost capacity is not immediately irreversible; rather, it can be regenerated by sufficient rest. Generally, capacity recovery is involved in the decrease of resistance, implying a correspondence between the two characteristics.
profiles at intervals of 10 cycles are presented to reveal the cells' evolution. Profiles of Cells 4 and 7 show not only a reduction in charge capacity but also evident changes in shape and position. They vary noticeably between the first 20 cycles and immediately
3.1.3. Charge profile evolution Fig. 7 presents the voltage profiles with cycling at 10 C. It is important to note that the profiles do not change continuously with the execution of standard tests; here, each first cycle after the standard test was discharged at room temperature. Therefore, the
Fig. 5. 30 s discharge resistance evolution at different SOC from the RPT for Cell 7.
Fig. 4. Evolution of CC charge time and charge capacity with cycles of the Cell 4.
Fig. 6. Changes in resistance at specific SOC compared with capacity retention across cycles for Cell 7.
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
following 30 cycles. Under 3.5 V, the charge profiles have a reduced voltage with an elevated trend up to 3.5 V. This leads to reaching the cut-off voltage in a very short period of time, resulting in a decrease in the available capacity. An increase in the charge internal resistance may have been a contributing factor. In contrast, cells showing a slight capacity loss exhibited less change in terms of the charge voltage, the shape of which remained stable upon cycling. The charge plateau was also elevated due to the rise in resistance, as in cells 5 and 6. 3.1.4. Cell disassembling examination Cell disassembling is a constructive and direct methodology to observe change in cells whereby three cells of varying degreesea fresh cell, Cells 2 and 7ewere examined. Before the operation, cells are fully discharged by the CC procedure at a 1/3 C current rate with a subsequent CV procedure at a 1/20 C cut-off current rate. After opening the cells, the current collectors are covered with anode and cathode electrodes and the separator samples are cut out. The cathode electrode appears almost unchanged upon a macroscopic check; the anode electrode, in contrast, changes significantly. The anode samples from the different cells after resting in the air for a certain period are shown in Fig. 8. The anode surface of the fresh cell is colored gray, the color of delithiated graphite [38], as shown in Fig. 8(a). The anode surface of Cell 2 is dark blue, but a few white dots are visible in Fig. 8(b). Fig. 8(c) depicts severe degradation in the form of a multitude of white dots on the anode surface of Cell 7; the dots are considered to be oxide products of the active substance in contact with the atmosphere. The color and morphology of the fresh cell and Cell 2 remain relatively steady in the air. In contrast, the anode changes in color and surface morphology due to exposure in the air. The details are
313
shown in Fig. 9. At the instant the cell is opened, denoted as 0 s, a majority of the anode appears dark yellow. In the subsequent 20 s, the yellow portion gradually becomes black. After approximately 80 s, some parts of the surface turn white so that concentrated white dots appear in a bubble formation and escape. The phenomenon suggests lithium plating in the anode. The above discharging procedure is not capable of rendering graphite in a complete state of deintercalation. By inference, the lithium layer on the surface of the graphite remains as the actual part of the anode that oxidizes during exposure. The details are discussed in Section 3.2.2. The above macroscopic changes reveal the strong reactions of the surface layer lithium and graphite intercalation with the atmosphere. In Fig. 10, the lithiated graphite appears as red or gold [38] together with the lithium layer, and the SEI film on the surface is dark yellow. Therefore, the silver metallic luster of the lithium does not appear. Lithium first reacts with N2 and then produces black Li3N, after which LiOH is generated by Li3N in contact with moisture in the air, with NH3 yielding. Finally, LiOH converts to Li2CO3 through the process of slow oxygenation [39], which is the origin of the aforementioned white dots. The microstructure of the anode and separator from the cells are analyzed by means of SEM. An ultrathin conductive coating was deposited on the surface of the samples by an ETD-800 automatic sputter coater from Elaborate Technology Development, Ltd. The SEM is a JSM-7401F of JEOL Co., Ltd. Electron micrographs of the separator and anode are shown in Fig. 10. Separator images of the fresh cell, Cells 2 and 7 are depicted in Fig. 10(a)e(c). The porous separator is gradually blocked by the products of side reactions, rendering Liþ more difficult to transport and leading to an increase in ohmic resistance. In Fig. 10(d)e(f), the
Fig. 7. Comparison of the charging profiles at low temperature among cells with different degrees of degradation: (a) and (b) represent serious degradation; (c) and (d) represent slight degradation; (a) Cell 4; (b) Cell 7; (c) Cell 6; (d) Cell 5.
314
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
Fig. 8. Photographs of the negative electrode from cells at different degrees of degradation: (a) fresh cell; (b) Cell 2 after the cycle test; (c) Cell 7 after the cycle test.
Fig. 9. Changes in the color and morphology of Cell 7's negative electrode surface in contact with the atmosphere from 0 s to 80 s.
SEM images of the anode surface are presented. The particle morphology of the graphite is intensely distinct in Fig. 10(d). As the capacity loss increases, something needle-like which might be metal lithium deposition are visible on the surface of particles, as in Fig. 10(e); images at a higher magnification are shown in Fig. 11. Comparatively speaking, in Fig. 10(f), the particulate graphite disappears. Instead, a film with holes of various sizes covers almost the entire surface area of the negative electrode. It can be deduced that the film consists of the primary SEI film and secondary SEI produced by the deposited lithium [40]. The holes are proved to be created by the bubbles that are the reaction product of the lithium in the air. The lithium deposition is creating a serious situation. At a low temperature, lithium is more likely to be deposited due to a decrease in the solid diffusion coefficient and low anode potential [26,41,42]. The morphology is determined by the current rate. Needle-like lithium deposition at high current rate may break the SEI film due to stress. As a result, the lithium grows out continuously [43]. Needle-like deposition is long and slender with branches and bending as shown in Fig. 11. In Fig. 11(a), deposition needles of different lengths and diameters are indicated by arrows. Needle-like deposition grows on the lithium substrate through the SEI film, as indicated in Fig. 11(b). Therefore, they lie down on the surface and may lack the mechanical strength to perforate the separator and resulting in an internal short-circuit. The dendrite lithium, however, may crack where it is fragile and isolate itself
from the lithium substrate, leading to a transition to the dead lithium [40]. Mechanisms of lithium deposition at room temperature and low temperature are different. Lithium deposition is generally caused by two dominant factors: 1) High rate charging leads to high lithium diffusion rate, and lithium deposition occurs on the surface of anode when the lithium formation rate exceeds its intercalation rate [44]; 2) Lithium deposition is more prone to happen at anode under local high over-potential, which is determined by the product of charge current and resistance at anode [26]. At room temperature, the lithium diffusion rate and the over-potential can be elevated by high current rate during rapid charging. However, at low temperature, the lithium diffusion rate decreases and resistance of anode increases [25,45]. Therefore, the condition of lithium deposition is stricter at low temperature than that at room temperature with same current rate. 3.2. Incremental capacity analysis of the aging mechanism at a low temperature In Section 3.1, the evolution of the capacity, resistance and charging curves are discussed and the SEM images of separators and anodes are analyzed. In addition, quantitative analyses must to be conducted for reliability. It is reported that the IC curve, i.e. dQ/ dV-V, was used to deduce cell aging mechanisms in several studies
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
315
Fig. 10. SEM images of separators and graphite electrodes from cells at different degrees of degradation caused by various cut-off voltages: (a) separator of the fresh cell; (b) separator of Cell 2; (c) separator of Cell 7; (d) graphite electrode of the fresh cell; (e) graphite electrode of Cell 2 (f) graphite electrode of Cell 7.
Fig. 11. SEM images of the needle-like deposition which might be lithium on the surface of the negative electrode Cell 5.
[1,13,37,46,47]. In this section, degradation mechanisms are investigated by the semi-quantitative means of ICA. ICA is a process to reduce the order of the voltage profile, reflecting the electrochemical procedure evolution more specifically during charging [12]. Peaks in the IC (dQ/dV) curve reveal the phase transitions and the magnitude, position, and shape, which reflect electrochemical dynamism [1,12,47]. 3.2.1. IC curves analysis using micro currents at room temperature Each peak in an IC curve is correlated with a specific phase transition, of which the magnitude, shape and position are important indicators [13]. Every standard test at room temperature containing a 1/20 C rate charge process offers close-to-equilibrium conditions to reduce the polarization impact for the purpose of ICA. Here, we derive dQ/dV curve using the probability density function (PDF) method referred to previous studies [16]. The IC curve is then smoothed. Fig. 12 displays the IC curves of the anode (negative electrode) and cathode (positive electrode). Different peaks in the IC curve of the anode manifest corresponding to different phases of the
transition process. The evolution in the concentration of intercalation lithium ultimately accounts for multiple steps of the phase transitions in the graphite negative electrode. It is reported that there are three voltage plateaus at 210 mV, 120 mV, 85 mV in graphite, corresponding to x from 0.08 to 1, where x appears in LixC6 [48]. Further, reactions in this range are divided into four regions in which different stages of lithium-graphite intercalation compounds exist with phase transitions from a higher stage to a lower stage in reaction. It should be noted that the voltage regions of peaks ②, ③ and ⑤ are slightly lower than those previous reported literature due to polarization. In Fig. 13, IC curves of Cells 5 and 7 are compared after different low temperature cycles, respectively. 4 peaks are distinctly observed and denoted as ①, ②, ③ and ⑤, corresponding to four phase transition processes in the anode. Peak ④ cannot be explicitly identified. As for the cathode, it has a FePO4eLiFePO4 two phase transition region offering a long potential plateau as the reference voltage during charge and discharge, which is denoted as Ⅱ. Therefore, four peaks are defined as ①*Ⅱ, ②*Ⅱ, ③*Ⅱ and ⑤*Ⅱ, as suggested in the literature [47].
316
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
Fig. 12. IC curves of the positive and negative electrode: (a) LFP positive electrode; (b) graphite negative electrode.
During 50 cycles at low temperature, the first peak in the IC curve of Cell 7, denoted as ①*Ⅱ, changed in density and position but not in shape, as shown in Fig. 13(a). The ①*Ⅱ peak density of Cell 7 decayed significantly until it disappeared, while the other three peaks remained stable in density in the preceding 40 cycles. Specifically, the density and position of the 50th item returning to the previous state are noticeably reflected. Deviation at a higher voltage occurs in this position, which suggests an increase in the battery resistance, as shown in previous HPPC results in Section 3.1.2. Changes of peaks along with HPPC results and SEM images indicate that the capacity decay originated in LLI from lithium deposition and that the thickness of the SEI film increased due to the reaction between the active deposited lithium and electrolytes, contributing to the raised battery resistance. First, the mechanism of capacity fade is LLI, otherwise, the peak magnitudes in Fig. 13 will change. Second, LLI comes from the lithium deposition on the surface of the negative electrode due to observable lithium dendrites in Fig. 11. Third, insoluble products, which comes from the reaction between the deposited lithium and electrolyte enhance the thickness of the SEI film and the internal resistance inferred from the HPPC test shown in Fig. 5. As Fig. 13(a) shows, peak ①*Ⅱ density of Cell 7 decays to a greater extent because the capacity degradation is more significant. The dominant reaction of the third stage in graphite is as follows [48],
LiC6 $C6 þ Li#2LiC6 Accompanied with the reduction in lithium ion, the phase transition process gradually decreases in number, making the
lithium-graphite compound at the end of discharge exist in the form of the second-stage phase, i.e., LiC12 [38]. 3.2.2. IC curves analysis using charge profiles at low temperature In 3.2.1, peaks featuring the evolution of the IC curve at a low charge rate (1/20 C) were used to elucidate the aging mechanisms of the LFP/graphite battery cycle at a low temperature, which offers a close-to-equilibrium process during charging and therefore renders each peak an accurate and distinct presence. As the rate increases, the polarization for charge and discharge retains a greater influence on the peaks, even forcing some small peaks to merge. However, the analysis is valid within a specific rate range. In our previous study [1], we performed an explicit analysis on a 1/3 C charge profile to identify the on-board aging mechanism in a battery management system (BMS). Here, it is the IC curve that is employed to explain the evolution of the charge curve at a low temperature in different cycles. Fig. 14 shows the difference between the 1st cycle immediately after the room temperature standard test and the subsequent cycles at a low temperature. The difference can be explained as more quantities of charge capacity into the battery in the 1st cycle after a full discharge at room temperature. The shape of the 1st charge curve is close to 2-10th cycles for Cell 3 because of less degradation as shown in Fig. 14(c). Deviation in the shape of the 1st charge curve from that of the 2-10th cycle for Cells 4 and 7 indicates more degradation. The shape of 2-10th charge curves has significant deviation from the 1st charge curve at low temperature cycling, as shown in Fig. 14(a) and (b). The IC curves of the different cycles of Cell 7 compared with the
Fig. 13. IC curves from the micro current test after various cycles of Cells 7 and 5: (a) Cell 7; (b) Cell 5.
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
317
Fig. 14. Low temperature charge curve from the 1st cycle to the 10th cycle: (a) Cell 4; (b) Cell 7; (c) Cell 3; (d) Cell 1.
positive electrode are displayed in Fig. 15. In particular, IC curves are calculated through non dimensional computing, which is convenient for comparing curves among the different capacity scales. In Fig. 15, the red line is the IC curve of the 1st cycle and the cyanic line is the 10th cycle. The green solid and dashed lines are the positive electrode IC curve. Compared with the 1st cycle, peak ③*Ⅱ decreased in cycle 10 and peak ②*Ⅱ became predominant. On the one hand, at a low temperature, the charge and discharge capacity hold at only 70% of their room temperature capacity. The charge capacity of the 1st cycle is 10.81 Ah with 8.95 Ah discharged, whereas a capacity of 9.17 Ah (only 84.83% of the 1st cycle) is charged in the 2nd cycle. The following capacity of discharge is
slightly less than that of the charge with a coulombic efficiency of 98e99%. Therefore, the discrepancy of the SOC range is probably responsible for the charge abnormality in the 1st cycle. The partial discharge in the 1st cycle accounts for the initial offset of the SOC charge to a higher point and the resulting reduction in the magnitude of peak ③*Ⅱ. On the other hand, the shape of the IC curve after the 1st cycle is similar to that of the LFP half-cell after being translated to a higher potential, considering the negative electrode potential and the polarization factor. In other words, lithium deposition on the anode surface after the 1st charge process changes the electrochemical dynamism of the electrodeelectrolyte interface; namely, the lithium ion intercalation into metal lithium increased, but less into the graphite layer to form a graphite compound. This validates the predominant aging mechanism mentioned in 3.2.1. Moreover, the relative magnitude of the characteristic peaks could be a new evolution criterion to detect lithium deposits, which is a potential application of BMS. Herein lies in an additional problem concerning the drastic changes in position in the cycles for Cells 4 and 7. Fig. 16 illustrates the evolution of the IC curves at every 10th cycle of Cell 4. The shape and magnitude of the characteristic peaks do not change significantly, but the positions of peaks 1 and 2 gradually move to a higher potential. This indicates that the rise in battery resistance, which is mainly due to the increase in the thickness of the SEI film, is a key reason behind this change. 3.3. Identification of aging parameters based on prognostic model using particle swarm optimization algorithm
Fig. 15. The evolution of the low temperature cycle of the 1st cycle and the 10th cycle compared with the positive electrode IC curve (solid blue line for the original curve and dashed blue line for the right move curve). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
According to our previous research [1], the method for reconstructing the CC charging voltage curve can be used to quantitatively identify the evolution of the capacity of the two electrodes,
318
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
the quantities of lithium ion and the impedance value. Based on the half electrode charge voltage curve, the LFP/ graphite potential versus Li/Liþ is a uniform function with the number of transferred lithium-ion. Therefore, it is possible to reconstruct the full cell charge voltage curve if the transferred lithium-ion is evaluated. The model is described by Equations (1)e(3):
x ¼ x0 þ
Q Cn
(1)
y ¼ y0
Q Cp
(2)
Vc ¼ Up ðyÞ Un ðxÞ þ IR
(3)
The equations are referred to our previous study [1]. Cn and Cp are estimated to be the capacity of the anode and cathode electrode, respectively. In addition, x0 and y0 represent the initial number value of the lithium ion insertion in the two electrodes, respectively. R is the battery internal resistance as a whole. Here, the parameter Q is the charge capacity at the moment when I is charging current value; both can be measured. It should be noted that the simulated x and y do not represent the real SOC, only the relative values for the evolution of lithium ion amount. Therefore, five total parameters of x0, y0, Cn, Cp and R will be estimated to minimize the cost function as Equation (4):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Ptn 2 t¼t1 ðVm ðtÞ Vc ðtÞÞ RMSE ¼ n
(4)
The cost function is actually the root mean squared error (RMSE) between the measured and calculated voltage. Here, the PSO algorithm is used to determine the optimal parameters. Fig. 17 presents a comparison between the measured and simulated results of Cell 7 from 6 standard tests after every 10 cycles. Only the CC charge section is used for the simulation because CV charge is invalid. The satisfied overlap indicates not only the appropriateness of the method but also the accuracy of the simulation. Based on the above analysis, we can find the effect of the different charge parameters on the cell degradation. It is necessary to further analyze the aging mechanisms at a low temperature. Although the mechanisms of capacity decay and battery aging vary, they can be divided into problems of the cathode electrode, anode
Fig. 16. Evolution of the IC curves at a 10 cycle intervals.
Fig. 17. Charge curve fitting results (green square) vs. experiments profiles (blue line) for Cell 7 after every 10 cycles. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
electrode and electrolytes. A rise in battery impedance may be accounted for by continuous SEI growth due to electrolyte decomposition [37], which may in turn cause capacity fade. In conclusion, changes in the two electrodes and impedance allow a quantitative analysis of aging. Fig. 18 displays the results of the evolution of the identified parameters with cycling. Before the cycle test, the identified original capacity of the negative electrode is approximately 20% beyond the positive, as shown in the figure, which is designed to prevent lithium deposition resulting from overcharging. The capacity of each electrode is higher than the nominal capacity of the entire battery, expressing that only part of the active material is available. After the tests, the relatively unchanged Cp and Cn show no apparent loss of active material (LAM) for both electrodes, and x0 also remains constant. The total lithium number (Cp*y0þCn*x0), however, reveals a clear decrease in Fig. 19, indicating the presence of LLI. The lithium number of Cell 5 at 1/5 C remains comparatively unchanged. It is concluded that the decrease in the rechargeable capacity is the result of increasing irreversible lithium ions. The RMSE between voltages estimated and derived from experiments is within 10 mV. The battery average resistance gradually increased, which verifies the resistance analysis found in the results of the previous HPPC experiment. Therefore, it is deduced from the analysis that the low temperature degradation of LFP/graphite is mainly due to LLI induced by lithium deposition on the surface of anode electrode. The solid diffusion coefficient in the anode, which is universally considered as a rate-limit factor, drops more severely than electrolyte diffusion. As a result, it becomes more difficult for the lithium ion to intercalate into the anode, accumulating on the surface of the graphite. Once it reaches lithium deposition potential level, the lithium metal is plated. It should be noted that the shape/dimension of the large format cell will influence the lithium deposition at low temperature during cycling. The lithium deposition at specific positions within the cell depends on the local current density and temperature. The distribution of current density and temperature depends on shape/ dimension of the large format cell [49]. Therefore the lithium deposition phenomenon within large format cells with different shapes/dimensions can be different, which needs further investigation. However, the main mechanism of the capacity degradation during low temperature cycling is similar as discussed in this paper. It should be also noted that lithium deposition process proceeds
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
319
Fig. 18. Evolution of battery parameters fitting results after 50 cycles: (a) x0; (b) y0; (c) capacity of the negative electrode; (d) capacity of the positive electrode; (e) average resistance; (f) RMSE.
inside the SEI due to its insulation property. As the metallic state, once the lithium makes contact with the electrolyte due to the collapse of the primary SEI, the secondary SEI film will form. Therefore, lithium is prevented from reacting further with the electrolyte by the compact SEI film. If the potential increases again, oxidation and electron transfer will occur on the surface, resulting in lithium dissolution and capacity recovery [43]. This explains the phenomenon that the battery capacity increases post-degradation after resting at room temperature for 10 days. If it is deposited as dendritic lithium on the surface, however, the SEI film may be damaged such that active metal lithium forms undissolved inorganics by reacting with the electrolyte, leading to irreversible loss [30,41,43,50]. The inorganics will also increase the thickness of SEI, causing an increase in the battery resistance, as Fig. 20 represents [51]. Dendritic lithium will also affect the security of the battery by perforating the separator if sufficient mechanical strength is achieved [40].
Fig. 19. Evolution of lithium numbers in the battery with cycles at different charge rate.
4. Conclusion Low temperature cycle life experiments were performed at 10 C, and quantitative methods were used to identify the LFP battery aging mechanism. Capacity fade was more severe with a higher charging rate and higher cut-off voltage, and the turning points for the current rate and cut-off voltage after which degradation accelerated were determined to be approximately 0.25 C and 3.55 V. After dissembling the batteries with different cut-off
Fig. 20. Schematic of lithium deposition on the surface of the graphite negative electrode with cycles at a low temperature. The permanent loss of lithium stems from the secondary SEI formation and dead lithium.
320
M. Ouyang et al. / Journal of Power Sources 286 (2015) 309e320
voltages, something needle-like which might be deposited lithium was observed and the morphologies of the anode with different degrees of degradation were compared with the SEM images. Subsequently, the degradation was examined by ICA using low current charging profiles at room temperature and cycle charging profiles at a low temperature. This revealed that the mechanism is the loss of lithium rather than loss of active electrode material. The loss of lithium stems from the lithium deposition on the surface at the anode. The electrochemical behavior of the anode was altered by the process of lithium deposition, rendering the graphite anode similar to the lithium metal anode. Finally, a prognostic and mechanic model was used to quantitatively analyze the aging mechanisms. The decline in numbers of reversible lithium was quantitatively illustrated, proving that the loss of capacity results from the loss of lithium. The deposited lithium in contact with the electrolyte to form a secondary SEI film and dead lithium from the lithium dendrite are supposed to be the two origins of the LLI. Acknowledgment This research is funded by the MOST (Ministry of Science and Technology) of China under the contract of No. 2014DFG71590 and Beijing science and technology plan No. Z121100007912001 and the National Support Plan 2013BAG16B01 and MOE (Ministry of Education) of China under the contract of No. 2012DFA81190. References [1] X. Han, M. Ouyang, L. Lu, J. Li, Y. Zheng, Z. Li, J. Power Sources 251 (2014) 38e54. [2] M. Dubarry, C. Truchot, B.Y. Liaw, J. Power Sources 219 (2012) 204e216. [3] V.A. Sethuraman, L.J. Hardwick, V. Srinivasan, R. Kostecki, J. Power Sources 195 (2010) 3655e3660. [4] K. Amine, J. Liu, I. Belharouak, Electrochem Commun. 7 (2005) 669e673. [5] N. Omar, M.A. Monem, Y. Firouz, J. Salminen, J. Smekens, O. Hegazy, H. Gaulous, G. Mulder, P. Van den Bossche, T. Coosemans, Appl. Energ. 113 (2014) 1575e1585. [6] H. Lin, D. Chua, M. Salomon, H.C. Shiao, M. Hendrickson, E. Plichta, S. Slane, Electrochem. Solid-State Lett. 4 (2001) A71eA73. [7] J. Li, E. Murphy, J. Winnick, P.A. Kohl, J. Power Sources 102 (2001) 294e301. [8] L. Yang, X. Cheng, Y. Ma, S. Lou, Y. Cui, T. Guan, G. Yin, J. Electrochem Soc. 160 (2013) A2093eA2099. [9] A.N. Jansen, D.W. Dees, D.P. Abraham, K. Amine, G.L. Henriksen, J. Power Sources 174 (2007) 373e379. €ltzsch, O. Kanoun, H. Tr€ [10] U. Tro ankler, Electrochim. Acta 51 (2006) 1664e1672. [11] S.S. Zhang, K. Xu, T.R. Jow, Electrochim. Acta 49 (2004) 1057e1061. [12] M. Dubarry, V. Svoboda, R. Hwu, B.Y. Liaw, Electrochem. solid-state Lett. 9 (2006) A454eA457. [13] M. Dubarry, B.Y. Liaw, J. Power Sources 194 (2009) 541e549. [14] A.H. Thompson, J. Electrochem Soc. 126 (1979) 608e616. [15] H.M. Dahn, A.J. Smith, J.C. Burns, D.A. Stevens, J.R. Dahn, J. Electrochem Soc.
159 (2012) A1405eA1409. [16] X. Feng, J. Li, M. Ouyang, L. Lu, J. Li, X. He, J. Power Sources 232 (2013) 209e218. [17] J. Remmlinger, M. Buchholz, T. Soczka-Guth, K. Dietmayer, J. Power Sources 239 (2013) 689e695. [18] K. Honkura, K. Takahashi, T. Horiba, J. Power Sources 196 (2011) 10141e10147. [19] I. Bloom, A.N. Jansen, D.P. Abraham, J. Knuth, S.A. Jones, V.S. Battaglia, G.L. Henriksen, J. Power Sources 139 (2005) 295e303. [20] J. Kim, N.P. Pieczonka, Z. Li, Y. Wu, S. Harris, B.R. Powell, Electrochim. Acta 90 (2013) 556e562. [21] J. Christensen, J. Newman, J. Electrochem Soc. 150 (2003) A1416eA1420. [22] J. Christensen, J. Newman, J. Electrochem Soc. 152 (2005) A818eA829. [23] X. Feng, J. Sun, M. Ouyang, X. He, L. Lu, X. Han, M. Fang, H. Peng, J. Power Sources 272 (2014) 457e467. [24] S. Tippmann, D. Walper, L. Balboa, B. Spier, W.G. Bessler, J. Power Sources 252 (2014) 305e316. [25] J. Fan, S. Tan, J. Electrochem Soc. 153 (2006) A1081eA1092. [26] S.S. Zhang, K. Xu, T.R. Jow, J. Power Sources 160 (2006) 1349e1354. [27] X. Zhao, G. Zhang, L. Yang, J. Qiang, Z. Chen, J. Therm. Analysis Calorim. 104 (2011) 561e567. [28] O.J. Kwon, W. Fang, C. Wang, Meeting Abstracts, MA2010-03, 2010, p. 279. [29] O.J. Kwon, C. Wang, Meeting Abstracts, MA2009-02, 2009, p. 544. [30] S. Tippmann, D. Walper, L. Balboa, B. Spier, W.G. Bessler, J. Power Sources 252 (2014) 305e316. [31] W. Zhang, J. Power Sources 196 (2011) 2962e2970. [32] F. Gao, Z. Tang, Electrochim. Acta 53 (2008) 5071e5075. [33] Freedom C.S. Department of Energy (October 2003) DOE/ID-11069, (2003). [34] N. Omar, M.A. Monem, Y. Firouz, J. Salminen, J. Smekens, O. Hegazy, H. Gaulous, G. Mulder, P. Van den Bossche, T. Coosemans, Appl. Energ. 113 (2014) 1575e1585. [35] M. Broussely, P. Biensan, F. Bonhomme, P. Blanchard, S. Herreyre, K. Nechev, R.J. Staniewicz, J. Power Sources 146 (2005) 90e96. [36] M. Wu, P.J. Chiang, J. Lin, J. Electrochem Soc. 152 (2005) A47eA52. € ller, J.O. Besenhard, M. Winter, [37] J. Vetter, P. Nov ak, M.R. Wagner, C. Veit, K.C. Mo M. Wohlfahrt-Mehrens, C. Vogler, A. Hammouche, J. Power Sources 147 (2005) 269e281. [38] S.J. Harris, A. Timmons, D.R. Baker, C. Monroe, Chem. Phys. Lett. 485 (2010) 265e274. [39] H. Osborg, Trans. Electrochem. Soc. 66 (1934) 91e105. pez, J.T. Vaughey, D.W. Dees, J. Electrochem Soc. 159 (2012) [40] C.M. Lo A873eA886. [41] Z. Li, J. Huang, B.Y. Liaw, V. Metzler, J. Zhang, J. Power Sources 254 (2014) 168e182. [42] C.K. Huang, J.S. Sakamoto, J. Wolfenstine, S. Surampudi, J. Electrochem Soc. 147 (2000) 2893e2896. [43] J. Yamaki, S. Tobishima, Rechargeable Lithium Anodes, Wiley-VCH Verlag GmbH & Co. KGaA, 2011, pp. 377e404. [44] B.K. Purushothaman, U. Landau, J. Electrochem Soc. 153 (2006) A533eA542. [45] A.N. Jansen, D.W. Dees, D.P. Abraham, K. Amine, G.L. Henriksen, J. Power Sources 174 (2007) 373e379. [46] M. Dubarry, V. Svoboda, R. Hwu, B.Y. Liaw, J. Power Sources 165 (2007) 566e572. [47] M. Dubarry, B.Y. Liaw, M. Chen, S. Chyan, K. Han, W. Sie, S. Wu, J. Power Sources 196 (2011) 3420e3425. [48] T. Ohzuku, Y. Iwakoshi, K. Sawai, J. Electrochem Soc. 140 (1993) 2490e2498. [49] U.S. Kim, C.B. Shin, C. Kim, J. Power Sources 189 (2009) 841e846. [50] G. Zheng, S.W. Lee, Z. Liang, H. Lee, K. Yan, H. Yao, H. Wang, W. Li, S. Chu, Y. Cui, Nat. Nanotechnol. 9 (2014) 618e623. [51] J. Christensen, J. Newman, J. Electrochem Soc. 150 (2003) A1416eA1420.