Investigating the Low-Temperature Impedance Increase of Lithium-Ion Cells

Journal of The Electrochemical Society, 155 共1兲 A41-A47 共2008兲 A41 0013-4651/2007/155共1兲/A41/7/$23.00 © The Electrochemical Society Investigating the Low-Temperature Impedance Increase of Lithium-Ion Cells D. P. Abraham,*,z J. R. Heaton, S.-H. Kang, D. W. Dees,* and A. N. Jansen* Chemical Engineering Division, Argonne National Laboratory, Argonne, Illinois 60439, USA Low-temperature performance loss is a significant barrier to commercialization of lithium-ion cells in hybrid electric vehicles. Increased impedance, especially at temperatures below 0°C, reduces the cell pulse power performance required for cold engine starts, quick acceleration, or regenerative braking. Here we detail electrochemical impedance spectroscopy data on binder- and carbon-free layered-oxide and spinel-oxide electrodes, obtained over the +30 to −30°C temperature range, in coin cells containing a lithium-preloaded Li4/3Ti5/3O4 composite 共LTOc兲 counter electrode and a LiPF6-bearing ethylene carbonate/ethyl methyl car- bonate electrolyte. For all electrodes studied, the impedance increased with decreasing cell temperature; the increases observed in the midfrequency arc dwarfed the increases in ohmic resistance and diffusional impedance. Our data suggest that the movement of lithium ions across the electrochemical interface on the active material may have been increasingly hindered at lower tempera- tures, especially below 0°C. Low-temperature performance may be improved by modifying the electrolyte-active material inter- face 共for example, through electrolyte composition changes兲. Increasing surface area of active particles 共for example, through nanoparticle use兲 can lower the initial electrode impedance and lead to lower cell impedances at −30°C. © 2007 The Electrochemical Society. 关DOI: 10.1149/1.2801366兴 All rights reserved. Manuscript submitted December 7, 2006; revised manuscript received July 27, 2007. Available electronically November 7, 2007. Increasing fuel costs have spurred the growth of hybrid electric deintercalation process; and 共iv兲 using nanostructured electrodes to vehicles 共HEVs兲 that carry a battery pack to supplement/ reduce the distance that the Li ion must diffuse into and out of the complement the performance of the internal combustion engine active particles. 共ICE兲. These HEVs provide better gas mileage and produce smaller The pulse power capability of a lithium-ion cell is limited by its quantities of noxious emissions than do their ICE-only counterparts. impedance; an impedance increase diminishes the cell’s ability to Because of their higher energy storage and power densities, lithium- deliver and accept current pulses at relatively high rates. Measure- ion battery packs are expected to replace the nickel/metal-hydride ments by Jansen et al. at Argonne have shown that cell impedance battery packs used in currently available HEVs. Lithium-ion cells increases significantly at temperatures below 0°C.21 A typical ac impedance dataset collected on a cell containing a can handle the big bursts of power that occur during quick accelera- Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2-based positive electrode, a MCMB-10 tion 共rapid discharge兲 and regenerative braking 共fast charging兲. graphite-based negative electrode, and a LiPF6-based carbonate However, the energy and power densities of these cells are signifi- electrolyte is shown in Fig. 1. The 30°C data 共inset, Fig. 1兲 shows cantly lower at temperatures below 0°C.1,2 According to the 共a兲 a high-frequency intercept on the real axis associated with the FreedomCar Energy Storage goals3 set by the U.S. Department of bulk cell resistance that includes ohmic contributions from the elec- Energy 共DOE兲, lithium-ion cells in power-assist HEVs should de- trolyte, separator, and electrode; 共b兲 a midfrequency arc related to liver 5 kW of cold cranking power during a 2 s pulse at −30°C; this electrode–electrolyte interfacial processes; and 共c兲 a low-frequency test protocol simulates the start of a cold engine. Furthermore, the sloping line 共Warburg tail兲 that is related mainly to diffusion pro- battery pack should survive at −46°C 共i.e., changes resulting from cesses in the electrolyte and through the active material particles. exposure to low temperatures should not damage the innards of the The main change observed upon lowering the temperature from +30 cell or the battery pack兲. Research to meet these stringent low- to −30°C is an increase in the width of the midfrequency arc, indi- temperature performance goals is being conducted on high-power cating an impedance increase at the electrode–electrolyte interface. lithium-ion cells at Argonne National Laboratory 共Argonne兲 as Increases in the ohmic impedance, seen in the high-frequency inter- part of the DOE’s Advanced Technology Development 共ATD兲 program.4-9 Researchers studying lithium-ion cells10-20 have attributed the performance loss at low temperatures to various factors, including the following: 共i兲 increased viscosity and reduced lithium-ion mo- bility in the electrolyte, 共ii兲 diminished transport properties of elec- trode passivation films, 共iii兲 slower lithium-ion diffusion in the ac- tive material particles, and 共iv兲 substantially increased charge- transfer resistances at the electrode–electrolyte interface. The increase in cell impedance at low temperatures is often attributed to impedance increases at the negative electrode, especially for carbon- based anodes, and sometimes to impedance increases at the positive electrode. Strategies to improve low-temperature performance in- clude the following: 共i兲 use of binary and higher mixtures of sol- vents to depress the freezing point of carbonate electrolytes; 共ii兲 addition of cosolvents with low freezing points and low viscosity at low temperatures, such as organic esters, to the carbonate electro- lytes; 共iii兲 modifying particle morphology to increase the electro- chemically active area for the lithium-ion intercalation– Figure 1. EIS data 共100 KHz to 5 mHz兲 from a Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 共+兲/graphite共−兲 cell at temperatures from +30 to −30°C. The +30°C spec- * Electrochemical Society Active Member. trum is visible only in the inset, which is an expanded view showing the z E-mail: [email protected] lower impedance end of the data. A42 Journal of The Electrochemical Society, 155 共1兲 A41-A47 共2008兲 cept 共inset, Fig. 1兲, are relatively small. The relative contributions of the LiMO2 共M = Ni, Co, Mn兲-based positive and the graphite- based negative electrodes to impedance increase have been deter- mined by incorporating a Li4.4Sn reference electrode into these cells.21 In general, the data show that the impedance of both nega- tive and positive electrodes increase by similar proportions with decreasing cell temperature. All the electrodes that we have studied typically contain an or- Figure 2. 共Color online兲 Schematic of coin cells studied: 共a兲 lithium- ganic binder and carbon additives to improve electronic conductiv- preloaded Li4/3Ti5/3O4 composite 共LTOc兲 symmetric electrode cell and 共b兲 ity; the contribution of these nonactive components to electrode BCF electrode 共+兲//LTOc共−兲 cell. impedance rise at low temperatures is not known. Some poly共vi- nylidene difluoride兲 共PVdF兲 binders are known to have a glass tran- sition temperature of around −30°C. Because the binder is in inti- mate contact with the active material particles, an increase in binder stabilizes after a time period that could range from several days to viscosity could significantly hinder lithium charge-transfer charac- many weeks; 共ii兲 Li metal electrodes display unstable/irreproducible teristics at the electrode–electrolyte interface. To eliminate contribu- impedance data during temperature cycling 共i.e., the Li electrode tions from the binder and carbon additives at low temperatures, we impedance depends on its temperature-cycle history, which made it examined the impedance behavior of binder- and carbon-free 共BCF兲 unsuitable for our measurements兲. In contrast to Li metal, the LTOc layered-oxide 关Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2兴 and spinel-oxide electrodes provided stable, reproducible, and reliable voltage, capac- 关LiMn2O4 and Li4/3Ti5/3O4兴 electrodes over the +30 to −30°C tem- ity, and impedance data. perature range. We studied the Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 and The LTOc electrode was prepared with ⬃1 ␮m size Li4/3Ti5/3O4 LiMn2O4 electrodes because the oxides are of commercial interest particles, carbon-conducting agents, and PVdF binder on an 18 ␮m for high-power applications. The Li4/3Ti5/3O4 electrodes were of thick Cu foil.7 When cycled vs Li, the electrode showed the ex- particular interest because the oxide’s stability in typical pected flat voltage profile at 1.55 V vs Li; electrode capacity was LiPF6-based carbonate electrolytes results in a relatively “clean” ⬃1.7 mAh/cm2 over this 1.55 V plateau. The lithium preloading interface; the oxide does not form a solid electrolyte interphase was achieved during the first charge cycle of a cell containing 共SEI兲 layer akin to that on carbon electrodes because its 1.55 V vs 32 cm2 area Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 positive and Li4/3Ti5/3O4 Li plateau is higher than the voltages for carbonate solvent reduc- negative composite electrodes and a 1.2 M LiPF6 in an ethylene tion. Impedance measurements were conducted in cells containing a carbonate/ethyl methyl carbonate 共EC:EMC兲 共3:7 by weight兲 elec- lithium-preloaded Li4/3Ti5/3O4 composite 共LTOc兲 counter electrode. trolyte 共henceforth referred to as “Gen2” electrolyte兲. The cell The possible causes of impedance rise and implications of the data was held at 2.32 V 共3.87 V vs Li兲 for 24 h, during which to improve low-temperature electrode performance are discussed be- ⬃0.8 mAh/cm2 of lithium was inserted into the LTOc electrode low. 共i.e., fewer than half of the available sites in the LTOc electrode were filled with lithium ions兲. Experimental Electrode preparation.— The BCF electrodes were prepared Cell preparation and electrochemical testing.— Electrochemical by nonaqueous sol-gel spin coating on a 50 ␮m thick Pt foil tests were conducted on 2032-type coin cells 共20 mm diam, 3.2 mm substrate, followed by heat-treatment in air.22 The thick兲 assembled with 1.27 cm2 area electrodes in an Ar-atmosphere Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 precursor sol was prepared by dissolv- glove box. The cells contained the following electrode combina- ing Li-acetate, Ni-acetate, Co-nitrate, Mn-acetate, and polyvinylpyr- tions: rolidone 共PVP兲-55000 in a mixture solution of ethanol and acetic acid. The LiMn2O4 precursor sol was prepared by dissolving Li- 共1兲 LTOc 共+兲 and 共−兲, symmetric arrangement acetate, Mn-chloride, and PVP-55000 in ethanol. The Li4/3Ti5/3O4 共2兲 BCF Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 共+兲 and LTOc 共−兲 precursor sol was prepared by dissolving Li-acetate, Ti- 共3兲 BCF LiMn2O4 共+兲 and LTOc 共−兲 isopropoxide, and PVP-55000 in an ethanol and acetic acid solution 共4兲 BCF Li4/3Ti5/3O4 共+兲 and LTOc 共−兲 mixture. The prepared sols were spin-coated on Pt foils at 3000 rpm for 30 s to obtain a gel film. The gel films were then heat-treated at In addition to the electrodes, each cell also contained Celgard 2325 600°C for ⬃1 h in air to burn off the organic components. The separator 共25 ␮m thick兲 and the Gen2 electrolyte. A schematic of spin-coating and heat-treatment processes were repeated several the coin cell electrode contents is shown in Fig. 2. Note that these times until a film of desired thickness was obtained. Each electrode are two-electrode coin cells 共i.e., a reference electrode was not used was subjected to a final heat-treatment in air to obtain the well- in this study兲. crystallized oxide product: the Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 elec- All cells were studied in a Tenney Environmental TJR tempera- trode was heated at 850°C for 2 h, the LiMn2O4 electrode at 750°C ture chamber; the temperature accuracy within the chamber was for 3 h, and the Li4/3Ti5/3O4 electrode at 700°C for 30 min. All determined to be within ±1°C of the set point. Charge–discharge electrodes were studied by X-ray diffraction to ensure oxide phase characteristics of the BCF electrode cells were determined at 30°C, purity. Electrode examination by scanning electron microscopy using channels on a Maccor cycler calibrated for small currents. The 共SEM兲 showed that the Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 and LiMn2O4 electrochemical impedance spectroscopy 共EIS兲 studies were per- particle size averaged ⬃0.3 ␮m, while the Li4/3Ti5/3O4 particles formed over a frequency range of 100 kHz to 5 mHz with an EG&G were smaller, averaging ⬃0.050 ␮m in size. All oxide coatings were 273A potentiostat and a Solartron SI1260 frequency response ana- typically less than 1 ␮m thick. The oxide content of these electrodes lyzer controlled by Zplot measurement software. A 10 mV perturba- was small 共typically ⬍1 mg兲 and often difficult to measure accu- tion was used in all EIS studies. Impedance data were obtained at rately. Therefore, absolute electrode capacity values 共in ␮Ah兲 are 30, 20, 10, 0, −10, −20, and −30°C and then again at 30°C to check reported in this article instead of normalized capacities values 共in for cell degradation; measurements were conducted only after the mAh/g兲. cells were allowed to equilibrate at the testing temperatures, typi- A lithium-preloaded LTOc electrode was used as a surrogate for cally for a period of 3 h. In general, the 30°C capacity and imped- lithium metal electrodes. Our experiments with lithium metal ance data obtained before and after the low-temperature tests counter electrodes have shown that 共i兲 the impedance of fresh Li showed only small differences, indicating that cell degradation was metal electrodes typically increases with cell age; the impedance small 共if any兲 during the test duration.  Journal of The Electrochemical Society, 155 共1兲 A41-A47 共2008兲 A43 Figure 3. Representative charge–discharge plot from the Figure 5. Representative charge–discharge plot from the Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2共+兲/LTOc共−兲 coin cell acquired with a 4 ␮A Li4/3Ti5/3O4共+兲/LTOc共−兲 coin cell acquired with a 6 ␮A current at 30°C. current at 30°C. The EIS data were obtained at location x on the plot. The EIS data were obtained at location x on the plot. Results second-order phase transformation leading to random lithium inser- tion into the spinel structure;24,25 the processes occur in reverse or- Capacity measurements.— Representative charge–discharge cy- der upon charge. Others attribute the plateaus to a cubic-to-cubic cling plots at 30°C on BCF electrode共+兲/LTOc共−兲 cells are shown phase transition; these distinct cubic phases have slightly differing in Fig. 3-5; the data are from the third cycle for each cell. The true lattice parameters.26,27 Lithium extraction from LiMn2O4 produces measured voltage for each cell is shown on the left y axis; the right an oxidation of Mn+3 ions to Mn+4; the reverse reaction occurs y axis shows the equivalent voltage vs Li, based on the 1.55 V during lithium intercalation. plateau observed for the LTOc vs Li cells. The Li4/3Ti5/3O4共+兲/LTOc共−兲 cells showed a capacity of The Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2共+兲/LTOc共−兲 cells showed ⬃56 ␮Ah when cycled between +0.3 and −0.3 V with a 6 ␮A cur- a capacity of ⬃33 ␮Ah when cycled between 1.5 and 2.6 V 共3.05– rent. The plateau is centered around 0 V because, when lithiated, 4.15 V vs Li兲 with a 4 ␮A current. The voltage varied in a smooth, both electrodes have a voltage of ⬃1.55 V vs Li corresponding to continuous manner during the charge and discharge cycles, which is the Li4/3Ti5/3O4 + Li+ + e− ↔ Li7/3Ti5/3O4 reaction.28,29 Lithium in- typical for Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2. Lithium extraction causes tercalation into the Li4/3Ti5/3O4 is associated with a reduction of a change in the Ni oxidation state from +2 to +4;23 the oxidation Ti+4 ions to Ti+3; the reverse reaction occurs during lithium extrac- states of Co and Mn are apparently unchanged in the voltage range tion. studied. Lithium intercalation into the oxide produces a reduction of In separate experiments we confirmed that, for the same potential Ni+4 ions to Ni+2. window, the BCF electrode capacities 共measured at 30°C兲 are simi- The LiMn2O4共+兲/LTOc共−兲 cells showed a capacity of lar for cells with lithium metal or LTOc counter electrodes; the ⬃102 ␮Ah when cycled between 1.5 and 2.6 V 共3.05–4.15 V vs Li兲 LTOc electrode acts as a virtual lithium-ion source 共and sink兲 for the with a 6 ␮A current. The data, typical for LiMn2O4, show two dis- BCF electrodes. This occurs because the BCF electrode capacities tinct, slightly sloping plateaus between 3.9 and 4.15 V vs Li on both are less than ⬃0.08 mAh/cm2 共i.e., an order of magnitude smaller charge and discharge. Some researchers attribute the discharge pla- than the capacity of the LTOc electrode兲. teaus to an ordered intercalation of lithium ions, followed by a Impedance measurements.— Bode plots–full cells.— Bode plots at various temperatures for the BCF electrode共+兲/LTOc共−兲 cells are shown in Fig. 6-8. For each cell, the absolute impedance, Zabs, and the phase shift, ␪, of the impedance are shown as a function of frequency, f. In the log Zabs vs log f plots, the highest-frequency data are an indicator of the bulk cell resistance. The sloping portions in the Zabs plots indicate capacitive behavior in the electrochemical system; multiple sloping portions in the plots are associated with multiple electrochemical phenomena on the various active material surfaces of the electrode. The peaks in the phase shift plots are associated with the sloping portions in the Zabs plots. The phase shift maximum, ␪max, corresponds approximately to the midpoint of a sloping portion. The low-frequency minimum, ␪min, provides an ap- proximate demarcation between interfacial and diffusion processes in the electrode. A study of Fig. 6-8 shows the similarities and differences be- tween the three BCF electrode cell data over the +30 to −30°C temperature range; note that these data include the contributions of both electrodes. With decreasing temperature increases in the bulk cell resistance are observed in the Zabs plots; the magnitude of this Figure 4. Representative charge–discharge plot from the resistance increase is roughly the same for all cells. At frequencies LiMn2O4共+兲/LTOc共−兲 coin cell acquired with a 6 ␮A current at 30°C. The lower than 10 kHz, a purely horizontal plateau is not observed, ei- EIS data were obtained at location x on the plot. ther in the Zabs or in the phase shift plots. The electrode–electrolyte A44 Journal of The Electrochemical Society, 155 共1兲 A41-A47 共2008兲 Figure 6. 共Color online兲 Bode plots for the Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 Figure 7. 共Color online兲 Bode plots for the LiMn2O4共+兲/LTOc共−兲 cell. The 共+兲/LTOc共−兲 cell. The phase shift maximum 共␪max兲 and minimum 共␪min兲 for phase shift maximum 共␪max兲 and minimum 共␪min兲 for the −30°C data are the 30°C data are shown. shown. interfacial processes are dominant in the frequency range from cells, the BCF electrode’s impedance at any frequency can be ob- ⬃1 kHz to the frequency corresponding to ␪min for the various cells. tained by subtracting the impedance of the LTOc electrode at that Significant increases in Zabs are observed in this frequency regime, frequency, which is known from the symmetric cell data obtained at especially at the lower test temperatures; the magnitude of this im- the relevant temperature. In these analyses, we assume the separator pedance increase is different for the different electrodes. contributions to the impedance data are negligible, which our prior The Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2共+兲/LTOc共−兲 cell phase shift data21 indicate is a valid assumption. data show a clear ␪max at 30°C; the plots become increasingly wider Impedance data acquired at various temperatures for a single with decreasing temperature, with ␪max shifting to lower frequen- LTOc electrode are shown in Fig. 9; the cell voltage is 0 V because cies. At −30°C the phase shift plots indicate the presence of mul- both electrodes are at the same potential 共1.55 V vs Li兲. The ohmic tiple peaks, which suggests multiple electrochemical phenomena at portion 共R⍀兲 of the impedance increases from ⬃1.3 ⍀ at 30°C to the electrode–electrolyte interface. The widening of the phase shift ⬃2.5 ⍀ at 0°C to ⬃8.5 ⍀ at −30°C. A conductivity decrease in the plots and the ␪max shift to lower frequencies with decreasing tem- bulk electrolyte is mainly responsible for the higher R⍀ at lower perature are also observed in the LiMn2O4共+兲/LTOc共−兲 and the temperatures. In the same temperature range, the midfrequency arc Li4/3Ti5/3O4共+兲/LTOc共−兲 data. Although the presence of multiple width along the real axis 共Rw兲 increases from ⬃0.6 ⍀ at 30°C to peaks is not as distinct in these data, the wider peaks are suggestive ⬃2.8 ⍀ at 0°C to ⬃50 ⍀ at −30°C 共i.e., the magnitude of the Rw of multiple interfacial phenomena that become increasingly impor- increase is significantly greater than that for R⍀兲. The length of the tant at the lower temperatures. Note that for all cells ␪min moves to Warburg impedance tail is also seen to increase, which indicates an lower frequencies, indicating that diffusion phenomena are not a impedance increase in electrode diffusion processes at lower cell major contributor to the Zabs increase observed with decreasing tem- temperatures. In general, at any given temperature, the LTOc elec- perature. trode impedance was much smaller than the BCF electrode imped- Nyquist plots–individual electrodes.— The EIS data obtained from ance. a two-electrode cell are the sum of the individual impedances of the The BCF electrode’s impedance, calculated by subtract- positive and negative electrodes 共i.e., the impedance data from our ing out the LTOc electrode’s contribution in the BCF electrode coin cells contains contributions from both electrodes兲.30 To obtain 共+兲/LTOc共−兲 cell data, is shown in Fig. 10-12. For each BCF the impedance data for individual electrodes, we need to know the electrode, the impedance data reported are at the location x marked contribution of the counter electrode. For the LTOc symmetric cell, on its cycling curves 共Fig. 3-5兲; the Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 the impedance of each electrode at any frequency is half that of the data were obtained at ⬃4 V vs Li, the LiMn2O4 data at the total impedance at that frequency 共assuming that both electrodes midrange of the upper voltage plateau 共⬃4.12 V vs Li兲, and the behave in the same manner兲. In the BCF electrode共+兲/LTOc共−兲 Li4/3Ti5/3O4 data at the midrange of its 1.55 V vs Li voltage plateau.  Journal of The Electrochemical Society, 155 共1兲 A41-A47 共2008兲 A45 Figure 10. EIS data 共100 kHz to 5 mHz兲 for the BCF Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 electrode at various temperatures. The −20°C 共to 1 mHz兲 and −30°C 共to 160 ␮Hz兲 impedance values are significantly larger and hence are shown in the inset. Rw dwarf the changes in R⍀ 共i.e., the changes at the electrode– electrolyte interface are the main contributors to cell impedance in- creases observed at lower temperatures兲. Impedance increases ob- served in the Warburg tail for the LiMn2O4 and the Li4/3Ti5/3O4 electrodes are again relatively small compared to the Rw increase. The speed of interfacial electrochemical processes can be gauged from the frequency response obtained from the EIS data. In systems wherein the electrochemical process can be modeled by a simple RC circuit, the midfrequency arc maximum, f max, is inversely propor- tional to the time constant of the electrochemical process. However, Figure 8. 共Color online兲 Bode plots for the Li4/3Ti5/3O4共+兲/LTOc共−兲 cell. The phase shift maximum 共␪max兲 and minimum 共␪min兲 for the −30°C data are the midfrequency arc impedance data from our electrodes do not shown. result from a single electrochemical process. Rather, the arc contains contributions from several electrochemical processes, each with its own time constant. The low-frequency minimum, f min, where the midfrequency arc intersects the Warburg tail, is a better demarcation To access the entire width of the midfrequency arc, data for between the speed of the interfacial and diffusion processes. Table I Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 were measured to 1 mHz at −20°C shows f max and f min data as a function of test temperature for the and to 160 ␮Hz at −30°C. For all BCF electrodes, the increases in LTOc and BCF electrodes. It is immediately evident that both f max and f min decrease by more than two orders of magnitude when the Figure 9. EIS plots 共100 kHz to 5 mHz兲 for a single electrode calculated from the symmetric electrode LTOc cell 共0 V兲 data acquired at various tem- Figure 11. EIS data 共100 kHz to 5 mHz兲 for the BCF LiMn2O4 electrode at peratures. f max, f min, and Rw are indicated for the −30°C data. Inset is an various temperatures. The 30, 20, and 10°C impedance data are shown in the expanded view showing the lower impedance end of the data. inset. A46 Journal of The Electrochemical Society, 155 共1兲 A41-A47 共2008兲 Figure 13. Arrhenius plot showing Rw 共real portion of the resistance asso- ciated with the midfrequency arc兲 as a function of test temperature for the Figure 12. EIS data 共100 kHz to 5 mHz兲 for the BCF Li4/3Ti5/3O4 electrode various electrodes examined. The activation energy 共Ea兲 data are in kcal/mol. at various temperatures. The −10, −20, and −30°C impedance data are shown in the inset. and to differences in particle orientation on the Pt substrate. The cell test temperature is lowered from +30 to −30°C 共i.e., the inter- three-dimensional nature of lithium diffusion from/to the Ti and Mn facial electrochemical processes slow down significantly at the spinels produces the lower Rw values. The smaller particle size 共and lower temperatures兲. hence, higher active surface area兲 in the Li4/3Ti5/3O4 electrodes re- sults in a lower Rw than for the LiMn2O4 electrode. The higher Rw Discussion values for Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 may result from the planar The plot in Fig. 13 relates Rw to test temperature T, according to nature of Li diffusion in the oxide. Only particles having Li layers the Arrhenius equation oriented normal to the substrate have the lowest impedance; random particle orientations on the Pt substrate produce higher Rw values. Rw = A exp − 冉 冊 Ea RT For the BCF electrodes, the Arrhenius equation relating Rw to temperature holds true for the entire range from +30 to −30°C. Furthermore, the BCF electrode plots are almost parallel to each where A is a constant, Ea is the activation energy, and R is the gas other; the similar activation energies indicate that regardless of the constant. The following points in the plot are noteworthy and are active material, the same activation process is dominant over the discussed below: 共i兲 at 30°C, Rw is smallest for the LTOc electrode temperature range studied. This observation is especially significant and highest for the Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 electrode; 共ii兲 the for Li4/3Ti5/3O4 because surface films are not expected to be signifi- log Rw vs 1/T plots are approximately linear for all BCF electrodes, cant on this electrode. That is, the increase in interfacial impedance and furthermore, the Ea values for all three electrodes are similar, with decreasing temperature cannot be attributed solely to the pres- considering the range of experimental uncertainty; and 共iii兲 a dis- ence of a SEI layer. tinct kink is observed in the LTOc plot at 0°C 共the Ea below 0°C is The slope change at 0°C seen in the LTOc plot has also been comparable with that for the BCF electrodes, and the Ea above 0°C observed for porous electrodes prepared with other active materials is smaller兲. 共oxides and graphites兲.21 In all cases, the activation energies below For all our electrodes, the midfrequency arc arises from interfa- 0°C are greater than those above 0°C. The similarities in Ea below cial phenomena that proceed simultaneously on the various active 0°C between the LTOc and BCF electrodes indicate that the binder particles. Besides electron-transfer kinetics, the transport of ionic and carbon present in the composite, porous electrodes are not major species by diffusion or migration across the electrode–electrolyte contributors 共i.e., the dominant process producing the impedance interface contributes to Rw. The value of Rw is lowest for the porous increase at temperatures below 0°C is at the oxide–electrolyte inter- LTOc electrode because of its larger electrochemically active sur- face兲. Above 0°C, however, the binder and carbon appear to play a face area; the BCF electrodes have higher Rw values because of their role in reducing the degree of impedance change with temperature. smaller surface areas. The differences in Rw for the various BCF Our data suggest that the limiting process is the electrochemical electrodes may be related to particle morphology/size differences reaction step at the interface 共i.e., moving the lithium ions across the Table I. EIS data on midfrequency arc frequency parameters, fmax and fmin (in Hz), obtained as a function of test temperature for the different materials. LTOc Li1.05共Ni1/3Co1/3Mn1/3兲0.95O2 LiMn2O4 Li4/3Ti5/3O4 T 共°C兲 f max f min f max f min f max f min f max f min 30 630 10 6.3 0.63 250 6.3 4000 250 20 250 6.3 2.5 0.25 100 4 630 63 10 158 4 1.6 0.063 40 1.6 158 40 0 40 1.6 0.4 0.025 16 1 63 10 −10 6.3 0.4 0.04 0.0063 6.3 0.25 25 2.5 −20 2.5 0.16 0.025 0.0025 2.5 0.1 16 0.63 −30 0.63 0.04 0.016 0.0006 0.63 0.025 6.3 0.16  Journal of The Electrochemical Society, 155 共1兲 A41-A47 共2008兲 A47 active material/electrolyte interface兲. When we discussed the Acknowledgments lithium-ion intercalation reaction at an oxide/electrolyte interface in This work was supported by the FreedomCAR and Vehicle Tech- a previous article,31 we suggested that Rw includes contributions nologies Program at the U.S. Department of Energy. The authors from the Li+ desolvation processes, diffusion/migration through sur- thank Gary Henriksen for his support and encouragement. The sub- face films, interfacial electron transfer, and Li+ conduction along the mitted manuscript has been created by Argonne National Labora- oxide surface. The activation energy that relates Rw to temperature tory, a U.S. Department of Energy, Office of Science laboratory 共Fig. 13兲 reflects the activation energy of the slowest or rate- operated by UChicago Argonne, LLC, under contract no. DE-AC02- determining step in the sequence. The similarity in activation energy 06CH11357. The U.S. Government retains for itself, and others act- values suggests that the phenomena responsible for the activation ing on its behalf, a paid-up nonexclusive, irrevocable worldwide energy barrier are the same for the various materials studied. The license in said article to reproduce, prepare derivative works, dis- data presented in this article are insufficient to definitively identify tribute copies to the public, and perform publicly and display pub- the interfacial phenomena responsible for this activation energy bar- licly, by or on behalf of the Government. rier. Argonne National Laboratory assisted in meeting the publication costs of It is evident from our data that changing the type of active ma- this article. terial, binder, or conducting carbon in the electrodes is not likely to prevent the interfacial impedance increase with decreasing tempera- References ture. To improve low-temperature performance, we need to reduce 1. G. Nagasubramanian, J. Appl. Electrochem., 31, 99 共2001兲. 2. J. Fan and S. Tan, J. Electrochem. Soc., 153, A1081 共2006兲. the high activation energy of the limiting process, change the limit- 3. FreedomCAR Battery Test Manual for Power-Assist Hybrid Electric Vehicles, ing process altogether, or increase the electrochemical active area of DOE/ID-11069, INEEL 共2003兲. the active material enough to allow it to operate at −30°C. Increas- 4. D. P. Abraham, J. Liu, C. H. Chen, Y. E. Hyung, M. Stoll, N. Elsen, S. MacLaren, ing the surface areas of active particles by decreasing particle size R. Twesten, R. Haasch, E. Sammann et al., J. Power Sources, 119–121, 511 共2003兲. 共use of nanomaterials,18,19 for example兲 can lead to lower electrode 5. D. Dees, E. Gunen, D. P. Abraham, A. N. Jansen, and J. Prakash, J. Electrochem. impedances at −30°C, as seen for the Li4/3Ti5/3O4 electrode data. Soc., 152, A1409 共2005兲. Optimizing electrolyte composition for low-temperature perfor- 6. I. Bloom, A. N. Jansen, D. P. Abraham, J. Knuth, S. A. Jones, V. S. Battaglia, and G. L. Henriksen, J. Power Sources, 139, 295 共2005兲. mance is another means of altering the limiting process 共or at least 7. D. P. Abraham, E. M. Reynolds, E. Sammann, A. N. Jansen, and D. W. Dees, its activation energy兲 and has been the approach of several Electrochim. Acta, 51, 502 共2005兲. investigators.10-14,32-34 8. A. M. Andersson, D. P. Abraham, R. Haasch, S. MacLaren, J. Liu, and K. Amine, J. Electrochem. Soc., 149, A1358 共2002兲. 9. D. P. Abraham, R. D. Twesten, M. Balasubramanian, J. Kropf, D. Fischer, J. McBreen, I. Petrov, and K. Amine, J. Electrochem. Soc., 150, A1450 共2003兲. Conclusion 10. Y. Ein-Eli, S. R. Thomas, R. Chadha, T. J. Blakley, and V. R. Koch, J. Electrochem. Soc., 144, 823 共1997兲. 11. M. C. Smart, B. V. Ratnakumar, and S. Surampudi, J. Electrochem. Soc., 146, 486 We have studied the low-temperature performance of binder- and 共1999兲. carbon-free layered-oxide and spinel-oxide electrodes in the tem- 12. M. C. Smart, B. V. Ratnakumar, S. Surampudi, Y. Wang, X. Zhang, S. G. Green- perature range of +30 to −30°C in coin cells containing a lithium- baum, A. Hightower, C. C. Ahn, and B. Fultz, J. Electrochem. Soc., 146, 3963 共1999兲. preloaded Li4/3Ti5/3O4 composite counter electrode and a 13. M. C. Smart, B. V. Ratnakumar, and S. Surampudi, J. Electrochem. Soc., 149, LiPF6-bearing EC:EMC electrolyte. Our conclusions from this study A361 共2002兲. are as follows: 14. S. S. Zhang, K. Xu, and T. R. Jow, Electrochim. Acta, 49, 1057 共2004兲. 1. The LTOc counter electrode is an excellent lithium metal sur- 15. L. F. Xiao, Y. L. Cao, X. P. Ai, and H. X. Yang, Electrochim. Acta, 49, 4857 共2004兲. rogate for low-temperature impedance measurements. Impedance 16. M. K. Stowe, P. Liu, and G. L. Baker, Chem. Mater., 17, 6555 共2005兲. data on the LTOc electrode can be obtained from cells containing 17. C.-K. Huang, J. S. Sakamoto, J. Wolfenstine, and S. Surampudi, J. Electrochem. these electrodes, arranged in a symmetric configuration. Soc., 147, 2893 共2000兲. 18. C. R. Sides and C. R. Martin, Adv. Mater. (Weinheim, Ger.), 17, 125 共2005兲. 2. Impedance for both the BCF and LTOc electrodes increases 19. J. Li, Z. Tang, and Z. Zhang, Electrochem. Solid-State Lett., 8, A570 共2005兲. with decreasing cell temperature. The changes observed in the mid- 20. S. H. Choi, J.-W. Son, Y. S. Yoon, and J. Kim, J. Power Sources, 158, 1419 frequency arc of the EIS data dwarf the changes in ohmic resistance 共2006兲. and diffusional impedance. 21. A. N. Jansen, D. W. Dees, D. P. Abraham, K. Amine, and G. L. Henriksen, J. Power Sources, Accepted. 3. The activation energies determined from Arrhenius plots 22. S.-H. Kang and D. P. Abraham, J. Power Sources, Accepted. 共log Rw vs 1/T兲 were similar for all BCF electrodes 共i.e., the activa- 23. W.-S. Yoon, M. Balasubramanian, K. Y. Chung, X. Q. Yang, J. McBreen, C. P. tion energy appears to be relatively independent of the active mate- Grey, and D. A. Fischer, J. Am. Chem. Soc., 127, 17479 共2005兲. rial studied兲. 24. W. Liu, K. Kowal, and G. C. Farrington, J. Electrochem. Soc., 145, 459 共1998兲. 25. M. M. Thackeray, Prog. Solid State Chem., 25, 1 共1997兲. 4. Activation energy similarities between the LTOc and BCF 26. K. Y. Chung, H. S. Lee, W.-S. Yoon, J. McBreen, and X.-Q. Yang, J. Electrochem. electrodes at temperatures below 0°C indicate that the binder and Soc., 153, A774 共2006兲. carbon present in the composite, porous electrodes are not signifi- 27. W. Choi and A. Manthiram, J. Electrochem. Soc., 153, A1760 共2006兲. 28. E. Ferg, R. J. Gummow, A. de Kock, and M. M. Thackeray, J. Electrochem. Soc., cant contributors to the low-temperature impedance increase. 141, L147 共1994兲. 5. The electrochemical reaction step at the electrolyte-active ma- 29. T. Ohzuku, A. Ueda, and N. Yamamoto, J. Electrochem. Soc., 142, 1431 共1995兲. terial interface appears to be responsible for electrode impedance 30. D. P. Abraham, S. D. Poppen, A. N. Jansen, J. Liu, and D. W. Dees, Electrochim. increase, especially at temperatures below 0°C. Acta, 49, 4763 共2004兲. 31. D. P. Abraham, E. M. Reynolds, P. L. Schultz, A. N. Jansen, and D. W. Dees, J. 6. Low-temperature performance may be improved by 共i兲 de- Electrochem. Soc., 153, A1610 共2006兲. creasing particle size or changing particle morphology to increase 32. M. C. Smart, B. V. Ratnakumar, V. S. Ryan-Mowrey, S. Surampudi, G. K. S. the electrochemically active surface area and 共ii兲 modifying the Prakash, J. Hu, and I. Cheung, J. Power Sources, 119–121, 359 共2003兲. 33. T. Abe, F. Sagane, M. Ohtsuka, Y. Iriyama, and Z. Ogumi, J. Electrochem. Soc., electrolyte–active material interface 共through electrolyte composi- 152, A2151 共2005兲. tion changes, for example兲 to reduce the high activation energy of 34. M. C. Smart, J. F. Whitacre, B. V. Ratnakumar, and K. Amine, J. Power Sources, the limiting process. 168, 501 共2007兲.