Electrode Side Reactions, Capacity Loss and Mechanical Degradation in Lithium-Ion Batteries

Jiagang Xu; Rutooj D. Deshpande; Jie Pan; Yang-Tse Cheng; Vincent S. Battaglia

doi:10.1149/2.0291510jes

Electrode Side Reactions, Capacity Loss and Mechanical Degradation in Lithium-Ion Batteries

Xu, Jiagang; Deshpande, Rutooj D.; Pan, Jie; Cheng, Yang-Tse; Battaglia, Vincent S. 2015-07-28 00:00:00 A2026 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) Electrode Side Reactions, Capacity Loss and Mechanical Degradation in Lithium-Ion Batteries a,= b,=,z a,∗ a Jiagang Xu, Rutooj D. Deshpande, Jie Pan, Yang-Tse Cheng, c,∗∗ and Vincent S. Battaglia Department of Chemical & Materials Engineering, University of Kentucky, Lexington, Kentucky 40506, USA Electriﬁed Powertrain Engineering, Ford Motor Company, Dearborn, Michigan 48124, USA Environmental Energy Technologies Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA For advancing lithium-ion battery (LIB) technologies, a detailed understanding of battery degradation mechanisms is important. In this article, experimental observations are provided to elucidate the relation between side reactions, mechanical degradation, and capacity loss in LIBs. Graphite/Li(Ni Mn Co )O cells of two very different initial anode/cathode capacity ratios (R, both R > 1/3 1/3 1/3 2 1) are assembled to investigate the electrochemical behavior. The initial charge capacity of the cathode is observed to be affected by the anode loading, indicating that the electrolyte reactions on the anode affect the electrolyte reactions on the cathode. Additionally, the rate of “marching” of the cathode is found to be affected by the anode loading. These ﬁndings attest to the “cross-talk” between the two electrodes. During cycling, the cell with the higher R value display a lower columbic efﬁciency, yet a lower capacity fade rate as compared to the cell with the smaller R. This supports the notion that columbic efﬁciency is not a perfect predictor of capacity fade. Capacity loss is attributed to the irreversible production of new solid electrolyte interphase (SEI) facilitated by the mechanical degradation of the SEI. The higher capacity fade in the cell with the lower R is explained with the theory of diffusion-induced stresses (DISs). © The Author(s) 2015. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0291510jes] All rights reserved. Manuscript submitted March 3, 2015; revised manuscript received July 13, 2015. Published July 28, 2015. Lithium-ion batteries (LIBs) are widely used in small, portable, tal observations of DISs and stress induced cracking of electrode electronic devices due to their high energy density, high voltage, low materials. For example, Sethuraman et al. experimentally measured 1,2 self-discharge rate, and good cycle performance. Yet, for automo- the stresses in thin ﬁlm electrodes using wafer curvature methods. In tive propulsion applications, LIBs need improvement in volumetric another publication, Li et al. observed the cracking pattern in silicon and gravimetric energy density and with performance over the life thin ﬁlm electrodes and correlated it with the electrode thickness. of the vehicle. This is one of the major motivations of researchers In this research, an attempt is made to understand the relation- worldwide focusing on developing cheaper and more durable LIBs ship between coulombic efﬁciency, capacity retention, and mechani- for applications such as hybrid vehicles, electric vehicles, and other cal stresses. Two types of graphite/Li(Ni Mn Co )O (NMC) full 1/3 1/3 1/3 2 large-scale energy storage systems. cells were designed with different anode to cathode capacity ratios Capacity decay during storage and charge-discharge cycling is (R). The cathode loading was the same in both cells, thus R was ad- one of the major challenges that limit the life of LIBs. Instability of justed by changing the anode loading. Both cells were charge limited the electrolyte at the operating potentials results in side reactions on by the amount of lithium that could be removed from the cathode at a the electrode surfaces, part of which lead to the formation of solid given cutoff voltage and discharge limited by amount of lithium could 4–7 electrolyte interphases (SEIs). Side reactions may result in lower be removed from the anode. We found that, during electrochemical coulombic efﬁciency, loss of usable capacity of the cell, and increase cycling, the cell with the higher anode loading (i.e., high R value) of cell impedance. Cell performance degradation due to side reactions had much lower coulombic efﬁciency as compared to the cell with the is termed as chemical degradation, which is known to be the main lower anode loading (i.e., small R value). Monitoring of the charge 5,6,8–10 cause of lithium loss in well-made LIBs. These side reactions and discharge endpoints vs. time revealed that the marching rate of 11,12 lead to marching of the charge and discharge endpoints. Such side the charge endpoint was affected by the amount of anode loading for a reactions and their effects on the cell life are relatively less explored. ﬁxed cathode loading. We propose that the electrolyte reactions on the The solid state diffusion of lithium atoms in and out of host elec- cathode are affected by the electrolyte reactions on the anode. Thus, trode particles results in diffusion induced stresses (DISs) and volume there is a “cross-talk” between the two electrodes. Interestingly, the changes in electrode particles during charge and discharge. Depending cell with the higher R value suffered a slower rate of capacity fade upon the operating conditions these stresses may have various effects during cycling, though its total rate of side reactions was greater. Here, on the electrodes such as mechanical fatigue and fracture of the elec- the total rate of side reactions is deﬁned as the sum of marching rate trode particles, electrode particle isolation from the composite matrix, of discharge and charge end points. These observations support recent 13,14 12 or SEI fracture, to name a few. DISs have been studied in detail by ﬁndings presented in Deshpande et al., where it was demonstrated mathematical modeling and experimental observations. Previous re- through the use of an electrolyte additive that coulombic efﬁciency is ports have discussed mechanical stresses in single electrode particles not a perfect indicator of battery life. Thus, a cell with higher rates 13,15–17 and thin ﬁlm electrodes. Apart from the loss of active electrode of side reactions can have less capacity loss than a cell of the same material, mechanical degradation of electrode particles facilitates ad- chemistry with lower rates of total side reactions, which supports the ditional side reactions by exposing new surface area to the electrolyte. premise that not all side reactions lead to capacity fade in a battery, Various mathematical models have been developed to explain the ca- at least in the early stages of cycle life. We propose that there are two pacity loss in batteries during electrochemical cycling with coupled types of side reactions: one which causes lithium loss from the system 4,18 chemical and mechanical degradation as a prevailing mechanism. and one which shuttles between the two electrodes without actually Along with mathematical models there have been several experimen- leading to irreversible lithium loss. Based on the cell test data, we conclude that the reactions caus- ing the irreversible lithium loss in the system are mainly linked to These authors contributed equally to this work. the formation of additional SEI on newly exposed electrode surfaces Electrochemical Society Student Member. on the anode. We develop a DIS model correlating electrode volume ∗∗ Electrochemical Society Active Member. E-mail: [email protected] expansion with the capacity loss of the cell. Lower capacity loss in Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2027 Table I. Anode/Cathode pair fabricated. Anode loading, Cathode loading, Calculated graphite 2 2 ∗ ∗ 2 mg/cm mg/cm Anode thickness, Cathode thickness, surface area, cm Capacity Ratio ( = 14.3 mm) ( = 12.7 mm) μm μm (approximation) R = 1.18 2.89 6.37 41 42 13.26 R = 4.54 11.83 6.77 71 42 54.20 - Diameter of electrode in coin cell. Thickness is measured after calendering; thickness of Cu/Al foil is included. The thickness shown here is the representative average value of the laminate from which the disk electrodes are punched. the cell with higher anode loading is explained with the DIS model separator using a micropipette. The electrolyte consists of 1 M LiPF which incorporates the increased electrode surface area due to stress in a mixture of ethylene carbonate and diethyl carbonate (EC:DEC cracking. We propose that the lower capacity loss with increasing = 1:1, v/v) (Daikin Industries, Japan). The larger anode electrode anode loading can be explained by two possible stress-related mech- was placed on top of the separator and covered with a stainless steel anisms: (i) Increased anode loading results in decreasing in current spacer and spring. Finally, the cell was sealed by a hydraulic crimping density on the anode surface, which may reduce the electrode particle machine (National Research Council of Canada). cracking during cycling. (ii) Increased anode loading decreases anode Two sets of cells with different initial anode loadings and similar utilization and thus reduces the electrode particle expansion and con- cathode loadings were fabricated. The capacity ratio of anode to cath- traction during the respective lithiation and delithiation, thus reducing ode was calculated based on the initial weight of anode and cathode the mechanical degradation. material and their cycleable capacities (based on half-cell data). Spe- ciﬁc capacities of 320 mAh/g and 165 mAh/g for composite graphite and NMC, respectively, were used to calculate the R values. We note Experimental that the cycleable anode capacity available for lithiation may be less than the calculated value since the anode area is much larger than the Electrodes preparation.— All the powders were dried and stored cathode area. Hence, the R values are representative R values. The in argon-ﬁlled glove box before making electrodes. Anodes were anode and cathode loadings in Table I are calculated based on the mixtures of 87.8 wt% CGP-8 graphite carbon (Conoco Phillips), weights of the pure graphite and pure NMC in the respective compos- 2.9 wt% battery grade acetylene black (Denka Singapore Private ite electrodes and the electrode area of each of the punched electrode. Limited), and 9.3 wt% polyvinylidene ﬂuoride (PVDF) (No. 1100, A summary of the loading levels, electrode size, and thicknesses is Kureha, Japan). Cathodes were composed of 92.8 wt% NMC (Ap- provided in Table I. plied Materials, Inc.), 3.2 wt% acetylene black, and 4 wt% PVDF. In Table I, the values for surface of the graphite electrode for Anhydrous N-Methyl-2-pyrrolidone (NMP) (99.5%, Sigma-Aldrich) two cells are calculated based on the graphite loading and with the was used as solvent to dissolve PVDF. A slurry casting method de- assumption that the diameter of spherical graphite particle is 10 μm, veloped at the Lawrence Berkeley National Laboratory was used and density of graphite is known as 2.1 g/cm . to prepare all of the electrodes. For mixing powders, ﬁrstly, graphite/NMC and acetylene black were well mixed in NMP us- ing a homogenizer at 2500 rpm (Polytron PT10-35). Secondly, Electrochemical performance measurement.— Electrochemical PVDF was added to the slurry, and the whole slurry was blended tests were carried out in galvanostatic mode (constant current) on with the homogenizer again until it became uniform. Thirdly, the a Bio-Logic potentiostat (MPG-2) at 25 C controlled by an environ- uniform slurry was casted on battery grade Cu foil (thickness, mental chamber (Test Equity). All the cells were stabilized with 8 15 μm) for anodes and Al foil (thickness, 18 μm) for cathodes with a formation cycles between 4.2 V and 2.0 V at a slow rate of C/10. height adjustable doctor-blade (Yoshimitsu, model YOA-B). Different After that, all the cells were cycled between the same voltage limits loading densities of anode/cathode can be obtained by changing the at a rate of 1C. The capacities used for calculating currents for the 1C height of the doctor-blade. After drying overnight in the glove box, cycling were based on the discharge capacity of the eighth formation electrode laminates were calendered to approximately 37% porosity cycle. An open-circuit rest period of 15 minutes was also included using a roll press (Innovative Machine Corp.). Then anode disks with after every charge and discharge period. Values of the current densi- diameter of 1.43 cm and cathode disks with diameter of 1.27 cm were ties for the formation cycling and the long-term cycling for the two punched from the calendered laminates. All anode and cathode disks cells with different R values are summarized in Table II.The values were dried in a vacuum oven at 130 C for 15 h and then stored in of current densities mentioned in the Table II are based on the weights another argon-ﬁlled glove box for coin cells fabrication. of active materials (as against the weights of composite electrodes). Coin cell assembly.— Anodes and cathodes were assembled into Results and Discussion CR2325 coin-type cells. The smaller cathode electrode was placed at the bottom of the cell casing ﬁrst. Then one piece of polypropylene Formation cycles.— From Table II we see that, the current densities separator (Celgard 2400) was placed on top of the cathode electrode. A based on the active NMC weight in cathode (mA/g) are similar for ﬁxed amount (90 μL) of electrolyte was added to wet the cathode and both the cells during formation cycles. On the other hand, for the same Table II. Discharge/charge current densities. Formation cycling, C/10 Long-term cycling, 1C Current Anode Cathode Current Anode Cathode Capacity Ratio μA mA/g mA/g mA mA/g mA/g R = 1.18 143.4 30.9 17.7 1.2 258.1 148.5 R = 4.54 152.0 8.0 17.7 1.0 53.8 119.0 A2028 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) R=1.18 R=1.18 R=4.54 R=4.54 0 65 0123456 7 8 9 0123 45678 9 Cycle number Cycle number (a) (b) Figure 1. Electrochemical characteristics during formation cycling at C/10 of cells with two different anode loadings, (a) Speciﬁc capacities vs. cycle number, open marks indicate charge capacity, ﬁlled-in marks indicate discharge capacity. (b) Coulombic efﬁciency (discharge capacity divided by previous charge capacity) vs. cycle number. electrodes, current densities are noticeably different for the long-term the cell with the higher R delivers more discharge capacity than the cycling, even though both cells are cycled at the 1C rate. As mentioned, cell with lower R. Cycle-by-cycle capacities of both cells, normalized the 1C currents are based on the cell discharge capacity at the end of by their respective initial capacities, are plotted in Fig. 2b. It is evident the last formation cycle. The difference in the 1C current densities that the cell with the smaller R value had a faster capacity fade rate of the two cells is due to the reduced discharge capacity in a cell upon cycling as compared to the cell with the larger R value. with the larger anode after eight formation cycles. Speciﬁcally, at the Coulombic efﬁciency, cycle efﬁciency, charge/discharge endpoint last formation cycle, the cell discharge capacities are 119 mAh/g and marching.—In Fig. 2c the coulombic efﬁciencies of the two cells are 148 mAh/g (Fig. 1a) for the cells with R equal to 4.54 and 1.18, plotted as a function of cycle number. The cell with the lower anode respectively. This is understandable because cell with higher anode loading (R = 1.18) has a higher coulombic efﬁciency than the cell loading forms more SEI, which causes the higher-capacity anode to with the higher anode loading (R = 4.54). Compared with the capac- consume additional lithium from the cathode compared to the lower- ity retention plots (Fig. 2b), we observe that the cell with the lower capacity anode. coulombic efﬁciency possesses a higher capacity retention rate. In Fig. 1, the formation cycle charge and discharge capacities of The cycle efﬁciency is deﬁned as the ratio of the discharge ca- the two cells are plotted as a function of cycle number. In Fig. 1a, pacities of two successive cycles. For the cell with the lower anode the cell with R = 1.18 shows a ﬁrst cycle charge capacity of 171.8 loading (R = 1.18) the average cycle efﬁciency is calculated to be ca. mAh/g (black open diamond); the cell with R = 4.54 shows a ﬁrst 99.87% for 600 cycles; for the cell with higher anode loading (R = cycle charge capacity of 184.8 mAh/g (red open triangle). It is worth 4.54) it is ca. 99.95%. These numbers translate to cycling inefﬁcien- noting that, although the cathodes in both cells are very similar in cies of 0.13% and 0.05%. In other words, for the cell with R = 1.18, electrode weights, thicknesses, composition, and area, the ﬁrst cycle discharge capacity in every cycle is, on average 0.13% lower than the charge capacities of the two electrodes are signiﬁcantly different. The previous cycle and for the cell with R = 4.54, it is, on average 0.05% only difference in these cells is the difference in anode weights/surface lower than the previous cycle. Thus, the cell with the lower anode area. Hence, it is apparent that even though these cells are cathode loading (R = 1.18) is ca. 2 to 3 times less efﬁcient in cycling than capacity limited, the anode can affect the ﬁrst cycle charging capacity the cell with higher anode loading (R = 4.54). This is contrary to the of the cell. We attribute the higher ﬁrst cycle capacity of the cell conventional assumption that a higher coulombic efﬁciency implies a with the larger anode (R = 4.54) to additional electrolyte oxidation better cycle efﬁciency, i.e., longer cycle life. This indicates that the in the cell which contributes to the measured charge. We suggest that coulombic efﬁciency is not only a poor predictor of cycle life but, reaction products from the reduction of the electrolyte on the anode under certain conditions, the coulombic efﬁciency can be misleading. have crossed over to the cathode (cross-talk) where they are oxidized Many half and full cell studies in the literature focus on improving requiring additional charge. This is a further conﬁrmation of a ‘shuttle the coulombic efﬁciency of cells. It is important to realize that an im- reaction’ mechanism proposed earlier by Deshpande et al. This is proved coulombic efﬁciency may not always translate to an increased also reﬂected in the difference in the ﬁrst cycle coloumbic efﬁciency, cycle life as demonstrated in this experiment. 86% and 66% for R = 1.18 and 4.54 cells, respectively (Fig. 1b). After In Figure 2c, for both the cells, we note that there is a decrease the ﬁrst cycle, the negative electrode surface is mostly passivated and in coulombic efﬁciency of the cell in cycles immediately after the the coulombic efﬁciency quickly climbs to above 95% for both cells. C/10 cycles. The efﬁciency increases to a steady value after a few 1C cycles. This behavior of the coulombic efﬁciency has been observed earlier and investigation of this phenomenon is beyond the scope of Cell cycling study.—Capacity retention.— The capacity and this paper. coulombic efﬁciency during long-term cycling of the graphite/NMC To elucidate marching of charge and discharge endpoints, we plot cells with different R values are plotted in Fig set 2.Atthe startof in Fig. 2d typical voltage proﬁles of a cell during electrochemical cy- long-term cycling, the 1C discharge capacity is 135 mAh/g for the cell cling against the net charge/discharge capacity of the cell. As can be with R = 1.18 and 111 mAh/g for the cell with R = 4.54. Though the observed, during each cycle, both charge and discharge endpoints of capacity for the cell with R = 1.18 (diamond mark) starts at a higher the cell move to higher net capacity than the previous cycle. For exam- initial capacity, it decays at a faster rate as compared to the cell with R nd ple, the charge end point of the 2 cycle shown in red is at higher net = 4.54 (triangle mark). The two curves cross at 480 cycles after which Specific capacity, mAh/g Coulombic efficiency, % Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2029 160 120 R=1.18 R=1.18 R=4.54 R=4.54 80 60 0 0 0 200 400 600 800 1000 0 200 400 600 800 1000 Cycle number Cycle number (a) (b) R=1.18 R=4.54 0 200 400 600 800 1000 Cycle number (c) (d) R=1.18 R=4.54 Charge Slope: 0.4008 mA/g 550 Discharge 0.4308 mA/g Charge 0.0932 mA/g Discharge 0.1489 mA/g -50 0300 600 900 1200 1500 Time, h (e) Figure 2. (a) Speciﬁc discharge capacity vs. cycle number for cells with two different anode loadings during long-term cycling at 1C. (b) Discharge capacity retention (%) vs. cycle number for cells with two different anode loadings during long-term cycling at 1C. (c) Coulombic efﬁciency vs. cycle number for cells with two different anode loadings during long-term cycling at 1C. (d) Voltage proﬁles of a cell during cycling vs. net charge/discharge capacity with zoomed-in view of the charge endpoints of each cycle. (e) Net charge/discharge capacity vs. time for cells with two different anode loadings during long-term cycling at 1C. (Every 50 cycles, cells are cycled at C/10, these data are plotted as systematic interruptions in a, b, c and e). Coulombic efficiency, % Discharging capacity, mAh/g Net capacity, mAh/g Discharging capacity retention, % A2030 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) 0.45 R=1.18 R=4.54 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 200 400 600 800 1000 Cycle number (a) (b) (c) st th Figure 3. (a) Cell charge-discharge voltage proﬁles at 1C rate at 1 cycle and at 1000 cycle. The average difference in charge and discharge voltages changes as the cell is cycled. (b) The difference between the average cell voltage during charge and the average cell voltage during discharge for cell with R = 4.54 (red th th th triangles) and cell with R = 1.18 (blue squares) are plotted against the cycle number. (c) Cell voltage proﬁle during 1C charging for 10 , 500 and 1000 cycle plotted again charging time for a cell with R = 4.54 during charging till the voltage limit of 4.2 V is reached. The cell voltage is allowed to relax for 15 minutes once the cell reaches the voltage limit. st st th capacity than the charge end point of the 1 cycle shown in blue. This the 1 and the 1000 cycle against the capacity. A small increase in the shifting of charge/discharge endpoints is referred to as “marching of difference in charge discharge voltage proﬁles can be observed over charge/discharge endpoints”. Deshpande et al. found that coulom- 1000 cycles (V >V ). To clarify the contribution of resistance 2 1 bic efﬁciency and cycle efﬁciency are related to the marching of the rise in cells presented here, we adopt the same methodology presented charge and discharge endpoints. To analyze coulombic efﬁciency and in the earlier publication and observe the difference between average cycle efﬁciency for the two cells of different R values, the movement charge and discharge voltages as a function of cycle number (Fig. 3b). of the charge and discharge endpoints (i.e., c , d , respectively) Since, the polarization overvoltage depends on the cell state of R R are plotted as a function of time (Fig. 2e). The capacities plotted in charge the average change in the difference of charge and discharge Fig. 2e are the cumulative speciﬁc capacities based on initial cath- voltages over cycling gives an approximate value of capacity loss due ode loadings for both cells. The upper curves represent the charge to resistance rise. From Fig. 3b, it can be observed that, for the cell endpoints and lower curves represent the discharge endpoints as a with R = 1.18, the difference in the average voltages rises by approxi- function of time. Charge and discharge endpoints for both cells move mately 80 mV at the end of 1000 cycles as compared to the beginning to higher cumulative capacities during electrochemical cycling of the of cycling. Roughly half of that voltage difference rise occurs during cells as a result of side reactions. charging. Since the slope of the charging curve at the end of charge is In an earlier publication, Matthieu et al. observed a large increase 0.0088 V/(mAh/g), 40 mV of change in average charge voltage con- in resistance of cells over cycling. Resistance rise in the cell can tributes to 0.04/0.0088 = 4.5 mAh/g of capacity loss which is about affect accessible capacity of the cell as well as the rate of marching of 4.5% of the total capacity loss of 100 mAh/g in 1000 cycles for that both, the charge and the discharge endpoints. For the cells presented cell. A more accurate method to calculate the capacity loss due to re- here, we observe that the difference in charge and discharge voltage sistance rise would be to use the polarization overvoltage at the end of proﬁles increases with increasing cycle due to the increase in the ef- charge. We observe that the relaxation voltage after charging current fective resistance of the cell (Fig. 3a). As an example, in Fig. 3a we is stopped before next discharge is started. Figure 3c shows charging th th plot the charge-discharge voltage proﬁles for a cell with R = 4.54 on and relaxation voltage curves for a cell with R = 4.54 at 10 , 500 and Average (V ) - Average (V ), V ch dis Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2031 th start from a fully discharged point, are charged to a cutoff voltage and 1000 cycles. Figure 3c also zooms-in on the relaxation parts for these then fully discharged back to the next discharge point. The coulombic cycles to observe the difference in relaxation voltages as a function efﬁciency is the discharge capacity divided by the charge capacity, or of cycle number. We ﬁnd that most of the polarization overvoltage is another way to think of it, the coulombic inefﬁciency is the difference relaxed in 15 minutes once the charge current is stopped with dV/dt between where the two full discharges end when the voltage versus values less than 1 mV/min at the end of 15 minutes. We ﬁnd that for capacity is plotted in a positive direction on charge and then superim- the cell with R = 1.18 the polarization overvoltage increases from 55 posed on itself in the negative direction on discharge. Since the change mV to 83 mV (i.e. increase of 28 mV) in 1000 cycles. We realize that a in position of the discharge end points is are dictated solely by the side relaxation time of 15 minutes might not be sufﬁcient for complete re- reactions on the anode, when the cell is anode limited on discharge, laxation of the overvoltage, yet it’s a good approximation since there the coulombic efﬁciency is dictated solely by the side reactions on the is very small rate voltage relaxation at the end of 15 minutes. The increase in overvoltage of 28 mV corresponds to 3.2 mAh/gm of loss anode. of capacity due to overvoltage increase. Furthermore, the overvoltage d d Qc − Qd 1 d d n n of 28 mV found at the end of charge is consistent with the 40 mV of dN CE = = = 1 − [1] the average overvoltage calculated based on Fig. 3b. Both methods Qc Qc Qc dN n n n described above suggest that the resistance contribution to capacity In this equation, Qd is the discharge capacity of any cycle n and loss is only a small percentage (<5%) of the total capacity loss. th Qc is the charge capacity before the discharge in the n cycle. In Thus, 95% of the capacity fade is not due to resistance rise, but the given examples, since the cell with the higher anode loading (R due to the difference between the marching rates at each electrode, = 4.54) marches faster to the right than the cell with lower anode which in turn is caused by side reactions. This result is in accordance d d d d R=4.54 R=1.18 loading (R = 1.18) i.e. > , the cell with R = 4.54 with previous ﬁndings, the resistance rise is a small contributor to dt dt has a lower coulombic efﬁciency than the cell with the lower anode the capacity loss in these cells and so is the loss of active material loading (R = 1.18). Thus, the coulombic efﬁciency does not include from particle isolation. the rate of side reactions on the cathode. Hence, the marching of both the charge and discharge endpoints is mainly attributed to side reactions on the cathode and anode, Side reactions: Salt consumption or shuttle reactions?—Another way respectively. to view the rate of capacity fade in a cell is to look at the rate at which the charge and discharge endpoints are moving toward each other. Marching of the charge endpoints.—As the cell is cathode capacity If the difference between the anode and cathode marching rates, i.e. limited and the voltage of the anode is fairly ﬂat near the top of d d d c d d d c charge, the charge endpoint is reached as a result of the rise in the ( − )or( − ), is positive, the capacity of the cell is dN dN dt dt d d cathode voltage during its delithiation causing the cell to reach the declining at a rate equal to the difference. represents the shift in dN cutoff voltage. The sliding to the right of the charge endpoint, also d d the discharge endpoint capacity per cycle and represents the rate dt referred to as marching, is the result of electrolyte oxidation reactions of shift in discharge capacity endpoints with respect to time.Ifthere is on the cathode surface that occur since the last time the cell was fully little to no impedance rise in the cell and no loss of sites for lithiation charged. Intuitively, the amount of electrolyte oxidation reactions on and delithiation in either electrode, the difference in the marching rate the cathode should only depend on cathode loading/cathode surface of the anode and cathode is strictly determined by side reactions and area. Interestingly, for the cell with the lower anode loading (R = 1.18), can be equated to a rate of loss of lithium inventory (RLLI) between d c R=1.18 the charging endpoint moves more slowly to the right, = dt the electrodes. Using the latter ﬁgure of merit of loss per time provides −1 0.0932 mAg , than the cell with the higher anode loading (R = 4.54), a rate of the side reactions in mA per gram. For the cells presented in d c R=4.54 −1 = 0.4008 mAg . This observation strongly implies that the this paper, the rate of loss of lithium inventory with respect to time is, dt amount of electrolyte reduction on the anode is affecting the amount d d d c of electrolyte oxidation reaction on the cathode and demonstrates RLL I = − R=4.54 that there is cross-talk between the two electrodes, as discussed by dt dt R=4.54 Deshpande et al. −1 −1 −1 Marching of the discharge endpoints.—Near the end of discharge, = 0.4308 mAg − 0.4008 mAg = 0.0300 mAg the voltage curve of the anode rises steeply as its stored lithium is and depleted. In this region, the voltage proﬁle of the cathode is relatively ﬂat. Thus the discharge capacity is strongly limited by what is hap- d d d c RLL I = − pening at the anode. Any reduction of the electrolyte on the surface of R=1.18 dt dt R=1.18 anode further delithiates the anode and accelerates its discharge. This leads to a sliding or marching to the right of the discharge end point. −1 −1 −1 = 0.1489 mAg − 0.0932 mAg = 0.0557 mAg To avoid any confusion, we want to clarify that the movement of the discharge endpoint does not imply that the reactions are taking place [2] during discharge. But any reduction reactions on the anode during the subsequent charge and discharge result in the movement of the dis- To determine the average amount of capacity fade per cycle, one charge endpoint to right. In Marching of the charge endpoint section would multiply these numbers by the total amount of time it takes and here, the charge/discharge voltage versus capacity curves are plot- to charge and discharge the cell per cycle. The data provided in this ted such that the charge data goes from left to right, the discharge data manuscript shows that although the charge and discharge endpoints is plotted from right to left, and each starts where the other ends. From march faster when the anode loading is higher (R = 4.54), the rate of Fig. 2e it is clear that for the cell with the higher anode loading, the dis- capacity fade is actually lower since the difference in marching rates d d −1 R=4.54 of the anode and the cathode is smaller, i.e. the anode and the cathode charge endpoint moves faster = 0.4308 mAg to the right dt are marching more quickly but at nearly the same rates. For the cell d d R=1.18 −1 than the cell with lower anode loading, = 0.1489 mAg . dt with the lower anode loading (R = 1.18), even though the individual This observation can be understood because the larger the loading of rates of marching of the charge and discharge endpoints are slower, the negative electrode, the more area is available for electrolyte re- the difference between their marching rates is larger, resulting in a duction, and the greater the marching rate to the right of the discharge greater overall rate of cell capacity fade. endpoint. The impact of the difference in the rate of marching of the anode Coulombic efﬁciency.—As has been explained in Deshpande et al., (discharge endpoint) and the marching of the cathode (charge end th the coulombic efﬁciency CE of the n cycle of a cell is a function n point) is clear: there is a loss of cycleable capacity. But what additional d d of only the discharge endpoints marching . In other words, cells impact may the absolute rate of the side reactions have on cell life? dN A2032 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) There are four general ways to look at this which include cross talk, which is close to the density of EC at room temperature. With the NMC to be described through a general set of reactions. weight of 8.577 mg, the solvent loss is calculated to be approximately 0.48 μL. Since ca. 90 μL of electrolyte is added to the cell during Anode side reaction: the fabrication step, then the solvent loss with this mechanism is Solvent (S) is reduced on the anode and in the process consumes negligible, i.e. 0.53%. Similarly for the cell with R = 1.18, the solvent two Li ions to form some partially soluble component (SLi )which loss is 0.9 μL, which is again about only 1% of the total initial solvent is associated with SEI formation: in the system. In this scenario, there is no net loss of electrolyte salt. − − Hence, if the ﬁrst scenario is the most accurate description of what is S + 2e + 2Li P F → SLi + 2PF 6 2 6 happening in the cell, then there should be no long-term effects of the Considering cross talk, some of the SLi dissolves in the electrolyte side reactions on cell life. and diffuses to the cathode. When it reaches the cathode it does one ii. S + 2Li P F → S + 2Li P F Shuttle with loss of solvent 6 6 of three things, 1) it reacts reversibly and regenerates the Li and S, 2) it reacts irreversibly forming S but still regenerates the Li ,or For the cell where there is a shuttle but loss of solvent, scenario 2, 3) it reacts irreversibly and consumes PF . the rate of solvent loss is equal to the rate of the total rate of anode marching divided by two electrons per reaction. This is because sol- Possible cathode reactions: vent is lost for both types of reactions, i.e., the capacity loss reaction + − 1. SLi → S + 2Li + 2e as well as the shuttle reaction. For a cell with R = 4.54, the rate of + − d d −1 2. SLi → S + 2Li + 2e marching of the anode is = 0.4308 mAg . Following similar as- dt − − 3. SLi + 2PF → S(Li P F ) + 2e sumptions as above, in 1500h of cycling, the solvent loss is calculated 2 6 6 2 to be approximately 6.9 μL, which is about 7.7% of the total initial And although a strong argument has been put forth for crosstalk, d d −1 solvent in the system. For R = 1.18, with = 0.1489 mAg , dt it is possible that above and beyond the cross talk there is fresh solvent loss in 1500h via this mechanism (scenario 2) is predicted electrolyte oxidation to be 2.4 μL, which is about the 2.7% of the total initial solvent in + − the system. In this scenario, there is no net loss of electrolyte salt. 4. S + 2Li P F → S(PF ) + 2Li + 2e 6 6 2 Hence, if the second scenario is the most accurate, then there is a Although it is possible for reaction (4) to occur in the cell, it net loss of solvent as a result of the cross talk. Since there is a lot of cannot occur faster than the reaction on the anode, otherwise, the solvent compared to salt, this would also take a long time to reveal its cathode would be marching faster than the anode and the cell would repercussions (most likely catastrophic cell failure). temporarily show an increase in capacity until the cathode marched iii. S + 2Li P F → S(Li P F ) Shuttle with loss of electrolyte 6 6 off the anode and the anode displayed lithium deposition. It is not known if these reactions are one-or two-electron reactions but chose For the cell where there is an irreversible shuttle resulting in loss two-electron for these examples as this is generally assumed for most of electrolyte, scenario 3, the rate of solvent loss is equal to the rate of SEI formation reactions. the anode marching divided by two electrons per reaction similar to As a result of these four scenarios on the cathode, there are four that in scenario 2. In 1500h of cycling, the solvent loss is about 7.7% total reactions in the cell. for the cell with R = 4.54 and 2.7% for cell with R = 1.18. At the same time, in scenario 3, there is a loss of the electrolyte Total reactions: salt due to shuttle reaction as shown above. Since cathode marching is governed by the magnitude of shuttle reaction, the electrolyte salt 1. S + 2Li P F → S + 2Li P F Benign shuttle 6 6 lost is equivalent to the marching rate of the cathode (For R = 4.54, 2. S + 2Li P F → S + 2Li P F Shuttle with loss of solvent 6 6 d c −1 3. S + 2Li P F → S(Li P F ) Shuttle with loss of electrolyte = 0.4008 mAg ). Interestingly, in this scenario, the solvent 6 6 2 dt + − 4. x (S + 2Li ) + y(S + 2PF ) → loss is related to the anode marching but the electrolyte salt lost xSLi +yS(PF ) +2(y−x )e Loss of electrolyte with is equivalent to the marching rate of the cathode. This is because, 2 6 2 no shuttle solvent is lost in all of the side reactions in the cell which includes capacity loss reactions and the shuttle reaction, while electrolyte salt If the ﬁrst scenario is the most accurate of what is happening in is lost only due to shuttle reactions. For the cell with R = 4.54, using the cell, then there are no long-term effects of the side reactions. If the the cathode marching rate, in 1500h, the electrolyte salt lost can be second scenario is the most accurate, then there is a net loss of solvent −5 calculated to be 19.23 × 10 mol.Since 90 μLof1MLiPF salt as a result of the cross talk. Since there is a lot of solvent compared to solution is used, this equates to an initial lithium salt quantity of only salt, this may take a long time to reveal its repercussions (most likely −5 9 × 10 mol . Thus, this model of salt depletion with an irreversible catastrophic cell failure). If the third scenario is correct, then there shuttle reaction leads to an estimation of salt loss than is greater than will be a loss of salt as well. This could lead to an earlier death of the what is in the electrolyte when ﬁrst assembled. According to this + + cell as there is typically less Li in the electrolyte than there is Li model, the cell should have died well before the 1500h of cycling. in a fresh cathode. If the fourth scenario is the most accurate, then a This indicates that the assumption of irreversible salt consumption by lot of solvent would be lost in addition to the salt, although there is the electrolyte reactions does not represent the correct physics inside generally an excess of solvent when compared to salt in the cell. the cell. The total electrolyte consumption is dependent on the scenario + − iv. x (S + 2Li ) + y(S + 2PF ) → considered. For all of the cells, where the anode marches faster than the cathode and leads to a net capacity fade, only solvent is lost from xSLi + yS(PF ) + 2(y − x )e Loss of electrolyte with 2 6 2 that difference in marching rates, as the Li in the anode reaction is no shuttle supplied by the cathode keeping the LiPF concentration constant. If the cathode is also marching, the following can be said about the For the cell where there is solvent and electrolyte loss at both electrolyte. electrodes and no shuttle or crosstalk, scenario 4, the net loss of solvent is related to the sum of the marching of the anode and cathode i. S + 2Li P F → S + 2Li P F Benign shuttle 6 6 and the salt lost is related to the cathode marching (For R = 4.54, d c −1 d d −1 For the cell with the benign shuttle (scenario 1), the net solvent = 0.4008 mAg , = 0.4308 mAg ). For R = 4.54, in dt dt lost is equal to the net rate of capacity fade in the cell (RLLI) in mAh a 1500h of cycling, solvent lost is about 13.3 μL i.e. 14.8% of the divided by the two electrons involved in the reaction. For a cell with initial solvent. And the electrolyte salt lost for the cell R = 4.54 is −1 −5 R = 4.54, rate of capacity fade 0.03 mA g . For the approximate same as scenario 3, which is 19.23 × 10 mol . This model of salt −1 calculations we assume the density of the solvent to be 1.32 gmL depletion with no shuttle reaction again leads to estimation of salt Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2033 loss that is greater than that which is available. Since the cell is still ical degradation models can give one of the possible explanations of running after 1500h, we conclude that the model of ‘salt consumption the differences in distribution of the reaction products between dis- with no shuttle’ does not represent the correct physics inside the cell solvable and non-dissolvable products in the two cells. We hypothe- either. size that the reaction conditions on the bare electrode surface would With the analysis above, it seems shuttle reactions in either scenario be more favorable to non-dissolvable product formation (such as car- 1 or scenario 2 are most likely the dominant reactions in the current bonates) thus forming a more stable SEI ﬁlm. This is consistent with cells leading to marching of the endpoints in both electrodes, while the large ﬁrst cycle capacity loss observed in cells indicating that bare irreversible salt consumption by the electrolyte reactions is highly electrode surface favors non-dissolvable product formation (such as st unlikely. The increased 1 cycle charge capacity as a result of a larger lithium carbonate). We also observe that there is a four-fold increase −1 −1 anode also supports the hypothesis of cross talk between the two in rate of shuttle reaction (0.0932 mAg to 0.4008 mAg )asthe 2 2 electrodes with the likely hood of migration and reverse reactions of anode surface area increases by four-fold (13.26 cm to 54.20 cm ). the electrode reaction products SLi forming at the cathode. In the case It suggests that the side reactions forming dissolvable products are of reversible reactions, electrolyte salt and solvent are not consumed proportional to the total surface area of the negative electrodes. Since in the overall reaction, hence they do not lead to catastrophic cell after formation cycles, most of the electrode surface would be covered failure. with SEI, reactions through SEI seem to favor dissolvable products Though we do not have direct observation of SEI dissolving, the more. numbers above do indicate the possibility of such a phenomenon. As discussed earlier, the difference between anode marching and Apart from that, the indications of dissolution of side reaction prod- cathode marching leads to capacity loss in the battery. It’s well estab- ucts (or SEI) have been observed/proposed earlier by several research lished that in LIBs side reactions take place during cell storage as well 23–25 groups. In fact, a strong possibility of some of the side reac- as cycling. During cycling, DISs are postulated to assist side reactions tions being reversible has been expressed earlier. For example, CO by exposing new bare electrode surface to the electrolyte resulting in 4,28 is known to reduce to oxalate, formate and carbonate on negative additional capacity loss as compared to storage loss. DIS models electrode at normal conditions and oxalate/formate is easily oxidized provide an explanation for the faster capacity fade observed during 26,27 back to CO on the positive electrode. In fact, it’s postulated that cycling as compared to storage. Thus during cycling, the capacity CO undergoes reduction more easily than the other components of decay of a cell depends on the amount of new surface of the electrode the electrolyte. The cell with more anode (R = 4.54) material may particles exposed to the electrolyte during each cycle. The theory of have generated more CO during formation, which stays in the cell DIS can therefore provide a possible explanation for the higher rate and acts as a shuttle. of degradation in a cell with a lower anode loading (R = 1.18). Solid state diffusion of lithium atoms in and out of the host electrode parti- Capacity loss and mechanical degradation.—The data presented cles results in DISs in the electrode particles as a result of the volume shows that, though the coulombic efﬁciency of the cell with R = change that occurs during charge and discharge. Depending upon the 4.54 is less than the coulombic efﬁciency that of cell with R = 1.18, operating conditions, these stresses might have different effects on the capacity fade rate in the former cell is lower. the electrodes such as mechanical fatigue and fracture of the electrode Thus we conclude that there are two types of side reaction products: particles, electrode particle isolation from the composite matrix, or 13,14 1. Reaction products: Those which dissolve back into the electrolyte SEI fracture, to name some of the more popular explanations. solvents at the operating conditions get oxidized on the cathode As would be expected, the cell with the higher anode loading (R and contribute to shuttle reaction. = 4.54) that delivers a lower ﬁrst cycle discharge capacity than the 2. Reaction products: Those which do not dissolve back into the cell with the lower anode loading, would have less overall lithiation/ electrolyte solvents and thus remain attached to the electrode delithiation of the individual anode particles as compared to the parti- surface forming a stable SEI ﬁlm. cles in the electrode with the lower anode loading (R = 1.18). In Fig. 4, we plot differential voltage curves (dV/dQ) for the two cells again For example, as mentioned earlier, CO in the cell may reduce to graphite cell capacity per gram of graphite material, i.e., mAh/g of oxalate, formate and carbonate on the negative electrode at normal the graphite material. At the end of formation, the charge capacities, conditions. Among these products, oxalates and formates are easily based on the initial weight of the anodes are calculated to be 260 26,27 oxidized back to CO on the positive electrode while carbonates mAh/g and 55 mAh/g for R = 1.18 and 4.54, respectively. This is do not easily dissolve in the electrolyte solvents at normal operating clearly seen in the dV/dQ curves shown in Fig. 4, which indicate that conditions. the cell with R = 1.18 shows at least two distinct peaks corresponding to at least two different phase transitions in the graphite electrode, 0.03 last formation cycle 0.025 R=4.54 0.02 (Reactions above are reproduced from Ref. 27) 0.015 If the CO reduction on the anode results in the formation of reaction products such as carbonates, it will result in capacity loss. 0.01 This is because carbonates do not dissolve in the electrolyte solvents at the normal operating conditions and thus do not participate in shuttle R=1.18 mechanism. On the other hand, if CO reduction on the anode results 0.005 in oxalate/formate formation, these reaction products might dissolve back and get oxidized to CO on the cathode. From the data presented it’s clear that the distribution of dissolv- 0 50 100 150 200 250 able and non-dissolvable products in both cells is not the same. The Charge capacity, mAh/g cell with R = 1.18 has more capacity loss, thus relatively more non- dissolvable products and the cell with R = 4.54 has relatively more Figure 4. Differential voltage curve, dV/dQ vs. Q, for R = 1.18 and R = 4.54, dissolvable products resulting in lower coulombic efﬁciency. Mechan- for the charge process of last formation cycle. dV/dQ, V/(mAh/g) A2034 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) Figure 5. Schematic diagram showing capacity loss due to crack propagation on the negative electrode. whereas the dV/dQ plot of a cell with R = 4.54 has only one visible electrode particles i peak, which indicates that the cell with R = 4.54 does not undergo cell the same number of phase transitions as the cell with R = 1.18. This σ ∝ [5] means that although the anodes are both fully delithiated at the end A electrode of discharge, the anode particles in a cell with R = 4.54 are much Deshpande et al. also suggested that the capacity loss due to crack less lithiated on average at the end of charge. This also means that, propagation in the negative electrode particles is actually a function of during each cycle, the electrode particles with the higher R (R = 4.54) maximum stress on the electrode particles. The capacity loss increases undergo less expansion and contraction. Here we describe a possible with increased stress magnitude mechanism of capacity loss due to DISs. Electrode particle cracking due to mechanical fatigue.—Deshpande dQ et al. proposed that DISs the particles experience at the surface during = f (σ )[6] θ,max dN SE I f ormation due to crack propagation each cycle result in crack propagation due to mechanical fatigue. Fig. 5 shows schematically crack propagation as a possible mechanism for dQ Since i < i ⇒ σ < σ ⇒ < R=4.54 R=1.18 θ θ R=4.54 R=1.18 dN R=4.54 capacity loss in electrodes. dQ The ion ﬂux density (A/m ) on the graphite electrodes can be esti- dN R=1.18 Among the two types of cells described in this paper, anode par- mated based on the surface area of the graphite electrodes in both cells ticles of the cell with the lower anode loading have a much higher and assuming a uniform current distribution in both electrodes. For current density and thus experience much higher maximum tangential the cell with the higher R (R = 4.54), the average ion ﬂux density on −2 −2 stress. With the theory of fatigue fracture in solid mechanics, the crack the negative electrode is much smaller, 0.3 × 10 mA cm , as com- −2 −2 propagation is a function of the maximum stress that the material ex- pared to the cell with the lower R (R = 1.18), 2.28 × 10 mA cm . periences. Higher stress magnitude increases the tendency of cracks To understand the effect of current density on the electrode particle, propagation. Thus in case of battery electrodes, the larger the stress, previously developed DISs model is implemented. the more surface of the anode is exposed to the electrolyte solvent It is assumed that the diffusion in the nearly spherical anode par- and thus more active Li is consumed by the solvent reactions toward ticles during a galvanostatic charge condition can be described by re-passivating the surface. This is one possible explanation for the Fick’s law of diffusion. The DISs in the particle can be calculated 4,29 higher capacity fade in the cells with the lower anode loading (R = using the thermal stress analogy as has been previously published. 1.18). Cheng and Verbrugge showed the maximum stress in a cell is at We recognize that there may be other possible mechanisms causing the surface of the particles in the tangential direction at the start of differences in cell performances as there is no direct evidence for discharge. The magnitude of this stress (σ ) is described as cracking of the particles. A thorough investigation of the electrode ∞ 2 −λ τ surface may give a detailed understanding of this phenomenon. Such 1 E iR 1 e α α p σ = 1 − 2x + 2 an investigation is beyond the scope of this work. 3 (1 − ν ) FD 5 λ sin (λ ) α n n n=1 In summary, we believe there are two types of side reactions in Li-ion cells. The ﬁrst type of reaction results in the coverage of freshly sin (λ x ) sin (λ x ) − (λ x ) cos (λ x ) n n n n exposed anode surfaces with a SEI and occurs the ﬁrst time the anode × − [3] 3 3 λ x λ x n reaches low voltages and also occurs where there are cracks in the SEI that expose fresh surface. Thus, electrodes with more cracks have where E is the Young’s modulus and ν is the Poisson’s ratio of α α more of this type of side reaction. The second type occurs as a result the electrode material, is the partial molar volume of the solute, of dissolution of the SEI that occurs at low state of charges (SOCs) R is the electrode particle radius, D is the diffusion coefﬁcient of but is reformed at high SOCs. The dissolved SEI diffuses across the the solute inside electrode particle, F is Faraday’s constant, x = cell where it is oxidized resulting in no loss of capacity. Thus, a cell r/R , τ = tD/R , r is any radial location from the center, t is p p with a large overall amount of anode material will experience a lower time during lithiation or delithiation, positive current densities i over coulombic efﬁciency because it has much more SEI that can then the particle surface denote charging, and λ (n = 1, 2, 3,...) are the dissolve and subsequently reformed, where the dissolved components positive roots of tan(λ ) = λ . For a thin ﬁlm electrode with a speciﬁc n n migrate to the cathode where it is reduced but does not result in surface area a , thickness L , and geometric area A and the cell current greater capacity fade. A cell with less anode material will experience I , the current density over a particle surface i can be calculated as cell higher current density at the particle level that results in diffusion induced cracking of either the SEI or particles which requires solvent cell i = [4] consumption and loss of lithium inventory in the cell and relative electrode marching of the anode versus the cathode. This work complements the work of Deshpande et al. where it was also hypothesized that here A is the electrode surface exposed to the electrolyte. electrode VC slows the rate of the second type of reaction but has no effect on At any given characteristic time τ, the tangential tensile stress σ the ﬁrst in a graphite/NMC cell. on the particle surface is proportional to the current density over the Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2035 Conclusions r, θ, ϕ spherical coordinates partial molar volume of the solute in phase ‘i’(m /mol ) Graphite/NMC full cells with two different anode loading lev- R radius of the spherical electrode particle (m) els were tested to establish a better understanding of coulombic ef- ν Poisson’s ratio of phase ‘i’ ﬁciency and cell capacity fade rate. The cell with a higher anode E Young’s modulus of phase ‘i’(N /m ) st i loading demonstrated more 1 cycle charging capacity. A cross-talk σ tangential stress (N /m ) between the anode and cathode is postulated to be the reason behind α 3 C (r, t ) solute concentration at radius r at time t (mol/m ) this additional capacity. The cell with the higher anode loading also C initial concentration of solute in active core phase experienced a greater loss of cycleable capacity upon completion of (mol/m ) the formation process. The capacity loss during the formation pro- C swing in state of litigation of the ‘i’ phase avg cess is directly related to the total surface area of anode exposed to D diffusion coefﬁcient of the solute in α phase (m /sec) the electrolyte. The cell with the higher anode loading demonstrated i current density on the electrode particle (A/m ) poorer coulombic efﬁciency during long-term cycling. By following I current applied to the cell (A) cell the charge and discharge endpoints, it is shown that the coulombic A electrode surface area (m ) electrode efﬁciency of a cell is a measure of the marching rate of the discharge F Faraday constant (C) endpoint, which for both types of cells was the anode. Any increase Q cell capacity (C) in the amount of reduction reactions causes the anode to meet the dis- x dimensionless radius charge limits earlier and thus increase its marching rate. The higher τ dimensionless time marching rate of anode for a cell with higher anode loading (R = 4.54) dQ rate of loss of discharge capacity is assigned to the greater surface area of the anode and the hypothesis dN that the SEI is constantly dissolving and reforming based on SOC. Consequently, by changing the anode loading, the marching rate of the charge end point, for which the cells were cathode limited, also References changed. Increased marching of the cathode due to increased loading 1. D. Linden and T. B. Reddy, Handbook of batteries, 3rd ed. (McGraw-Hill, New York, of anode strongly suggests that the amount of electrolyte oxidation on 2002). the cathode surface is dependent on the amount of anode present in the 2. M. Armand and J. M. Tarascon, Nature 451(7179), 652 (2008). cell. Through some simple calculations, it was shown that the salt was 3. B. Dunn, H. Kamath, and J. M. Tarascon, Science 334(6058), 928 (2011). not consumed at the full rate of the side reactions on both electrodes 4. R. Deshpande, M. Verbrugge, Y. T. Cheng, J. Wang, and P. Liu, Journal of the Electrochemical Society 159(10), A1730 (2012). or the cells would consumed all of the salt in the cells much sooner 5. D. Aurbach, J Power Sources 89(2), 206 (2000). than when they actually reached the end of test. Thus, we re-afﬁrm 6. E. Markervich, G. Salitra, M. D. Levi, and D. Aurbach, Journal of Power Sources the hypotheses that a shuttle reaction between the anode and cathode 146(1), 146 (2005). occurs without the consumption of electrolytic salt. We believe that 7. J. Vetter, P. Nova ´k,M. R.Wagner, C. Veit,K. C.Moller, ¨ J. O. Besenhard, M. Winter, M. Wohlfahrt-Mehrens, C. Vogler, and A. Hammouche, Journal of Power Sources the electrode reaction products that make up the SEI on the anode, 147(1), 269 (2005). SLi , dissolve in the electrolyte, migrate to the cathode and undergo a 8. M. Winter, J. O. Besenhard, M. E. Spahr, and P. Novak, Adv. Mater. 10(10), 725 reaction that recovers the salt. Such a shuttle, reversible or not, does (1998). not lead to capacity loss in the short term. 9. M. Wachtler, J. O. Besenhard, and M. Winter, Journal of Power Sources 94(2), 189 (2001). Consistent with one of the main conclusions in Reference 12, 10. J. Yan, B. J. Xia, Y. C. Su, X. Z. Zhou, J. Zhang, and X. G. Zhang, Electrochimica we establish that the coulombic efﬁciency is not a good indicator of Acta 53(24), 7069 (2008). battery capacity failure. Additionally, we demonstrate that a cell with 11. J. C. Burns, N. N. Sinha, D. J. Coyle, G. Jain, C. M. VanElzen, W. M. Lamanna, higher coulombic efﬁciency can possibly have a higher capacity fade A. Xiao, E. Scott, J. P. Gardner, and J. R. Dahn, Journal of the Electrochemical Society 159(2), A85 (2012). rate. 12. R. D. Deshpande, P. Ridgway, Y. B. Fu, W. Zhang, J. S. Cai, and V. Battaglia, Journal of The Electrochemical Society 162(3), A330 (2015). 13. J. Christensen and J. Newman, J. Solid State Electr 10(5), 293 (2006). Acknowledgments 14. R. D. Deshpande, University of Kentucky, Lexington, Kentucky (USA), 2011. 15. R. Deshpande, Y. T. Cheng, and M. W. Verbrugge, Journal of Power Sources 195(15), The authors would like to express gratitude to the entire battery 5081 (2010). research members of Dr. Battaglia and Dr. Gao Liu in the Electrochem- 16. R. Deshpande, Y. Qi, and Y. T. Cheng, Journal of the Electrochemical Society 157(8), ical Technologies Group of Lawrence Berkeley National Lab. Special A967 (2010). thanks to Dr. Yanbao Fu, Xiangyun Song, and visiting student Min 17. V. A. Sethuraman, N. Van Winkle, D. P. Abraham, A. F. Bower, and P. R. Guduru, Journal of Power Sources 206, 334 (2012). Ling, who were very helpful with the running of the experiments. The 18. P. Liu, J. Wang, J. Hicks-Garner, E. Sherman, S. Soukiazian, M. Verbrugge, H. Tataria, authors acknowledge the BATT Program and the U.S. Department of J. Musser, and P. Finamore, J Electrochem Soc 157(4), A499 (2010). Energy, as well as National Science Foundation Award No. 1355438 19. J. Li, A. K. Dozier, Y. Li, F. Yang, and Y.-T. Cheng, Journal of The Electrochemical (Powering the Kentucky Bioeconomy for a Sustainable Future) for Society 158(6), A689 (2011). 20. H. Zheng, G. Liu, S. Carwford, and V. S. Battaglia, (http://bestar.lbl.gov/ partially funding this project. vbattaglia/cell-analysis-tools/). 21. M. Dubarry and B. Y. Liaw, Journal of Power Sources 194(1), 541 (2009). 22. V. Pop, H. Bergveld, J. O. het Veld, P. Regtien, D. Danilov, and P. Notten, Journal of The Electrochemical Society 153(11), A2013 (2006). List of Symbols 23. R. Spotnitz, Journal of Power Sources 113(1), 72 (2003). R ratio of initial anode capacity to cathode capacity 24. R. Yazami and Y. F. Reynier, Electrochimica Acta 47(8), 1217 (2002). d c 25. M. Lu, H. Cheng, and Y. Yang, Electrochimica Acta 53(9), 3539 (2008). rate of marching of charge endpoint (A/g) dt 26. S. E. Sloop, J. B. Kerr, and K. Kinoshita, Journal of Power Sources 119–121(0), 330 d d (2003). rate of marching of discharge endpoint (A/g) dt 27. A. Gennaro, A. A. Isse, J.-M. Saveant, ´ M.-G. Severin, and E. Vianello, Journal of CE coulombic efﬁciency the American Chemical Society 118(30), 7190 (1996). th Qd discharge capacity of n cycle (Ah) 28. M. Safari and C. Delacourt, Journal of The Electrochemical Society 158(10), A1123 th Qc charge capacity of n cycle (Ah) n (2011). 29. Y.-T. Cheng and M. W. Verbrugge, Journal of Power Sources 190(2), 453 (2009). RLL I rate of loss of lithium inventory (A/g) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of the Electrochemical Society IOP Publishing http://www.deepdyve.com/lp/iop-publishing/electrode-side-reactions-capacity-loss-and-mechanical-degradation-in-mo3qjFAhoo

Loading next page...

References (26)

M. Wachtler, J. Besenhard, M. Winter (2001)
Tin and tin-based intermetallics as new anode materials for lithium-ion cells
Journal of Power Sources, 94
Mi Lu, Hu Cheng, Yong Yang (2008)
A comparison of solid electrolyte interphase (SEI) on the artificial graphite anode of the aged and cycled commercial lithium ion cells
Electrochimica Acta, 53
Yang-Tse Cheng, M. Verbrugge (2009)
Evolution of stress within a spherical insertion electrode particle under potentiostatic and galvanostatic operation
Journal of Power Sources, 190
M. Armand, J. Tarascon (2008)
Building better batteries
Nature, 451
M. Safari, C. Delacourt (2011)
Aging of a Commercial Graphite/LiFePO4 Cell
Journal of The Electrochemical Society, 158
Jian Yan, B. Xia, Yuchang Su, Xiao-Zhong Zhou, Jian Zhang, Xigui Zhang (2008)
Phenomenologically modeling the formation and evolution of the solid electrolyte interface on the graphite electrode for lithium-ion batteries
Electrochimica Acta, 53
A. Gennaro, A. Isse, b Savéant, A. Severin, A. Vianello (1996)
Homogeneous Electron Transfer Catalysis of the Electrochemical Reduction of Carbon Dioxide. Do Aromatic Anion Radicals React in an Outer-Sphere Manner?
Journal of the American Chemical Society, 118
R. Spotnitz (2003)
Simulation of capacity fade in lithium-ion batteries
Journal of Power Sources, 113
B. Dunn, H. Kamath, J. Tarascon (2011)
Electrical Energy Storage for the Grid: A Battery of Choices
Science, 334
R. Deshpande, Y. Qi, Yang-Tse Cheng (2010)
Effects of Concentration-Dependent Elastic Modulus on Diffusion-Induced Stresses for Battery Applications
Journal of The Electrochemical Society, 157
D. Aurbach (2000)
Review of selected electrode–solution interactions which determine the performance of Li and Li ion batteries
Journal of Power Sources, 89
M. Dubarry, B. Liaw (2009)
Identify capacity fading mechanism in a commercial LiFePO4 cell
Journal of Power Sources, 194
R. Deshpande, M. Verbrugge, Yang-Tse Cheng, John Wang, Ping Liu (2012)
Battery Cycle Life Prediction with Coupled Chemical Degradation and Fatigue Mechanics
Journal of The Electrochemical Society, 159
R. Yazami, Y. Reynier (2002)
Mechanism of self-discharge in graphite–lithium anode
Electrochimica Acta, 47
J. Burns, N. Sinha, D. Coyle, Gaurav Jain, Collette VanElzen, W. Lamanna, A. Xiao, E. Scott, J. Gardner, J. Dahn (2011)
The Impact of Varying the Concentration of Vinylene Carbonate Electrolyte Additive in Wound Li-Ion Cells
Journal of The Electrochemical Society, 159
John Christensen, J. Newman (2005)
Stress generation and fracture in lithium insertion materials
Journal of Solid State Electrochemistry, 10
R. Deshpande, P. Ridgway, Yan-bao Fu, Wei Zhang, J. Cai, V. Battaglia (2015)
The Limited Effect of VC in Graphite/NMC Cells
Journal of The Electrochemical Society, 162
J. Vetter, P. Novák, M. Wagner, C. Veit, K. Möller, J. Besenhard, M. Winter, M. Wohlfahrt‐Mehrens, C. Vogler, A. Hammouche (2005)
Ageing mechanisms in lithium-ion batteries
Journal of Power Sources, 147
M. Winter, Jürgen Besenhard, M. Spahr, P. Novák (1998)
Insertion Electrode Materials for Rechargeable Lithium Batteries
Advanced Materials, 10
Juchuan Li, A. Dozier, Yunchao Li, Fuqian Yang, Yang-Tse Cheng (2011)
Crack Pattern Formation in Thin Film Lithium-Ion Battery Electrodes
Journal of The Electrochemical Society, 158
E. Markervich, G. Salitra, M. Levi, D. Aurbach (2005)
Capacity fading of lithiated graphite electrodes studied by a combination of electroanalytical methods, Raman spectroscopy and SEM
Journal of Power Sources, 146
R. Deshpande, Yang-Tse Cheng, M. Verbrugge (2010)
Modeling diffusion-induced stress in nanowire electrode structures
Journal of Power Sources, 195
S. Sloop, J. Kerr, K. Kinoshita (2003)
The role of Li-ion battery electrolyte reactivity in performance decline and self-discharge
Journal of Power Sources, 119
V. Pop, H. Bergveld, J. Veld, P. Regtien, D. Danilov, P. Notten (2006)
Modeling Battery Behavior for Accurate State-of-Charge Indication
Journal of The Electrochemical Society, 153
V. Sethuraman, N. Winkle, D. Abraham, A. Bower, P. Guduru (2012)
Real-Time Stress Measurements in Lithium-ion Battery Negative-electrodes
Journal of Power Sources, 206
Ping Liu, John Wang, Jocelyn Hicks-Garner, E. Sherman, Souren Soukiazian, M. Verbrugge, H. Tataria, James Musser, Peter Finamore (2010)
Aging Mechanisms of LiFePO4 Batteries Deduced by Electrochemical and Structural Analyses
Journal of The Electrochemical Society, 157

Copyright: Copyright © The Author(s) 2015. Published by ECS.
ISSN: 0013-4651
eISSN: 1945-7111
DOI: 10.1149/2.0291510jes
Publisher site: See Article on Publisher Site

Abstract

A2026 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) Electrode Side Reactions, Capacity Loss and Mechanical Degradation in Lithium-Ion Batteries a,= b,=,z a,∗ a Jiagang Xu, Rutooj D. Deshpande, Jie Pan, Yang-Tse Cheng, c,∗∗ and Vincent S. Battaglia Department of Chemical & Materials Engineering, University of Kentucky, Lexington, Kentucky 40506, USA Electriﬁed Powertrain Engineering, Ford Motor Company, Dearborn, Michigan 48124, USA Environmental Energy Technologies Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA For advancing lithium-ion battery (LIB) technologies, a detailed understanding of battery degradation mechanisms is important. In this article, experimental observations are provided to elucidate the relation between side reactions, mechanical degradation, and capacity loss in LIBs. Graphite/Li(Ni Mn Co )O cells of two very different initial anode/cathode capacity ratios (R, both R > 1/3 1/3 1/3 2 1) are assembled to investigate the electrochemical behavior. The initial charge capacity of the cathode is observed to be affected by the anode loading, indicating that the electrolyte reactions on the anode affect the electrolyte reactions on the cathode. Additionally, the rate of “marching” of the cathode is found to be affected by the anode loading. These ﬁndings attest to the “cross-talk” between the two electrodes. During cycling, the cell with the higher R value display a lower columbic efﬁciency, yet a lower capacity fade rate as compared to the cell with the smaller R. This supports the notion that columbic efﬁciency is not a perfect predictor of capacity fade. Capacity loss is attributed to the irreversible production of new solid electrolyte interphase (SEI) facilitated by the mechanical degradation of the SEI. The higher capacity fade in the cell with the lower R is explained with the theory of diffusion-induced stresses (DISs). © The Author(s) 2015. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0291510jes] All rights reserved. Manuscript submitted March 3, 2015; revised manuscript received July 13, 2015. Published July 28, 2015. Lithium-ion batteries (LIBs) are widely used in small, portable, tal observations of DISs and stress induced cracking of electrode electronic devices due to their high energy density, high voltage, low materials. For example, Sethuraman et al. experimentally measured 1,2 self-discharge rate, and good cycle performance. Yet, for automo- the stresses in thin ﬁlm electrodes using wafer curvature methods. In tive propulsion applications, LIBs need improvement in volumetric another publication, Li et al. observed the cracking pattern in silicon and gravimetric energy density and with performance over the life thin ﬁlm electrodes and correlated it with the electrode thickness. of the vehicle. This is one of the major motivations of researchers In this research, an attempt is made to understand the relation- worldwide focusing on developing cheaper and more durable LIBs ship between coulombic efﬁciency, capacity retention, and mechani- for applications such as hybrid vehicles, electric vehicles, and other cal stresses. Two types of graphite/Li(Ni Mn Co )O (NMC) full 1/3 1/3 1/3 2 large-scale energy storage systems. cells were designed with different anode to cathode capacity ratios Capacity decay during storage and charge-discharge cycling is (R). The cathode loading was the same in both cells, thus R was ad- one of the major challenges that limit the life of LIBs. Instability of justed by changing the anode loading. Both cells were charge limited the electrolyte at the operating potentials results in side reactions on by the amount of lithium that could be removed from the cathode at a the electrode surfaces, part of which lead to the formation of solid given cutoff voltage and discharge limited by amount of lithium could 4–7 electrolyte interphases (SEIs). Side reactions may result in lower be removed from the anode. We found that, during electrochemical coulombic efﬁciency, loss of usable capacity of the cell, and increase cycling, the cell with the higher anode loading (i.e., high R value) of cell impedance. Cell performance degradation due to side reactions had much lower coulombic efﬁciency as compared to the cell with the is termed as chemical degradation, which is known to be the main lower anode loading (i.e., small R value). Monitoring of the charge 5,6,8–10 cause of lithium loss in well-made LIBs. These side reactions and discharge endpoints vs. time revealed that the marching rate of 11,12 lead to marching of the charge and discharge endpoints. Such side the charge endpoint was affected by the amount of anode loading for a reactions and their effects on the cell life are relatively less explored. ﬁxed cathode loading. We propose that the electrolyte reactions on the The solid state diffusion of lithium atoms in and out of host elec- cathode are affected by the electrolyte reactions on the anode. Thus, trode particles results in diffusion induced stresses (DISs) and volume there is a “cross-talk” between the two electrodes. Interestingly, the changes in electrode particles during charge and discharge. Depending cell with the higher R value suffered a slower rate of capacity fade upon the operating conditions these stresses may have various effects during cycling, though its total rate of side reactions was greater. Here, on the electrodes such as mechanical fatigue and fracture of the elec- the total rate of side reactions is deﬁned as the sum of marching rate trode particles, electrode particle isolation from the composite matrix, of discharge and charge end points. These observations support recent 13,14 12 or SEI fracture, to name a few. DISs have been studied in detail by ﬁndings presented in Deshpande et al., where it was demonstrated mathematical modeling and experimental observations. Previous re- through the use of an electrolyte additive that coulombic efﬁciency is ports have discussed mechanical stresses in single electrode particles not a perfect indicator of battery life. Thus, a cell with higher rates 13,15–17 and thin ﬁlm electrodes. Apart from the loss of active electrode of side reactions can have less capacity loss than a cell of the same material, mechanical degradation of electrode particles facilitates ad- chemistry with lower rates of total side reactions, which supports the ditional side reactions by exposing new surface area to the electrolyte. premise that not all side reactions lead to capacity fade in a battery, Various mathematical models have been developed to explain the ca- at least in the early stages of cycle life. We propose that there are two pacity loss in batteries during electrochemical cycling with coupled types of side reactions: one which causes lithium loss from the system 4,18 chemical and mechanical degradation as a prevailing mechanism. and one which shuttles between the two electrodes without actually Along with mathematical models there have been several experimen- leading to irreversible lithium loss. Based on the cell test data, we conclude that the reactions caus- ing the irreversible lithium loss in the system are mainly linked to These authors contributed equally to this work. the formation of additional SEI on newly exposed electrode surfaces Electrochemical Society Student Member. on the anode. We develop a DIS model correlating electrode volume ∗∗ Electrochemical Society Active Member. E-mail: [email protected] expansion with the capacity loss of the cell. Lower capacity loss in Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2027 Table I. Anode/Cathode pair fabricated. Anode loading, Cathode loading, Calculated graphite 2 2 ∗ ∗ 2 mg/cm mg/cm Anode thickness, Cathode thickness, surface area, cm Capacity Ratio ( = 14.3 mm) ( = 12.7 mm) μm μm (approximation) R = 1.18 2.89 6.37 41 42 13.26 R = 4.54 11.83 6.77 71 42 54.20 - Diameter of electrode in coin cell. Thickness is measured after calendering; thickness of Cu/Al foil is included. The thickness shown here is the representative average value of the laminate from which the disk electrodes are punched. the cell with higher anode loading is explained with the DIS model separator using a micropipette. The electrolyte consists of 1 M LiPF which incorporates the increased electrode surface area due to stress in a mixture of ethylene carbonate and diethyl carbonate (EC:DEC cracking. We propose that the lower capacity loss with increasing = 1:1, v/v) (Daikin Industries, Japan). The larger anode electrode anode loading can be explained by two possible stress-related mech- was placed on top of the separator and covered with a stainless steel anisms: (i) Increased anode loading results in decreasing in current spacer and spring. Finally, the cell was sealed by a hydraulic crimping density on the anode surface, which may reduce the electrode particle machine (National Research Council of Canada). cracking during cycling. (ii) Increased anode loading decreases anode Two sets of cells with different initial anode loadings and similar utilization and thus reduces the electrode particle expansion and con- cathode loadings were fabricated. The capacity ratio of anode to cath- traction during the respective lithiation and delithiation, thus reducing ode was calculated based on the initial weight of anode and cathode the mechanical degradation. material and their cycleable capacities (based on half-cell data). Spe- ciﬁc capacities of 320 mAh/g and 165 mAh/g for composite graphite and NMC, respectively, were used to calculate the R values. We note Experimental that the cycleable anode capacity available for lithiation may be less than the calculated value since the anode area is much larger than the Electrodes preparation.— All the powders were dried and stored cathode area. Hence, the R values are representative R values. The in argon-ﬁlled glove box before making electrodes. Anodes were anode and cathode loadings in Table I are calculated based on the mixtures of 87.8 wt% CGP-8 graphite carbon (Conoco Phillips), weights of the pure graphite and pure NMC in the respective compos- 2.9 wt% battery grade acetylene black (Denka Singapore Private ite electrodes and the electrode area of each of the punched electrode. Limited), and 9.3 wt% polyvinylidene ﬂuoride (PVDF) (No. 1100, A summary of the loading levels, electrode size, and thicknesses is Kureha, Japan). Cathodes were composed of 92.8 wt% NMC (Ap- provided in Table I. plied Materials, Inc.), 3.2 wt% acetylene black, and 4 wt% PVDF. In Table I, the values for surface of the graphite electrode for Anhydrous N-Methyl-2-pyrrolidone (NMP) (99.5%, Sigma-Aldrich) two cells are calculated based on the graphite loading and with the was used as solvent to dissolve PVDF. A slurry casting method de- assumption that the diameter of spherical graphite particle is 10 μm, veloped at the Lawrence Berkeley National Laboratory was used and density of graphite is known as 2.1 g/cm . to prepare all of the electrodes. For mixing powders, ﬁrstly, graphite/NMC and acetylene black were well mixed in NMP us- ing a homogenizer at 2500 rpm (Polytron PT10-35). Secondly, Electrochemical performance measurement.— Electrochemical PVDF was added to the slurry, and the whole slurry was blended tests were carried out in galvanostatic mode (constant current) on with the homogenizer again until it became uniform. Thirdly, the a Bio-Logic potentiostat (MPG-2) at 25 C controlled by an environ- uniform slurry was casted on battery grade Cu foil (thickness, mental chamber (Test Equity). All the cells were stabilized with 8 15 μm) for anodes and Al foil (thickness, 18 μm) for cathodes with a formation cycles between 4.2 V and 2.0 V at a slow rate of C/10. height adjustable doctor-blade (Yoshimitsu, model YOA-B). Different After that, all the cells were cycled between the same voltage limits loading densities of anode/cathode can be obtained by changing the at a rate of 1C. The capacities used for calculating currents for the 1C height of the doctor-blade. After drying overnight in the glove box, cycling were based on the discharge capacity of the eighth formation electrode laminates were calendered to approximately 37% porosity cycle. An open-circuit rest period of 15 minutes was also included using a roll press (Innovative Machine Corp.). Then anode disks with after every charge and discharge period. Values of the current densi- diameter of 1.43 cm and cathode disks with diameter of 1.27 cm were ties for the formation cycling and the long-term cycling for the two punched from the calendered laminates. All anode and cathode disks cells with different R values are summarized in Table II.The values were dried in a vacuum oven at 130 C for 15 h and then stored in of current densities mentioned in the Table II are based on the weights another argon-ﬁlled glove box for coin cells fabrication. of active materials (as against the weights of composite electrodes). Coin cell assembly.— Anodes and cathodes were assembled into Results and Discussion CR2325 coin-type cells. The smaller cathode electrode was placed at the bottom of the cell casing ﬁrst. Then one piece of polypropylene Formation cycles.— From Table II we see that, the current densities separator (Celgard 2400) was placed on top of the cathode electrode. A based on the active NMC weight in cathode (mA/g) are similar for ﬁxed amount (90 μL) of electrolyte was added to wet the cathode and both the cells during formation cycles. On the other hand, for the same Table II. Discharge/charge current densities. Formation cycling, C/10 Long-term cycling, 1C Current Anode Cathode Current Anode Cathode Capacity Ratio μA mA/g mA/g mA mA/g mA/g R = 1.18 143.4 30.9 17.7 1.2 258.1 148.5 R = 4.54 152.0 8.0 17.7 1.0 53.8 119.0 A2028 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) R=1.18 R=1.18 R=4.54 R=4.54 0 65 0123456 7 8 9 0123 45678 9 Cycle number Cycle number (a) (b) Figure 1. Electrochemical characteristics during formation cycling at C/10 of cells with two different anode loadings, (a) Speciﬁc capacities vs. cycle number, open marks indicate charge capacity, ﬁlled-in marks indicate discharge capacity. (b) Coulombic efﬁciency (discharge capacity divided by previous charge capacity) vs. cycle number. electrodes, current densities are noticeably different for the long-term the cell with the higher R delivers more discharge capacity than the cycling, even though both cells are cycled at the 1C rate. As mentioned, cell with lower R. Cycle-by-cycle capacities of both cells, normalized the 1C currents are based on the cell discharge capacity at the end of by their respective initial capacities, are plotted in Fig. 2b. It is evident the last formation cycle. The difference in the 1C current densities that the cell with the smaller R value had a faster capacity fade rate of the two cells is due to the reduced discharge capacity in a cell upon cycling as compared to the cell with the larger R value. with the larger anode after eight formation cycles. Speciﬁcally, at the Coulombic efﬁciency, cycle efﬁciency, charge/discharge endpoint last formation cycle, the cell discharge capacities are 119 mAh/g and marching.—In Fig. 2c the coulombic efﬁciencies of the two cells are 148 mAh/g (Fig. 1a) for the cells with R equal to 4.54 and 1.18, plotted as a function of cycle number. The cell with the lower anode respectively. This is understandable because cell with higher anode loading (R = 1.18) has a higher coulombic efﬁciency than the cell loading forms more SEI, which causes the higher-capacity anode to with the higher anode loading (R = 4.54). Compared with the capac- consume additional lithium from the cathode compared to the lower- ity retention plots (Fig. 2b), we observe that the cell with the lower capacity anode. coulombic efﬁciency possesses a higher capacity retention rate. In Fig. 1, the formation cycle charge and discharge capacities of The cycle efﬁciency is deﬁned as the ratio of the discharge ca- the two cells are plotted as a function of cycle number. In Fig. 1a, pacities of two successive cycles. For the cell with the lower anode the cell with R = 1.18 shows a ﬁrst cycle charge capacity of 171.8 loading (R = 1.18) the average cycle efﬁciency is calculated to be ca. mAh/g (black open diamond); the cell with R = 4.54 shows a ﬁrst 99.87% for 600 cycles; for the cell with higher anode loading (R = cycle charge capacity of 184.8 mAh/g (red open triangle). It is worth 4.54) it is ca. 99.95%. These numbers translate to cycling inefﬁcien- noting that, although the cathodes in both cells are very similar in cies of 0.13% and 0.05%. In other words, for the cell with R = 1.18, electrode weights, thicknesses, composition, and area, the ﬁrst cycle discharge capacity in every cycle is, on average 0.13% lower than the charge capacities of the two electrodes are signiﬁcantly different. The previous cycle and for the cell with R = 4.54, it is, on average 0.05% only difference in these cells is the difference in anode weights/surface lower than the previous cycle. Thus, the cell with the lower anode area. Hence, it is apparent that even though these cells are cathode loading (R = 1.18) is ca. 2 to 3 times less efﬁcient in cycling than capacity limited, the anode can affect the ﬁrst cycle charging capacity the cell with higher anode loading (R = 4.54). This is contrary to the of the cell. We attribute the higher ﬁrst cycle capacity of the cell conventional assumption that a higher coulombic efﬁciency implies a with the larger anode (R = 4.54) to additional electrolyte oxidation better cycle efﬁciency, i.e., longer cycle life. This indicates that the in the cell which contributes to the measured charge. We suggest that coulombic efﬁciency is not only a poor predictor of cycle life but, reaction products from the reduction of the electrolyte on the anode under certain conditions, the coulombic efﬁciency can be misleading. have crossed over to the cathode (cross-talk) where they are oxidized Many half and full cell studies in the literature focus on improving requiring additional charge. This is a further conﬁrmation of a ‘shuttle the coulombic efﬁciency of cells. It is important to realize that an im- reaction’ mechanism proposed earlier by Deshpande et al. This is proved coulombic efﬁciency may not always translate to an increased also reﬂected in the difference in the ﬁrst cycle coloumbic efﬁciency, cycle life as demonstrated in this experiment. 86% and 66% for R = 1.18 and 4.54 cells, respectively (Fig. 1b). After In Figure 2c, for both the cells, we note that there is a decrease the ﬁrst cycle, the negative electrode surface is mostly passivated and in coulombic efﬁciency of the cell in cycles immediately after the the coulombic efﬁciency quickly climbs to above 95% for both cells. C/10 cycles. The efﬁciency increases to a steady value after a few 1C cycles. This behavior of the coulombic efﬁciency has been observed earlier and investigation of this phenomenon is beyond the scope of Cell cycling study.—Capacity retention.— The capacity and this paper. coulombic efﬁciency during long-term cycling of the graphite/NMC To elucidate marching of charge and discharge endpoints, we plot cells with different R values are plotted in Fig set 2.Atthe startof in Fig. 2d typical voltage proﬁles of a cell during electrochemical cy- long-term cycling, the 1C discharge capacity is 135 mAh/g for the cell cling against the net charge/discharge capacity of the cell. As can be with R = 1.18 and 111 mAh/g for the cell with R = 4.54. Though the observed, during each cycle, both charge and discharge endpoints of capacity for the cell with R = 1.18 (diamond mark) starts at a higher the cell move to higher net capacity than the previous cycle. For exam- initial capacity, it decays at a faster rate as compared to the cell with R nd ple, the charge end point of the 2 cycle shown in red is at higher net = 4.54 (triangle mark). The two curves cross at 480 cycles after which Specific capacity, mAh/g Coulombic efficiency, % Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2029 160 120 R=1.18 R=1.18 R=4.54 R=4.54 80 60 0 0 0 200 400 600 800 1000 0 200 400 600 800 1000 Cycle number Cycle number (a) (b) R=1.18 R=4.54 0 200 400 600 800 1000 Cycle number (c) (d) R=1.18 R=4.54 Charge Slope: 0.4008 mA/g 550 Discharge 0.4308 mA/g Charge 0.0932 mA/g Discharge 0.1489 mA/g -50 0300 600 900 1200 1500 Time, h (e) Figure 2. (a) Speciﬁc discharge capacity vs. cycle number for cells with two different anode loadings during long-term cycling at 1C. (b) Discharge capacity retention (%) vs. cycle number for cells with two different anode loadings during long-term cycling at 1C. (c) Coulombic efﬁciency vs. cycle number for cells with two different anode loadings during long-term cycling at 1C. (d) Voltage proﬁles of a cell during cycling vs. net charge/discharge capacity with zoomed-in view of the charge endpoints of each cycle. (e) Net charge/discharge capacity vs. time for cells with two different anode loadings during long-term cycling at 1C. (Every 50 cycles, cells are cycled at C/10, these data are plotted as systematic interruptions in a, b, c and e). Coulombic efficiency, % Discharging capacity, mAh/g Net capacity, mAh/g Discharging capacity retention, % A2030 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) 0.45 R=1.18 R=4.54 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 200 400 600 800 1000 Cycle number (a) (b) (c) st th Figure 3. (a) Cell charge-discharge voltage proﬁles at 1C rate at 1 cycle and at 1000 cycle. The average difference in charge and discharge voltages changes as the cell is cycled. (b) The difference between the average cell voltage during charge and the average cell voltage during discharge for cell with R = 4.54 (red th th th triangles) and cell with R = 1.18 (blue squares) are plotted against the cycle number. (c) Cell voltage proﬁle during 1C charging for 10 , 500 and 1000 cycle plotted again charging time for a cell with R = 4.54 during charging till the voltage limit of 4.2 V is reached. The cell voltage is allowed to relax for 15 minutes once the cell reaches the voltage limit. st st th capacity than the charge end point of the 1 cycle shown in blue. This the 1 and the 1000 cycle against the capacity. A small increase in the shifting of charge/discharge endpoints is referred to as “marching of difference in charge discharge voltage proﬁles can be observed over charge/discharge endpoints”. Deshpande et al. found that coulom- 1000 cycles (V >V ). To clarify the contribution of resistance 2 1 bic efﬁciency and cycle efﬁciency are related to the marching of the rise in cells presented here, we adopt the same methodology presented charge and discharge endpoints. To analyze coulombic efﬁciency and in the earlier publication and observe the difference between average cycle efﬁciency for the two cells of different R values, the movement charge and discharge voltages as a function of cycle number (Fig. 3b). of the charge and discharge endpoints (i.e., c , d , respectively) Since, the polarization overvoltage depends on the cell state of R R are plotted as a function of time (Fig. 2e). The capacities plotted in charge the average change in the difference of charge and discharge Fig. 2e are the cumulative speciﬁc capacities based on initial cath- voltages over cycling gives an approximate value of capacity loss due ode loadings for both cells. The upper curves represent the charge to resistance rise. From Fig. 3b, it can be observed that, for the cell endpoints and lower curves represent the discharge endpoints as a with R = 1.18, the difference in the average voltages rises by approxi- function of time. Charge and discharge endpoints for both cells move mately 80 mV at the end of 1000 cycles as compared to the beginning to higher cumulative capacities during electrochemical cycling of the of cycling. Roughly half of that voltage difference rise occurs during cells as a result of side reactions. charging. Since the slope of the charging curve at the end of charge is In an earlier publication, Matthieu et al. observed a large increase 0.0088 V/(mAh/g), 40 mV of change in average charge voltage con- in resistance of cells over cycling. Resistance rise in the cell can tributes to 0.04/0.0088 = 4.5 mAh/g of capacity loss which is about affect accessible capacity of the cell as well as the rate of marching of 4.5% of the total capacity loss of 100 mAh/g in 1000 cycles for that both, the charge and the discharge endpoints. For the cells presented cell. A more accurate method to calculate the capacity loss due to re- here, we observe that the difference in charge and discharge voltage sistance rise would be to use the polarization overvoltage at the end of proﬁles increases with increasing cycle due to the increase in the ef- charge. We observe that the relaxation voltage after charging current fective resistance of the cell (Fig. 3a). As an example, in Fig. 3a we is stopped before next discharge is started. Figure 3c shows charging th th plot the charge-discharge voltage proﬁles for a cell with R = 4.54 on and relaxation voltage curves for a cell with R = 4.54 at 10 , 500 and Average (V ) - Average (V ), V ch dis Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2031 th start from a fully discharged point, are charged to a cutoff voltage and 1000 cycles. Figure 3c also zooms-in on the relaxation parts for these then fully discharged back to the next discharge point. The coulombic cycles to observe the difference in relaxation voltages as a function efﬁciency is the discharge capacity divided by the charge capacity, or of cycle number. We ﬁnd that most of the polarization overvoltage is another way to think of it, the coulombic inefﬁciency is the difference relaxed in 15 minutes once the charge current is stopped with dV/dt between where the two full discharges end when the voltage versus values less than 1 mV/min at the end of 15 minutes. We ﬁnd that for capacity is plotted in a positive direction on charge and then superim- the cell with R = 1.18 the polarization overvoltage increases from 55 posed on itself in the negative direction on discharge. Since the change mV to 83 mV (i.e. increase of 28 mV) in 1000 cycles. We realize that a in position of the discharge end points is are dictated solely by the side relaxation time of 15 minutes might not be sufﬁcient for complete re- reactions on the anode, when the cell is anode limited on discharge, laxation of the overvoltage, yet it’s a good approximation since there the coulombic efﬁciency is dictated solely by the side reactions on the is very small rate voltage relaxation at the end of 15 minutes. The increase in overvoltage of 28 mV corresponds to 3.2 mAh/gm of loss anode. of capacity due to overvoltage increase. Furthermore, the overvoltage d d Qc − Qd 1 d d n n of 28 mV found at the end of charge is consistent with the 40 mV of dN CE = = = 1 − [1] the average overvoltage calculated based on Fig. 3b. Both methods Qc Qc Qc dN n n n described above suggest that the resistance contribution to capacity In this equation, Qd is the discharge capacity of any cycle n and loss is only a small percentage (<5%) of the total capacity loss. th Qc is the charge capacity before the discharge in the n cycle. In Thus, 95% of the capacity fade is not due to resistance rise, but the given examples, since the cell with the higher anode loading (R due to the difference between the marching rates at each electrode, = 4.54) marches faster to the right than the cell with lower anode which in turn is caused by side reactions. This result is in accordance d d d d R=4.54 R=1.18 loading (R = 1.18) i.e. > , the cell with R = 4.54 with previous ﬁndings, the resistance rise is a small contributor to dt dt has a lower coulombic efﬁciency than the cell with the lower anode the capacity loss in these cells and so is the loss of active material loading (R = 1.18). Thus, the coulombic efﬁciency does not include from particle isolation. the rate of side reactions on the cathode. Hence, the marching of both the charge and discharge endpoints is mainly attributed to side reactions on the cathode and anode, Side reactions: Salt consumption or shuttle reactions?—Another way respectively. to view the rate of capacity fade in a cell is to look at the rate at which the charge and discharge endpoints are moving toward each other. Marching of the charge endpoints.—As the cell is cathode capacity If the difference between the anode and cathode marching rates, i.e. limited and the voltage of the anode is fairly ﬂat near the top of d d d c d d d c charge, the charge endpoint is reached as a result of the rise in the ( − )or( − ), is positive, the capacity of the cell is dN dN dt dt d d cathode voltage during its delithiation causing the cell to reach the declining at a rate equal to the difference. represents the shift in dN cutoff voltage. The sliding to the right of the charge endpoint, also d d the discharge endpoint capacity per cycle and represents the rate dt referred to as marching, is the result of electrolyte oxidation reactions of shift in discharge capacity endpoints with respect to time.Ifthere is on the cathode surface that occur since the last time the cell was fully little to no impedance rise in the cell and no loss of sites for lithiation charged. Intuitively, the amount of electrolyte oxidation reactions on and delithiation in either electrode, the difference in the marching rate the cathode should only depend on cathode loading/cathode surface of the anode and cathode is strictly determined by side reactions and area. Interestingly, for the cell with the lower anode loading (R = 1.18), can be equated to a rate of loss of lithium inventory (RLLI) between d c R=1.18 the charging endpoint moves more slowly to the right, = dt the electrodes. Using the latter ﬁgure of merit of loss per time provides −1 0.0932 mAg , than the cell with the higher anode loading (R = 4.54), a rate of the side reactions in mA per gram. For the cells presented in d c R=4.54 −1 = 0.4008 mAg . This observation strongly implies that the this paper, the rate of loss of lithium inventory with respect to time is, dt amount of electrolyte reduction on the anode is affecting the amount d d d c of electrolyte oxidation reaction on the cathode and demonstrates RLL I = − R=4.54 that there is cross-talk between the two electrodes, as discussed by dt dt R=4.54 Deshpande et al. −1 −1 −1 Marching of the discharge endpoints.—Near the end of discharge, = 0.4308 mAg − 0.4008 mAg = 0.0300 mAg the voltage curve of the anode rises steeply as its stored lithium is and depleted. In this region, the voltage proﬁle of the cathode is relatively ﬂat. Thus the discharge capacity is strongly limited by what is hap- d d d c RLL I = − pening at the anode. Any reduction of the electrolyte on the surface of R=1.18 dt dt R=1.18 anode further delithiates the anode and accelerates its discharge. This leads to a sliding or marching to the right of the discharge end point. −1 −1 −1 = 0.1489 mAg − 0.0932 mAg = 0.0557 mAg To avoid any confusion, we want to clarify that the movement of the discharge endpoint does not imply that the reactions are taking place [2] during discharge. But any reduction reactions on the anode during the subsequent charge and discharge result in the movement of the dis- To determine the average amount of capacity fade per cycle, one charge endpoint to right. In Marching of the charge endpoint section would multiply these numbers by the total amount of time it takes and here, the charge/discharge voltage versus capacity curves are plot- to charge and discharge the cell per cycle. The data provided in this ted such that the charge data goes from left to right, the discharge data manuscript shows that although the charge and discharge endpoints is plotted from right to left, and each starts where the other ends. From march faster when the anode loading is higher (R = 4.54), the rate of Fig. 2e it is clear that for the cell with the higher anode loading, the dis- capacity fade is actually lower since the difference in marching rates d d −1 R=4.54 of the anode and the cathode is smaller, i.e. the anode and the cathode charge endpoint moves faster = 0.4308 mAg to the right dt are marching more quickly but at nearly the same rates. For the cell d d R=1.18 −1 than the cell with lower anode loading, = 0.1489 mAg . dt with the lower anode loading (R = 1.18), even though the individual This observation can be understood because the larger the loading of rates of marching of the charge and discharge endpoints are slower, the negative electrode, the more area is available for electrolyte re- the difference between their marching rates is larger, resulting in a duction, and the greater the marching rate to the right of the discharge greater overall rate of cell capacity fade. endpoint. The impact of the difference in the rate of marching of the anode Coulombic efﬁciency.—As has been explained in Deshpande et al., (discharge endpoint) and the marching of the cathode (charge end th the coulombic efﬁciency CE of the n cycle of a cell is a function n point) is clear: there is a loss of cycleable capacity. But what additional d d of only the discharge endpoints marching . In other words, cells impact may the absolute rate of the side reactions have on cell life? dN A2032 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) There are four general ways to look at this which include cross talk, which is close to the density of EC at room temperature. With the NMC to be described through a general set of reactions. weight of 8.577 mg, the solvent loss is calculated to be approximately 0.48 μL. Since ca. 90 μL of electrolyte is added to the cell during Anode side reaction: the fabrication step, then the solvent loss with this mechanism is Solvent (S) is reduced on the anode and in the process consumes negligible, i.e. 0.53%. Similarly for the cell with R = 1.18, the solvent two Li ions to form some partially soluble component (SLi )which loss is 0.9 μL, which is again about only 1% of the total initial solvent is associated with SEI formation: in the system. In this scenario, there is no net loss of electrolyte salt. − − Hence, if the ﬁrst scenario is the most accurate description of what is S + 2e + 2Li P F → SLi + 2PF 6 2 6 happening in the cell, then there should be no long-term effects of the Considering cross talk, some of the SLi dissolves in the electrolyte side reactions on cell life. and diffuses to the cathode. When it reaches the cathode it does one ii. S + 2Li P F → S + 2Li P F Shuttle with loss of solvent 6 6 of three things, 1) it reacts reversibly and regenerates the Li and S, 2) it reacts irreversibly forming S but still regenerates the Li ,or For the cell where there is a shuttle but loss of solvent, scenario 2, 3) it reacts irreversibly and consumes PF . the rate of solvent loss is equal to the rate of the total rate of anode marching divided by two electrons per reaction. This is because sol- Possible cathode reactions: vent is lost for both types of reactions, i.e., the capacity loss reaction + − 1. SLi → S + 2Li + 2e as well as the shuttle reaction. For a cell with R = 4.54, the rate of + − d d −1 2. SLi → S + 2Li + 2e marching of the anode is = 0.4308 mAg . Following similar as- dt − − 3. SLi + 2PF → S(Li P F ) + 2e sumptions as above, in 1500h of cycling, the solvent loss is calculated 2 6 6 2 to be approximately 6.9 μL, which is about 7.7% of the total initial And although a strong argument has been put forth for crosstalk, d d −1 solvent in the system. For R = 1.18, with = 0.1489 mAg , dt it is possible that above and beyond the cross talk there is fresh solvent loss in 1500h via this mechanism (scenario 2) is predicted electrolyte oxidation to be 2.4 μL, which is about the 2.7% of the total initial solvent in + − the system. In this scenario, there is no net loss of electrolyte salt. 4. S + 2Li P F → S(PF ) + 2Li + 2e 6 6 2 Hence, if the second scenario is the most accurate, then there is a Although it is possible for reaction (4) to occur in the cell, it net loss of solvent as a result of the cross talk. Since there is a lot of cannot occur faster than the reaction on the anode, otherwise, the solvent compared to salt, this would also take a long time to reveal its cathode would be marching faster than the anode and the cell would repercussions (most likely catastrophic cell failure). temporarily show an increase in capacity until the cathode marched iii. S + 2Li P F → S(Li P F ) Shuttle with loss of electrolyte 6 6 off the anode and the anode displayed lithium deposition. It is not known if these reactions are one-or two-electron reactions but chose For the cell where there is an irreversible shuttle resulting in loss two-electron for these examples as this is generally assumed for most of electrolyte, scenario 3, the rate of solvent loss is equal to the rate of SEI formation reactions. the anode marching divided by two electrons per reaction similar to As a result of these four scenarios on the cathode, there are four that in scenario 2. In 1500h of cycling, the solvent loss is about 7.7% total reactions in the cell. for the cell with R = 4.54 and 2.7% for cell with R = 1.18. At the same time, in scenario 3, there is a loss of the electrolyte Total reactions: salt due to shuttle reaction as shown above. Since cathode marching is governed by the magnitude of shuttle reaction, the electrolyte salt 1. S + 2Li P F → S + 2Li P F Benign shuttle 6 6 lost is equivalent to the marching rate of the cathode (For R = 4.54, 2. S + 2Li P F → S + 2Li P F Shuttle with loss of solvent 6 6 d c −1 3. S + 2Li P F → S(Li P F ) Shuttle with loss of electrolyte = 0.4008 mAg ). Interestingly, in this scenario, the solvent 6 6 2 dt + − 4. x (S + 2Li ) + y(S + 2PF ) → loss is related to the anode marching but the electrolyte salt lost xSLi +yS(PF ) +2(y−x )e Loss of electrolyte with is equivalent to the marching rate of the cathode. This is because, 2 6 2 no shuttle solvent is lost in all of the side reactions in the cell which includes capacity loss reactions and the shuttle reaction, while electrolyte salt If the ﬁrst scenario is the most accurate of what is happening in is lost only due to shuttle reactions. For the cell with R = 4.54, using the cell, then there are no long-term effects of the side reactions. If the the cathode marching rate, in 1500h, the electrolyte salt lost can be second scenario is the most accurate, then there is a net loss of solvent −5 calculated to be 19.23 × 10 mol.Since 90 μLof1MLiPF salt as a result of the cross talk. Since there is a lot of solvent compared to solution is used, this equates to an initial lithium salt quantity of only salt, this may take a long time to reveal its repercussions (most likely −5 9 × 10 mol . Thus, this model of salt depletion with an irreversible catastrophic cell failure). If the third scenario is correct, then there shuttle reaction leads to an estimation of salt loss than is greater than will be a loss of salt as well. This could lead to an earlier death of the what is in the electrolyte when ﬁrst assembled. According to this + + cell as there is typically less Li in the electrolyte than there is Li model, the cell should have died well before the 1500h of cycling. in a fresh cathode. If the fourth scenario is the most accurate, then a This indicates that the assumption of irreversible salt consumption by lot of solvent would be lost in addition to the salt, although there is the electrolyte reactions does not represent the correct physics inside generally an excess of solvent when compared to salt in the cell. the cell. The total electrolyte consumption is dependent on the scenario + − iv. x (S + 2Li ) + y(S + 2PF ) → considered. For all of the cells, where the anode marches faster than the cathode and leads to a net capacity fade, only solvent is lost from xSLi + yS(PF ) + 2(y − x )e Loss of electrolyte with 2 6 2 that difference in marching rates, as the Li in the anode reaction is no shuttle supplied by the cathode keeping the LiPF concentration constant. If the cathode is also marching, the following can be said about the For the cell where there is solvent and electrolyte loss at both electrolyte. electrodes and no shuttle or crosstalk, scenario 4, the net loss of solvent is related to the sum of the marching of the anode and cathode i. S + 2Li P F → S + 2Li P F Benign shuttle 6 6 and the salt lost is related to the cathode marching (For R = 4.54, d c −1 d d −1 For the cell with the benign shuttle (scenario 1), the net solvent = 0.4008 mAg , = 0.4308 mAg ). For R = 4.54, in dt dt lost is equal to the net rate of capacity fade in the cell (RLLI) in mAh a 1500h of cycling, solvent lost is about 13.3 μL i.e. 14.8% of the divided by the two electrons involved in the reaction. For a cell with initial solvent. And the electrolyte salt lost for the cell R = 4.54 is −1 −5 R = 4.54, rate of capacity fade 0.03 mA g . For the approximate same as scenario 3, which is 19.23 × 10 mol . This model of salt −1 calculations we assume the density of the solvent to be 1.32 gmL depletion with no shuttle reaction again leads to estimation of salt Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2033 loss that is greater than that which is available. Since the cell is still ical degradation models can give one of the possible explanations of running after 1500h, we conclude that the model of ‘salt consumption the differences in distribution of the reaction products between dis- with no shuttle’ does not represent the correct physics inside the cell solvable and non-dissolvable products in the two cells. We hypothe- either. size that the reaction conditions on the bare electrode surface would With the analysis above, it seems shuttle reactions in either scenario be more favorable to non-dissolvable product formation (such as car- 1 or scenario 2 are most likely the dominant reactions in the current bonates) thus forming a more stable SEI ﬁlm. This is consistent with cells leading to marching of the endpoints in both electrodes, while the large ﬁrst cycle capacity loss observed in cells indicating that bare irreversible salt consumption by the electrolyte reactions is highly electrode surface favors non-dissolvable product formation (such as st unlikely. The increased 1 cycle charge capacity as a result of a larger lithium carbonate). We also observe that there is a four-fold increase −1 −1 anode also supports the hypothesis of cross talk between the two in rate of shuttle reaction (0.0932 mAg to 0.4008 mAg )asthe 2 2 electrodes with the likely hood of migration and reverse reactions of anode surface area increases by four-fold (13.26 cm to 54.20 cm ). the electrode reaction products SLi forming at the cathode. In the case It suggests that the side reactions forming dissolvable products are of reversible reactions, electrolyte salt and solvent are not consumed proportional to the total surface area of the negative electrodes. Since in the overall reaction, hence they do not lead to catastrophic cell after formation cycles, most of the electrode surface would be covered failure. with SEI, reactions through SEI seem to favor dissolvable products Though we do not have direct observation of SEI dissolving, the more. numbers above do indicate the possibility of such a phenomenon. As discussed earlier, the difference between anode marching and Apart from that, the indications of dissolution of side reaction prod- cathode marching leads to capacity loss in the battery. It’s well estab- ucts (or SEI) have been observed/proposed earlier by several research lished that in LIBs side reactions take place during cell storage as well 23–25 groups. In fact, a strong possibility of some of the side reac- as cycling. During cycling, DISs are postulated to assist side reactions tions being reversible has been expressed earlier. For example, CO by exposing new bare electrode surface to the electrolyte resulting in 4,28 is known to reduce to oxalate, formate and carbonate on negative additional capacity loss as compared to storage loss. DIS models electrode at normal conditions and oxalate/formate is easily oxidized provide an explanation for the faster capacity fade observed during 26,27 back to CO on the positive electrode. In fact, it’s postulated that cycling as compared to storage. Thus during cycling, the capacity CO undergoes reduction more easily than the other components of decay of a cell depends on the amount of new surface of the electrode the electrolyte. The cell with more anode (R = 4.54) material may particles exposed to the electrolyte during each cycle. The theory of have generated more CO during formation, which stays in the cell DIS can therefore provide a possible explanation for the higher rate and acts as a shuttle. of degradation in a cell with a lower anode loading (R = 1.18). Solid state diffusion of lithium atoms in and out of the host electrode parti- Capacity loss and mechanical degradation.—The data presented cles results in DISs in the electrode particles as a result of the volume shows that, though the coulombic efﬁciency of the cell with R = change that occurs during charge and discharge. Depending upon the 4.54 is less than the coulombic efﬁciency that of cell with R = 1.18, operating conditions, these stresses might have different effects on the capacity fade rate in the former cell is lower. the electrodes such as mechanical fatigue and fracture of the electrode Thus we conclude that there are two types of side reaction products: particles, electrode particle isolation from the composite matrix, or 13,14 1. Reaction products: Those which dissolve back into the electrolyte SEI fracture, to name some of the more popular explanations. solvents at the operating conditions get oxidized on the cathode As would be expected, the cell with the higher anode loading (R and contribute to shuttle reaction. = 4.54) that delivers a lower ﬁrst cycle discharge capacity than the 2. Reaction products: Those which do not dissolve back into the cell with the lower anode loading, would have less overall lithiation/ electrolyte solvents and thus remain attached to the electrode delithiation of the individual anode particles as compared to the parti- surface forming a stable SEI ﬁlm. cles in the electrode with the lower anode loading (R = 1.18). In Fig. 4, we plot differential voltage curves (dV/dQ) for the two cells again For example, as mentioned earlier, CO in the cell may reduce to graphite cell capacity per gram of graphite material, i.e., mAh/g of oxalate, formate and carbonate on the negative electrode at normal the graphite material. At the end of formation, the charge capacities, conditions. Among these products, oxalates and formates are easily based on the initial weight of the anodes are calculated to be 260 26,27 oxidized back to CO on the positive electrode while carbonates mAh/g and 55 mAh/g for R = 1.18 and 4.54, respectively. This is do not easily dissolve in the electrolyte solvents at normal operating clearly seen in the dV/dQ curves shown in Fig. 4, which indicate that conditions. the cell with R = 1.18 shows at least two distinct peaks corresponding to at least two different phase transitions in the graphite electrode, 0.03 last formation cycle 0.025 R=4.54 0.02 (Reactions above are reproduced from Ref. 27) 0.015 If the CO reduction on the anode results in the formation of reaction products such as carbonates, it will result in capacity loss. 0.01 This is because carbonates do not dissolve in the electrolyte solvents at the normal operating conditions and thus do not participate in shuttle R=1.18 mechanism. On the other hand, if CO reduction on the anode results 0.005 in oxalate/formate formation, these reaction products might dissolve back and get oxidized to CO on the cathode. From the data presented it’s clear that the distribution of dissolv- 0 50 100 150 200 250 able and non-dissolvable products in both cells is not the same. The Charge capacity, mAh/g cell with R = 1.18 has more capacity loss, thus relatively more non- dissolvable products and the cell with R = 4.54 has relatively more Figure 4. Differential voltage curve, dV/dQ vs. Q, for R = 1.18 and R = 4.54, dissolvable products resulting in lower coulombic efﬁciency. Mechan- for the charge process of last formation cycle. dV/dQ, V/(mAh/g) A2034 Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) Figure 5. Schematic diagram showing capacity loss due to crack propagation on the negative electrode. whereas the dV/dQ plot of a cell with R = 4.54 has only one visible electrode particles i peak, which indicates that the cell with R = 4.54 does not undergo cell the same number of phase transitions as the cell with R = 1.18. This σ ∝ [5] means that although the anodes are both fully delithiated at the end A electrode of discharge, the anode particles in a cell with R = 4.54 are much Deshpande et al. also suggested that the capacity loss due to crack less lithiated on average at the end of charge. This also means that, propagation in the negative electrode particles is actually a function of during each cycle, the electrode particles with the higher R (R = 4.54) maximum stress on the electrode particles. The capacity loss increases undergo less expansion and contraction. Here we describe a possible with increased stress magnitude mechanism of capacity loss due to DISs. Electrode particle cracking due to mechanical fatigue.—Deshpande dQ et al. proposed that DISs the particles experience at the surface during = f (σ )[6] θ,max dN SE I f ormation due to crack propagation each cycle result in crack propagation due to mechanical fatigue. Fig. 5 shows schematically crack propagation as a possible mechanism for dQ Since i < i ⇒ σ < σ ⇒ < R=4.54 R=1.18 θ θ R=4.54 R=1.18 dN R=4.54 capacity loss in electrodes. dQ The ion ﬂux density (A/m ) on the graphite electrodes can be esti- dN R=1.18 Among the two types of cells described in this paper, anode par- mated based on the surface area of the graphite electrodes in both cells ticles of the cell with the lower anode loading have a much higher and assuming a uniform current distribution in both electrodes. For current density and thus experience much higher maximum tangential the cell with the higher R (R = 4.54), the average ion ﬂux density on −2 −2 stress. With the theory of fatigue fracture in solid mechanics, the crack the negative electrode is much smaller, 0.3 × 10 mA cm , as com- −2 −2 propagation is a function of the maximum stress that the material ex- pared to the cell with the lower R (R = 1.18), 2.28 × 10 mA cm . periences. Higher stress magnitude increases the tendency of cracks To understand the effect of current density on the electrode particle, propagation. Thus in case of battery electrodes, the larger the stress, previously developed DISs model is implemented. the more surface of the anode is exposed to the electrolyte solvent It is assumed that the diffusion in the nearly spherical anode par- and thus more active Li is consumed by the solvent reactions toward ticles during a galvanostatic charge condition can be described by re-passivating the surface. This is one possible explanation for the Fick’s law of diffusion. The DISs in the particle can be calculated 4,29 higher capacity fade in the cells with the lower anode loading (R = using the thermal stress analogy as has been previously published. 1.18). Cheng and Verbrugge showed the maximum stress in a cell is at We recognize that there may be other possible mechanisms causing the surface of the particles in the tangential direction at the start of differences in cell performances as there is no direct evidence for discharge. The magnitude of this stress (σ ) is described as cracking of the particles. A thorough investigation of the electrode ∞ 2 −λ τ surface may give a detailed understanding of this phenomenon. Such 1 E iR 1 e α α p σ = 1 − 2x + 2 an investigation is beyond the scope of this work. 3 (1 − ν ) FD 5 λ sin (λ ) α n n n=1 In summary, we believe there are two types of side reactions in Li-ion cells. The ﬁrst type of reaction results in the coverage of freshly sin (λ x ) sin (λ x ) − (λ x ) cos (λ x ) n n n n exposed anode surfaces with a SEI and occurs the ﬁrst time the anode × − [3] 3 3 λ x λ x n reaches low voltages and also occurs where there are cracks in the SEI that expose fresh surface. Thus, electrodes with more cracks have where E is the Young’s modulus and ν is the Poisson’s ratio of α α more of this type of side reaction. The second type occurs as a result the electrode material, is the partial molar volume of the solute, of dissolution of the SEI that occurs at low state of charges (SOCs) R is the electrode particle radius, D is the diffusion coefﬁcient of but is reformed at high SOCs. The dissolved SEI diffuses across the the solute inside electrode particle, F is Faraday’s constant, x = cell where it is oxidized resulting in no loss of capacity. Thus, a cell r/R , τ = tD/R , r is any radial location from the center, t is p p with a large overall amount of anode material will experience a lower time during lithiation or delithiation, positive current densities i over coulombic efﬁciency because it has much more SEI that can then the particle surface denote charging, and λ (n = 1, 2, 3,...) are the dissolve and subsequently reformed, where the dissolved components positive roots of tan(λ ) = λ . For a thin ﬁlm electrode with a speciﬁc n n migrate to the cathode where it is reduced but does not result in surface area a , thickness L , and geometric area A and the cell current greater capacity fade. A cell with less anode material will experience I , the current density over a particle surface i can be calculated as cell higher current density at the particle level that results in diffusion induced cracking of either the SEI or particles which requires solvent cell i = [4] consumption and loss of lithium inventory in the cell and relative electrode marching of the anode versus the cathode. This work complements the work of Deshpande et al. where it was also hypothesized that here A is the electrode surface exposed to the electrolyte. electrode VC slows the rate of the second type of reaction but has no effect on At any given characteristic time τ, the tangential tensile stress σ the ﬁrst in a graphite/NMC cell. on the particle surface is proportional to the current density over the Journal of The Electrochemical Society, 162 (10) A2026-A2035 (2015) A2035 Conclusions r, θ, ϕ spherical coordinates partial molar volume of the solute in phase ‘i’(m /mol ) Graphite/NMC full cells with two different anode loading lev- R radius of the spherical electrode particle (m) els were tested to establish a better understanding of coulombic ef- ν Poisson’s ratio of phase ‘i’ ﬁciency and cell capacity fade rate. The cell with a higher anode E Young’s modulus of phase ‘i’(N /m ) st i loading demonstrated more 1 cycle charging capacity. A cross-talk σ tangential stress (N /m ) between the anode and cathode is postulated to be the reason behind α 3 C (r, t ) solute concentration at radius r at time t (mol/m ) this additional capacity. The cell with the higher anode loading also C initial concentration of solute in active core phase experienced a greater loss of cycleable capacity upon completion of (mol/m ) the formation process. The capacity loss during the formation pro- C swing in state of litigation of the ‘i’ phase avg cess is directly related to the total surface area of anode exposed to D diffusion coefﬁcient of the solute in α phase (m /sec) the electrolyte. The cell with the higher anode loading demonstrated i current density on the electrode particle (A/m ) poorer coulombic efﬁciency during long-term cycling. By following I current applied to the cell (A) cell the charge and discharge endpoints, it is shown that the coulombic A electrode surface area (m ) electrode efﬁciency of a cell is a measure of the marching rate of the discharge F Faraday constant (C) endpoint, which for both types of cells was the anode. Any increase Q cell capacity (C) in the amount of reduction reactions causes the anode to meet the dis- x dimensionless radius charge limits earlier and thus increase its marching rate. The higher τ dimensionless time marching rate of anode for a cell with higher anode loading (R = 4.54) dQ rate of loss of discharge capacity is assigned to the greater surface area of the anode and the hypothesis dN that the SEI is constantly dissolving and reforming based on SOC. Consequently, by changing the anode loading, the marching rate of the charge end point, for which the cells were cathode limited, also References changed. Increased marching of the cathode due to increased loading 1. D. Linden and T. B. Reddy, Handbook of batteries, 3rd ed. (McGraw-Hill, New York, of anode strongly suggests that the amount of electrolyte oxidation on 2002). the cathode surface is dependent on the amount of anode present in the 2. M. Armand and J. M. Tarascon, Nature 451(7179), 652 (2008). cell. Through some simple calculations, it was shown that the salt was 3. B. Dunn, H. Kamath, and J. M. Tarascon, Science 334(6058), 928 (2011). not consumed at the full rate of the side reactions on both electrodes 4. R. Deshpande, M. Verbrugge, Y. T. Cheng, J. Wang, and P. Liu, Journal of the Electrochemical Society 159(10), A1730 (2012). or the cells would consumed all of the salt in the cells much sooner 5. D. Aurbach, J Power Sources 89(2), 206 (2000). than when they actually reached the end of test. Thus, we re-afﬁrm 6. E. Markervich, G. Salitra, M. D. Levi, and D. Aurbach, Journal of Power Sources the hypotheses that a shuttle reaction between the anode and cathode 146(1), 146 (2005). occurs without the consumption of electrolytic salt. We believe that 7. J. Vetter, P. Nova ´k,M. R.Wagner, C. Veit,K. C.Moller, ¨ J. O. Besenhard, M. Winter, M. Wohlfahrt-Mehrens, C. Vogler, and A. Hammouche, Journal of Power Sources the electrode reaction products that make up the SEI on the anode, 147(1), 269 (2005). SLi , dissolve in the electrolyte, migrate to the cathode and undergo a 8. M. Winter, J. O. Besenhard, M. E. Spahr, and P. Novak, Adv. Mater. 10(10), 725 reaction that recovers the salt. Such a shuttle, reversible or not, does (1998). not lead to capacity loss in the short term. 9. M. Wachtler, J. O. Besenhard, and M. Winter, Journal of Power Sources 94(2), 189 (2001). Consistent with one of the main conclusions in Reference 12, 10. J. Yan, B. J. Xia, Y. C. Su, X. Z. Zhou, J. Zhang, and X. G. Zhang, Electrochimica we establish that the coulombic efﬁciency is not a good indicator of Acta 53(24), 7069 (2008). battery capacity failure. Additionally, we demonstrate that a cell with 11. J. C. Burns, N. N. Sinha, D. J. Coyle, G. Jain, C. M. VanElzen, W. M. Lamanna, higher coulombic efﬁciency can possibly have a higher capacity fade A. Xiao, E. Scott, J. P. Gardner, and J. R. Dahn, Journal of the Electrochemical Society 159(2), A85 (2012). rate. 12. R. D. Deshpande, P. Ridgway, Y. B. Fu, W. Zhang, J. S. Cai, and V. Battaglia, Journal of The Electrochemical Society 162(3), A330 (2015). 13. J. Christensen and J. Newman, J. Solid State Electr 10(5), 293 (2006). Acknowledgments 14. R. D. Deshpande, University of Kentucky, Lexington, Kentucky (USA), 2011. 15. R. Deshpande, Y. T. Cheng, and M. W. Verbrugge, Journal of Power Sources 195(15), The authors would like to express gratitude to the entire battery 5081 (2010). research members of Dr. Battaglia and Dr. Gao Liu in the Electrochem- 16. R. Deshpande, Y. Qi, and Y. T. Cheng, Journal of the Electrochemical Society 157(8), ical Technologies Group of Lawrence Berkeley National Lab. Special A967 (2010). thanks to Dr. Yanbao Fu, Xiangyun Song, and visiting student Min 17. V. A. Sethuraman, N. Van Winkle, D. P. Abraham, A. F. Bower, and P. R. Guduru, Journal of Power Sources 206, 334 (2012). Ling, who were very helpful with the running of the experiments. The 18. P. Liu, J. Wang, J. Hicks-Garner, E. Sherman, S. Soukiazian, M. Verbrugge, H. Tataria, authors acknowledge the BATT Program and the U.S. Department of J. Musser, and P. Finamore, J Electrochem Soc 157(4), A499 (2010). Energy, as well as National Science Foundation Award No. 1355438 19. J. Li, A. K. Dozier, Y. Li, F. Yang, and Y.-T. Cheng, Journal of The Electrochemical (Powering the Kentucky Bioeconomy for a Sustainable Future) for Society 158(6), A689 (2011). 20. H. Zheng, G. Liu, S. Carwford, and V. S. Battaglia, (http://bestar.lbl.gov/ partially funding this project. vbattaglia/cell-analysis-tools/). 21. M. Dubarry and B. Y. Liaw, Journal of Power Sources 194(1), 541 (2009). 22. V. Pop, H. Bergveld, J. O. het Veld, P. Regtien, D. Danilov, and P. Notten, Journal of The Electrochemical Society 153(11), A2013 (2006). List of Symbols 23. R. Spotnitz, Journal of Power Sources 113(1), 72 (2003). R ratio of initial anode capacity to cathode capacity 24. R. Yazami and Y. F. Reynier, Electrochimica Acta 47(8), 1217 (2002). d c 25. M. Lu, H. Cheng, and Y. Yang, Electrochimica Acta 53(9), 3539 (2008). rate of marching of charge endpoint (A/g) dt 26. S. E. Sloop, J. B. Kerr, and K. Kinoshita, Journal of Power Sources 119–121(0), 330 d d (2003). rate of marching of discharge endpoint (A/g) dt 27. A. Gennaro, A. A. Isse, J.-M. Saveant, ´ M.-G. Severin, and E. Vianello, Journal of CE coulombic efﬁciency the American Chemical Society 118(30), 7190 (1996). th Qd discharge capacity of n cycle (Ah) 28. M. Safari and C. Delacourt, Journal of The Electrochemical Society 158(10), A1123 th Qc charge capacity of n cycle (Ah) n (2011). 29. Y.-T. Cheng and M. W. Verbrugge, Journal of Power Sources 190(2), 453 (2009). RLL I rate of loss of lithium inventory (A/g)

Journal

Journal of the Electrochemical Society – IOP Publishing

Published: Jul 28, 2015

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Electrode Side Reactions, Capacity Loss and Mechanical Degradation in Lithium-Ion Batteries

Electrode Side Reactions, Capacity Loss and Mechanical Degradation in Lithium-Ion Batteries

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Electrode Side Reactions, Capacity Loss and Mechanical Degradation in Lithium-Ion Batteries

Electrode Side Reactions, Capacity Loss and Mechanical Degradation in Lithium-Ion Batteries

References (26)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies