Li Metal
`Free Rechargeable LiMn
`?O?
`2
`4
`Their Understanding and Optimization
`
`−
`
`?/?Carbon Cells:
`
`D. Guyomard and J. M. Tarascon
`
`1992, Volume 139, Issue 4, Pages 937-948.J. Electrochem. Soc.
`
`doi: 10.1149/1.2069372
`
`Email alerting
`service
`
`Receive free email alerts when new articles cite this article - sign up
`
` click here
`in the box at the top right corner of the article or
`
`To subscribe to go to: Journal of The Electrochemical Society
`
`http://jes.ecsdl.org/subscriptions
`
`
`
`© 1992 ECS - The Electrochemical Society
`
`APPLE 1015
`
`1
`
`

`

` TECHNICAL PAPERS
`
`ELECTROCHEMICAL SCIENCE AND TECHNOLOGY
`
`Li Metal-Free Rechargeable LiMn20,/Carbon Cells: Their
`Understanding and Optimization
`
`
`
`D. Guyomard and J. M. Tarascon
`Bellcore, Red Bank, New Jersey 07701
`
`ABSTRACT
`
`So-called “rocking-chair” rechargeable batteries that use lithium intercalation compoundsfor the positive and nega-
`tive electrodes should be safer than batteries that contain free-lithium metal. Such a cell, using the spinel LiMn,0,as the
`positive electrode and carbon as the negative electrode, was optimized as a function of various operating parameters.
`These cells reversibly insert 0.32 Li atoms per Mn at an average output voltage of 3.7 V, yielding an effective specific
`energy of 250 mWh/g of electrode materials (3 times that of Ni-Cd). They can sustain high currentrates similar to Ni-Cd
`batteries, and can be discharged to 0 V without any degradation of their operating conditions. By systematically studying
`the stability of several electrolyte systems, we were able to minimize electrolyte decomposition (by controlling drastically
`the chargecut-off voltage) so that these cells show a promisingcycle life even at 55°C while maintaining 75% oftheir initial
`capacity.
`
`It is now recognized that ambient temperature non-
`aqueous secondary lithium cells using lithium metal ex-
`hibit problems which makeit difficult for their wide usage
`in the consumer market. These problems are: the short
`cycle life for deeply discharged cells, the high cost of the
`technology and, aboveall, the unsafe operating character-
`istics due to the high reactivity of lithium metal (1). To alle-
`viate this problem,a “rocking-chair” battery that uses an-
`other intercalation compound as the anode has been
`proposed (2-8). Among the alternative materials that could
`replace lithium metal as the negative electrode, carbon
`provides the best compromise between large specific ca-
`pacity and reversible cycling behavior (9-11). However, a
`price is paid in terms of average output voltage and energy
`density as comparedto a lithium metalcell, and this is the
`reason why a strongly oxidizing intercalation compound
`must be used as the positive electrode. The layered com-
`pounds LiNiO, and LiCoO, and the three-dimensional
`compound LiMn,0, with the spinel structure, are the only
`phasespresently knownto intercalate Li reversibly at volt-
`age larger than 3.5 V vs. Li.
`Recently, two battery companies, Moli Energy Ltd. and
`Sony Energytec Inc. have announced the future commer-
`cialization of Li* ion rechargeable batteries using a carbon
`intercalation compound as the negative electrode, and
`LINiO, and LiCoO, or LiNip2Cop,02 materials as the posi-
`tive electrodes, respectively (12-14). We also have recently
`demonstrated the feasibility of such recking-chair cells
`using the spinel LiMn,O, and its lithiated parent com-
`pound Li,Mn,0,as the positive electrode at 25°C (15).
`These rocking-chair batteries do not require a stringent
`manufacturing environment because the starting elec-
`trode materials (i.e. the lithiated manganese oxide and the
`carbon) are stable in ambient atmosphere. Thecell is as-
`sembled in its discharged state, where the output voltage
`is close to 0 V, and, consequently, can be handled before
`use without any fear of irreversible degradation due to a
`short circuit. The battery is activated during the first
`charge. This is similar to the well-known and widely used
`Ni-Cd batteries that need to be charged priorto their first
`use.
`The use of the spinel manganese oxide material offers
`the following advantages: LiMn,0,is easy to prepare, com-
`pared to the several steps required for the synthesis of
`LiNiO,, and Mn is more abundant and cheaper than Co, so
`
`that the overail electrode cost would be minimized if
`LiMn,0,is used; The theoretical energy density of LiNiO,
`and LiCoO,is twice as high as that of LiMn,O,, but in prac-
`tice only half of the Li content can be removed, thus lead-
`ing to the same order of usable energy density for the three
`materials. Moreover, the energy density of spinel-based
`manganeseoxide cells can be increased if LizMn,O, is used
`(15). The ability to synthesize LiMn,O, at temperatures as
`low as 400°C (lower than the temperature required to ob-
`tain LiNiO, and LiCoO,) also makes this oxide an attrac-
`tive material for microbatteries compatible with semicon-
`ductor
`technologies
`(e.g.,
`self-powered electronics).
`Finally, most of the battery community is already familiar
`with manganese-based oxides since most of the commer-
`cial primary batteries (MnO,/Li or MnO,/Zn) use a manga-
`nese-based oxide as the positive electrode so that toxicity
`df any) or recycling problemsare already well known.
`The use of a strongly oxidizing compound such as
`LiMn,0, as the positive electrode is worrisome since the
`delithiation of this compound occursat a very high voltage
`(about 4.1 V vs. Li/Li*) where the electrolyte oxidation can
`occur (16). Moreover, such an electrochemical reaction
`would be accelerated uponincreasing the temperature. We
`have systematically studied the electrolyte oxidation as a
`function of temperature and the liquid electrolyte used.
`We succeeded in minimizing the electrolyte decom-
`position and we show that the LiMn,0O,/carbon recharge-
`able cells can operate at 55°C with good reversibility and
`without measurable electrolyte oxidation.
`This paper details the electrochemical and cycling be-
`havior of the LiMn,0, spinel manganese oxide/carbon
`rocking-chair cells as a function of various parameters:
`The ratio of positive and negative electrode masses; The
`electrode thicknesses; The charge and discharge cut-off
`voltages; The current and; The operating temperature. We
`also report the figure-of-merit of these cells as compared to
`the widely used Ni-Cd battery and to the recently an-
`nounced Lit ion rechargeable battery (13). The electro-
`chemical behavior of each individual LiMn,O, and carbon
`electrode vs. Li in two different electrolytes will be pre-
`sented first and separately, with emphasis on the influence
`of the operating temperature and the voltage limits for cy-
`cling. Along with this study, we also have determined the
`chemical diffusion of Li* in the spinel LiMn,O, and in pe-
`troleum coke.
`
`J. Electrochem. Soc., Voi. 139, No. 4, April 1992 © The Electrochemical Society, Inc.
`
`937
`
`2
`
`

`

`938
`
`J. Electrochem. Soc., Vol. 139, No. 4, April 1992 © The Electrochemical Society, Inc.
`
`Two Electrodes
`
`Three Electrodes
`
`Spring
`
`Counter Electrode
`
`Current Collector
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Plungers ——___|
`
`Fig. 1. The drawing of 2 and 3 electrodes Swagelock cells are shown
`in (a) and (b), respectively.
`
`spectively, (for example a composite electrode containing
`90% LiMn,Q,, 8% carbon black, and 2% EPDM will be de-
`noted as [90, 8, 2]. A survey of several compositions was
`done for both composite electrodes in order to optimize
`the cyclability performance of the composite electrodes
`with respect to Li. The results of such a survey show that
`
`the compositions [89, 10, 1] and [94, 5, 1] give the best re-
`Data
`versibility vs. Li for the positive and negative electrodes,
`Separators —<-———|
`Acquisitlan
`Working Electrode
`respectively, as well as the lowest polarization (minimum
`Ref. Electrode (Li)
`difference in voltage between charge and discharge
`curves). These compositions were usedin all the results re-
`ported here. Although thermogravimetric analysis meas-
`urements (TGA)of the composite electrodes did not reveal
`any trace of water, all the electrodes were heated at 150°C
`during 1h, prior to being used as electrodes in rocking-
`chair batteries to eliminate any possibility of water con-
`tamination. For all the cells using LiMn,0, as the positive
`electrode, we took the Li content «x = 1 as the initial stoi-
`chiometry, and then the composition was automatically
`determined from the current, the massof the active cath-
`ode material used, and the elapsed time. From the values
`of x, the equivalent capacity in mAh/g reported in several
`Experimental
`figures is simply obtained by multiplying the theoretical
`The battery cycling studies presented herein have been
`capacity 148.2 mAh/g by Ax.
`made using standard and three electrode Swagelock cells
`Propylene carbonate (PC), with its low vapor pressure
`shownin Fig. 1. The plungers were madeout of a special
`and highflash point, was used for the solvent. To increase
`stainless steel (Carpenter, Ref. 20Cb-3), that allows oxida-
`the ionic conductivity of an electrolyte without changing
`tive voltages as large as 4.7 V vs. Li without any leakage
`the nature and content of the salt, mixtures of solvents are
`current due to the electrolyte oxidation. In such cells, the
`generally used. For instance, high dielectric constant sol-
`positive electrode (LiMn,0,) is in direct contact with the
`vents like ethylene carbonate (EC) or PC are mixed with
`plunger, while the negative electrode (Li or carbon) is in
`contact with a half-inch diameter nickel disk. Both elec-
`low dielectric constant and low viscosity solvents (like 2-
`methyl-tetrahydrofuran (2Me-THF) or dimethoxyethane
`trodes are separated by a 0.3 mm thick porous glass paper
`(DME)) (20). In this study, diethoxyethane (DEE) wasse-
`soaked in the electrolyte selected for the study. To ensure
`lected instead of the commonly used 2Me-THF or DME,
`physical contacts between the cell components,a stainless
`because its boiling point is higher (121°C compared to
`steel spring that allows pressures of about 1 kg/cm? is
`80°C), thereby limiting safety hazards(e.g., explosions) and
`placed between the nickel disk and the second plunger. In
`the solvent mixture consisting of EC and DEE, 50-50 by
`the three electrode design,the reference electrode,a Li foil
`weight, was tried. The solvents (Aldrich) were dried using
`plated onto a nickel wire sandwiched between two glass
`standard methods. For EC and PC, the solvents were de-
`paper disks is placed between the positive and negative
`gassed by 3 pump/refill cycles using Aras the inert gas. EC
`electrode pellets. These three electrode cells allow thesi-
`multaneousin situ determination of the three characteris-
`was warmed to above the melting point. The solvents were
`dried over 4 A molecular sieves for 48 h and then weredis-
`tics, Voositive US. Li, Vnegative vs. Li, and Voositive vs. Vnegatives pro-
`tilled under reduced pressure(15 Torr). DEE was degassed
`viding us the information necessary to fully understand,
`by 3 freeze-pump-thaw cycles, and was distilled under Ar
`and optimize, the behavior of these secondary cells. For
`from K benzophenoneketyl.
`the cyclovoltammetry measurements, a three electrode
`Different electrolytes have been made by adding various
`glass cell, consisting of a working electrode, a reference
`widely used Li salts (heat-vacuum dried prior to their use),
`electrode (piece of lithium foil plated onto a nickel wire),
`such as LiAsF.,, LiCF;SO3, LiBFy, LiPF., LiN(CF3S0.)2,
`and a counterelectrode (made the same way as the refer-
`and LiClO, to PC or EC + DEE (50:50). After the drying
`ence electrode) was used. The working electrode was ei-
`process, the x-ray powderdiffractograms were controlled
`ther a glassy carbon electrode (to study the oxidation limit
`of some electrolytes) or a pellet of tested material com-
`to exclude any possibility of degradation. The best Li-salt,
`defined in this study as the one allowing the largest elec-
`pacted onto a platinum gauze (for the electrochemical
`troactivity range at high voltage, was selected as follows.
`characterization of this material). Both the Swagelockcells
`First, the oxidation limit of the electroactivity range of
`and the three electrode glass cell were always assembled
`these electrolytes has been determined by cyclic voltam-
`in a helium atmosphere. Then, the air-tight Swagelock
`metry on a glassy carbon electrode. The oxidation current
`cells were removed from the dry box to be tested. No loss
`of the electrolytes is located in the range 4.6 to 5.0 V vs. Li,
`of electrolyte was detected for these cells even after a
`showingthatall these salts could be used up toavoltage of
`series of tests at 55°C during 1 month.
`4.5 V vs. Liin any battery. This measurement, however,is
`LiMn,O, materials were prepared by reactingin air stoi-
`not fully indicative since it does not reflect the true operat-
`chiometric amounts of LizCO3; and MnO,placed into an
`alumina crucible. In accordance with our previous work
`ing conditions because, the catalytic activity of any posi-
`tive electrode material could be different from that of the
`(17), because of their larger capacity, LiMn,O, powders
`glassy carbon, and the standard scan rate used for metal
`(1 to 2 wm particle size) prepared by three consecutive an-
`neals at 800°C (24h each) were used as the positive elec-
`electrodes (0.1 V/s) is much higher than that used for cy-
`trodes. Petroleum coke (Conoco, —200/400 mesh; i.e. parti-
`cling an intercalation electrode material (10-* to 10-5 V/s).
`Thus, these different electrolytes have been tested under
`cle size between 38 and 75 ym) was used instead of
`graphite as the negative electrode, because of a lower loss
`real operating conditions in LiMn.O/electrolyte/Li cells
`of capacity compared to graphite at the first cycle, al-
`up to 4.5 V and thesalt allowing for largest charge cut-off
`though its capacity is twice as small (its maximum Li com-
`voltage on the LiMn,0, material, over a large numberof
`position has been reported to be LigsC,g instead of LiC,
`cycles without any oxidation of the electrolyte, is LiClO.
`(18, 19)). Composite electrode materials containing the ac-
`Othersalts can also be used with LiMn,O,, but they allow a
`tive material, carbon black, and EPDM (ethylene propyl-
`smaller electroactivity range. Consequently, from this
`study, the two electrolytes PC + 1M LiClO, and EC + DEE
`ene diene monomer) rubber were made by mixing the
`(50:50) + 1M LiClO, were selected and Fig. 2 shows their
`powdersin a solution of EPDM in cyclohexane, evaporat-
`oxidation limit measured by cyclic voltammetry at the two
`ing the solvent, sieving the resulting powders to 100 pm
`temperatures used in this work. The PC-based electrolyte
`and pressing them at a pressure of 10 tons/cm?into pellets.
`The resulting composite material will be denoted by three
`allows higher voltage values, but for both electrolytes the
`numbers[a, b, c} where a, b, and c will refer to the weight
`oxidation limit is lowered by 0.2 V when the operating
`temperature is increased to 55°C.
`ratio of the active material, carbon black, and EPDM,re-
`
`3
`
`

`

`J. Electrochem. Soc., Vol. 139, No. 4, April 1992 © The Electrochemical Society, Inc.
`
`
`
`
`
`
`!|
`
`9+ —-—~--
`
`_
`
`Voltage (volts vs. Li)
`
`Fig. 2. Cyclic voltammograms at glassy carbon electrode of EC +
`DEE (50:50) + 1M LiClO, and PC + 1M LiCIO,electrolytes at 25 and
`55°C. Scan rate = 0.1 V/s.
`
`The ionic conductivity and the water contentfor the se-
`lected electrolytes used for our experiments were deter-
`mined using a high frequency impedance analyzer and a
`Karl-Fisher coulometer, respectively. Ionic conductivity
`values of 10.4 and 5.62 mS/cm at room temperature (25°C)
`and water contents of45 and 40 ppm were obtained,for the
`twoelectrolytes EC + DEE (50:50) + 1M LiClO,, and PC +
`1M LiClO,, respectively.
`The chemical diffusion coefficients of Li* ions in both
`LiMn,0, and carbon electrodes were measured using the
`potentiostatic intermittent
`titration technique (PITT)
`(21, 22) by meansof a special multichannel galvanostatic/
`potentiostatic system “Mac-Pile.” In this system the po-
`tentiostatic mode acts as a coulometer so that the experi-
`mental parameteris the charge incrementvs. time and not
`the current vs. time as usually measured. Thus, in our ex-
`periment, the potential of the material/Licell is monitored
`by steps, and the decay of the charge increment(the inte-
`gral of the current) during each step is measured as a func-
`tion of time. The resolution of Fick’s second law for times
`t << P/D and t >> I/D leads to the Eq.[1] and [2], respec-
`tively, for the variation of the charge incrementas a func-
`tion of time
`
`a(t)
`
`_ 2FSD"AC
`T12
`
`Vt
`
`(t)
`
`a
`
`8FSIAC
`7
`
`( =)
`4?
`
`P
`
`[1]
`
`[2]
`
`where t is the elapsed time from the beginning ofthe step,
`F is the Faraday constant, S is the total surface area of the
`grains of the active material, AC is the variation of Li con-
`centration into the material during the voltage step, D is
`the chemical diffusion coefficient of Li* into the material,
`and lis the size of the material grains. The chemical diffu-
`sion coefficient D of Lit can then be determined by the
`slope of the linear plot of q vs. Vt in the short time approxi-
`mation, and by the slope of the linearplot of log q vs. t in
`the long time approximation. The accuracy of this tech-
`nique directly depends on the evaluation of the total sur-
`face area. In this work, we have estimated this surface area
`by assuming that the grains are spherical with a diameter
`equal to the averageparticle size.
`Results and Discussion
`Behavior of LiMn,O, vs. Li—Figure 3a shows the room
`temperature cycling behavior of a LiMn,O,/Li cell under
`constant current
`in the two electrolytes, EC + DEE
`(50:50) + 1M LiClO, and PC + 1M LiClO,. The cellis first
`charged from its rest voltage (close to 3.3 V vs. Li) to 4.5 V
`
`0.2
`
`0.4
`
`0.6
`
`0.8
`
`1
`
`x in Li, Mn2 Og
`
`Fig. 3. Typical cycling behavior at 25 (a) and 55°C (b) of LiMn,0,/Li
`cells using EC + DEE (50:50) + 1M LiCIO, and PC + 1M LiCIO,elec-
`trodes. In (a) the cells contain 30 mg (of active material) and were cy-
`cled between 3.5 and 4.5 V at a current of 0.4 mA/cm? (Ax = 1
`in
`22 h). In (b) the cells contain 10 mg (of active material) and were cy-
`cled between 3.5 and 4.3 V at a current of 0.2 mA/em? (Ax = 1
`in
`15 h).
`
`vs. Li. The removal of one Li' from the structure (leading
`to A-MnO,) occurs in two steps close to 4.1V vs. Li as
`shownin previous work (17, 23-26). On discharge, \-MnO,
`converts back reversibly to LiMn,.Q, in two steps. Inde-
`pendently of the electrolyte used, the material shows the
`same reversible capacity after the first charge-discharge
`cycle, Ax = 0.8, corresponding to 119 mAh/g of active ma-
`terial. The loss in capacity during thefirst cycle (Ax = 0.2)
`may arise from particles that becomeelectrically discon-
`nected andalso results from the fact that the battery is cy-
`eled under constant current without any relaxation so as to
`allow the system to reach equilibrium upon Li* interca-
`lation. For composite LiMn,O, electrodes of the same
`thickness, the polarization (Venarge — Vaischarge) 1s about three
`times lower in the EC- than in the PC-based electrolyte
`(Fig. 3a), consistent with the higher ionic conductivity
`measured for the EC- than the PC-basedelectrolyte.
`In order to determine the maximumvoltage to which re-
`chargeable Li cells based on LiMn,0,as the positive elec-
`trode can operate safely, we havetried several cut-off volt-
`ages on charge. Figure 4 shows what happensif the charge
`cut-off voltage of a LiMn.O,/Li cell is increased after a few
`cycles. A plateau is obtained at 4.59 V vs. Li, correspond-
`ing to an irreversible electrochemical reaction. Asa result,
`the next discharge curveis shifted to lower values of x,
`with a value corresponding to the capacity length of this
`plateau. This reaction could be the oxidative degradation
`of the )-MnO, framework or the oxidation of the electro-
`lyte at the surface of the material. The x-ray diffraction pat-
`tern of the sampleafter this treatment shows nodifference
`
` “oa
`
`939
`
`
`
`5.0
`
`Specific Capacity (mAh/g}
`100
`50
`
`0
`
`Discharge
`Electrolyte = PC + LiICIOg
`
`Discharge
`Electralyte = EC +DEE + LiClO,
`
`Electrolyie = PC + LiClO,
`
`2 g
`
`g £
`
`_
`4
`2
`Oo
`=®
`
`> £
`
`g3
`
`>
`
`|
`:
`<
`ios
`i
`
`$10—Salt= LiClO, | |
`
`5»>
`yoy
`E
`S| eee PC at 55°C
`myo
`
`=
`—-—-— PC at 25°C
`Al
`Electrolyte = EC + DEE + LiClO,
`
`6|----- EC + DEE at 25°C il:
`oD
`2 0.5 +} ——— EC +DEE at 55°C
`Abdo
`5
`|
`,
`+
`oO
`
`4
`
`

`

`J. Electrochem. Soc., Vol. 139, No. 4, April 1992 © The Electrochemical Society, tnc.
`
`T
`
`Specific Capacity (mAh/g)
`
`200
`160
`120
`80
`40
`0
`| (a)
`25°C
`
`5.0. —7TtT T T T_T 71 ii
`
`05
`'
`LiMnzO4/Li CELL AT 25°C
`< |
`j
`IN EC + DEE + LiclO,g
`—€
`|
`a
`5 ob
`a
`5
`
`940
`
`4.6L
`a 4wo KN
`
`@ 42}KN
`s
`|SSS
`2 3.8}
`S>
`3.4)-
`
`4
`
`
`
`
` Pq TT Thy 7
`ral
`
`PC+ LICIO,
`
`pooLb
`
` ewn
`
`
`0.4 mAfom?
`
`3.0
`‘
`
`Lop ft
`0
`0.2
`0.4
`0.6
`0.8
`4.0
`xin LiyMngQ4
`
`Fig. 4. Cycling behavior at 25°C of a LiMn20,/Li cell in EC + DEE
`(50:50) + 1M LiClO, when the charge cut-off voltage is increased
`above 4.5 V. The pellet was 30 mg (active material) and a current rate
`of 0.4 mA/cm? was used (Ax = 1
`in 22h). The charge cut-off voltage
`was 4.5 V for the 2 first charges, then 4.6 V for the third charge, and
`then 4.4 V for the following charges. The discharge cut-off voltage was
`always 3.5 Y.
`
`
`
`Current(mA)
`
`|
`
`0.5
`
`QoTo
`
`with the one for \-MnO,, excluding thefirst possibility. In
`an attempt to avoid further oxidation ofthe electrolyte, the
`following charges were limited to 4.4 V vs. Li. However, at
`each successive cycle, even if the voltage is kept lower
`than 4.4 V vs. Li, the voltage-composition curves show
`lower capacity and larger polarization than a cell that has
`not been charged above 4.5 V vs. Li. In addition, these
`curves are shifted to larger capacity values, due to someir-
`reversible behavior coming from the oxidation of the elec-
`trolyte. After a few cycles, the high voltage cannot be
`reached and the delithiation of the spinel no longer occurs
`(e.g. the cell cannot be used any longer). These results indi-
`cate that once the oxidation of the electrolyte has been ini-
`tiated by an increase in cut-off voltage, its decomposition
`voltage is lowered into the cycling window so that it can-
`not be stopped even by lowering the cut-off voltage on
`subsequent cycles. This emphasizes the importance of
`carefully controlling the overvoltage in charging these
`cells: 4.45 V vs. Li is a safe limit at room temperature that
`allows one to use the whole capacity withoutrisk of elec-
`trolyte decomposition.
`The behaviorof the LiMn,O, material at high voltage has
`been studied further by cyclic voltammetry at room tem-
`perature. The three electrode technique allows one to
`study only the contribution of the LiMn,O,/electrolyte in-
`terface and consequently avoid any phenomenon occur-
`ring at the Li electrode or any secondary reaction suscep-
`tible to occur at the metal current collector in Swagelock
`test cells. The voltammogramsof Fig. 5 show two peaksin
`oxidation and two peaks in reductian, when the electrode
`material is cycled between 3.3 and 4.45 V vs. Li, which are
`the signature of the two-step reversible transformation be-
`tween LiMn,O, and 4-MnO,(17, 21-23). If the higher volt-
`age limit is increased above 4.5 V vs. Li, an oxidation cur-
`rent appears on the anodic sean direction. On the
`following cathodic scan, above 4.5 V vs. Li, this current is
`always oxidative, so it leads to irreversible capacity. This
`current is intrinsic to the material, due to oxidation of the
`electrolyte on the electrode material surface, because the
`oxidation of the electrolyte on the Pt holder remains slow
`in the same voltage range (dot-dashed line in Fig. 5). On
`the subsequent cycles, the current peaks becomelarger
`and badly defined, in contrast to that observedif the volt-
`age is always kept below the limit 4.5 V vs. Li. This degra-
`dation of the behavior of the manganese oxide could be
`due to the deposition of an insulating layer, composed of
`the products of the electrolyte oxidation, at the surface of
`the grains. Comparing the curvesobtainedin both electro-
`lytes, we see that the voltage separation between the peaks
`in the oxidation scan direction and in the reduction scan
`
`Voltage (volts vs. Li)
`
`Fig. 5. Cyclic voltammograms at LiMn,0, in EC + DEE (50:50) +
`1M LiClO, and PC + 1M LiClO, electrolytes at 25°C (a) and at 55°C
`{b). For all the voltammograms the starting voltage is close to 3.2 V
`and the scan rate is of 0.1 mV/s. The solid line, dashed line and dotline
`refer to the Ist, 2nd, and 4th cycle, respectively, while the dot-dashed
`line is obtained as a blank on the Pt grid holder alone until very high
`voltageis used. For the voltammograms {a} the voitageis first scanned
`to 4.45 V and back to 3.3 V (solid line) and then to 4.75 (for EC-based
`electrolyte) or 4.95 V (for PC-based electrolyte) for following sweeps
`in the positive direction. For the voltammograms (b)the voltageisfirst
`scanned to 4.42 (for EC-based electrolyte) or 4.37 V (for PC-based
`electrolyte) and back to 3.2 V (solid line) and then to 4.7 (for EC-) or
`4.85 V (for PC-) for following sweeps in the positive direction.
`
`direction(e.g., the polarization) is lower in EC- than in PC-
`based electrolyte, because the ionic conductivities of the
`two media are different. Moreover, the oxidation of the
`electrolyte on the material surface occurs at a lowervolt-
`age and the degradation of the shape of the voltammo-
`gramsoccurs faster in the EC-based electrolyte. From this
`study, we conclude that the degradation of the reversible
`eycling behavior of the LiMn,O, material after an over-
`charge exceeding 45V vs. Li
`is
`intrinsic to the
`LiMn,0,/electrolyte interface and arises from the electro-
`lyte oxidative decomposition at the surface of the material
`grains.
`Another important question to be addressed is what
`happensat higher operating temperature where a battery
`operates in practice. Figure 3b showsthe cycling behavior
`of LiMn,0,/Li cells at 55°C under constant current in the
`two electrolytes we selected. Note that a lower charge cut-
`off voltage than the one used at room temperature (4.3 V
`vs. Li instead of 4.5 V vs. Li) was sufficient to obtain the
`complete capacity. Again, the polarization is greater in PC-
`than in EC-based electrolytes but less than that observed
`at room temperature, as expected, because of an increase
`in the ionie diffusion rate with increasing temperature.
`The loss of capacity during the first few cyclesis slightly
`larger than at room temperature, and is slightly larger in
`PC- than in EC-based electrolytes.
`The samecyclic voltammetry study as was done at room
`temperature shows that at 55°C the electrolyte oxidation
`limit is effectively lowered by about 0.2 V in EC and 0.25 V
`in PC (Fig. 5b). As observed at room temperature the elec-
`trolyte oxidation still occurs at a higher voltage in PC than
`
`5
`
`

`

`J. Electrochem. Soc., Vol. 139, No. 4, April 1992 © The Electrochemical Society, Inc.
`
`941
`
`Specific Capacity (mAh/g)
`100
`200
`300
`0.5 mAfem
`
`55°C,
`
`
`
`
`PC + LICIO,
`
`Electrolyte =
`EC + DEE + LiClO,
`
`Electrolyte =
`
`4
`
`4
`
`4
`se.
`4
`
`4
`
`Fig. 6. Typical cycling behavior
`at 25°C (a) and at 55°C (b) of car-
`bon
`(petroleum coke)/Li
`cells
`using the electrolytes EC + DEE
`(50:50) + 1M LiClO, and PC +
`1M LiClO,. All the cells contain
`30 mg (of active material) and are
`cycled between 0.02 V and 1.8 V
`at
`a
`current
`of
`0.5 mA/cm?
`(Ax = 0.5 in Li,C. was obtained in
`22 h).
`
`0
`
`Discharge
`
`(voltsvsLi)
`
`Specific Capacity (mAh/g)
`100
`200
`
`0
`
`300
`
`0.5 mA/cm?
`
`Electrolyte =
`EC + DEE + LiClO,
`
`Electrolyte =
`PC + LICIO,
`
`0
`
`0.2
`
`0.4
`
`0.6
`
`0.8
`
`Voltage
`
`0
`
`0.2
`
`0.4
`
`0.6
`
`0.8
`
`xin Li,Cg
`
`x in Li, Cg
`
`EC, and the polarization is always higher in PC than in EC.
`Compared to what happened at room temperature, the
`degradation of the cycling behavior of the cell occurred
`faster. Note that at 55°C, using a scan rate of 0.1 mV/s, cor-
`responding to a charge-discharge cycle of the manganese
`oxide of less than 2h instead of 18 h under galvanostatic
`control (Fig. 3) the current trace in EC (Fig. 5b) does not
`reach a low value at 4.5 V vs. Li suggesting that the oxida-
`tion of the electrolyte begins before the end of the delithi-
`ation of the spinel. However, at such high scan rates the
`width of the peaksare larger and, equivalently, the electro-
`chemical reactions occur in a larger voltage range. The use
`of an adequate charge cut-off voltage (4.3 V vs. Li in the
`present conditions), will avoid the electrolyte oxidation
`and yetstill allow an acceptable capacity for the deinterca-
`lation reaction.
`From this study and the study of the cycling behavior of
`LiMn,0,/Li cells under constant current, we conclude that
`the degradation of the electrolyte begins at a voltage far
`enough from the delithiation voltage of the spinel, so that
`LiMn,0O,can be used as a positive electrode, even for cells
`operating at 55°C.
`
`ilar to that at room temperature. However, the irreversibil-
`ity at the first discharge is slightly higher than at room
`temperature (27% instead of 23% in EC + DEE (50:50) and
`25% instead of 21% in PC) while the capacity loss during
`the 5 first cycles is slightly higher than at room tempera-
`ture. The reversible capacity in EC + DEE (50:50) + 1M
`LiCl0,at 55°C is found to be Ax = 0.55 in Li,Cs, about 10%
`higher than what is observed at room temperature.
`The voltage limits, in which the carbon electrode can
`work properly, have been studied carefully. The Lit ions
`are completely removed from the carbon electrode at a
`voltage higher than 1.5 V vs. Li, but the charge can bein-
`creased until 3.9 V vs. Li with negligible excess capacity
`and without affecting the reversible behavior upon cy-
`cling, because 1.5 to 3.9 V is an electrochemically inactive
`voltage range for carbon. Figure 7a showsthefirst cycle
`downto very low voltage of a carbon/Li cell obtained using
`periodic times of constant current and relaxation under
`open circuit. The cell has been discharged to —0.045 V vs.
`Li undercurrent, with a relaxation open-circuit voltage al-
`wayspositive vs. Li. No changein the slope of the voltage-
`composition curve is observed, indicating that over this
`range of potential the intercalation of Li* ions into the car-
`Behavior of carbon vs. Li.—Figure 6a showsthe cycling
`bon structure is the only electrochemical process occur-
`behavior at room temperature of a carbon (petroleum
`ring. A negative voltage vs. Li is meaningless. In fact, the
`coke)/Li cell under constant current in the two electrolytes
`intrinsic voltage is always positive, as expected, but the IR
`EC + DEE (50:50) + 1M LiClO,, and PC + 1M LiClO,. The
`term (R being the internal resistance of the cell and I the
`cell is first discharged from its rest voltage (usually in the
`current flow) always addsto theintrinsic voltage and gives
`range 3.0 to 3.3 V vs. Li) to 0.02 V vs. Li. The intercalation
`a negative measured value. The negative voltage limit used
`of Li" ionsin the structure occurs without the formation of
`here depends strongly then on the experimental condi-
`any staged-phase becauseofthe lack of crystalline orderof
`tions through the term IR (e.g., R increases with the thick-
`the material, as was shown in previous work (11, 18). The
`ness of the pellet). The structural changes in the carbon
`charge is much shorterthan thefirst discharge, but the fol-
`during this deep discharge were followed using an in situ
`lowing cycles show excellent reversibility over Ax = 0.5 in
`x-ray cell over the range of 20 correspondingto the [002]
`Li,Ce, corresponding to 186 mAh/g of active material. The
`Bragg peak whose changes in position directly reflect
`large irreversible capacity loss after the first cycle has been
`changesin the spacing between two consecutive graphite
`attributed to the formation of a passivating film at the
`layers. The width at half maximum ofthis diffraction peak
`grain surfaces, often called solid electrolyte interface
`was found to remain constant during the cell discharge in
`(SED, arising from the reduction ofthe solvent at a voltage
`contrast to its position. Figure 8 showsthat the interlayer
`lower than 1.2 V vs. Li (18). On the following cycles, this
`distance d(002) varies continuously as a function of x in the
`passivating layer protects the carbon from any direct con-
`range 0 < x<0.7, confirming that Li intercalation is the
`tact with the electrolyte, preventing its further reduction
`only phenomenon occurring in this voltage range. When
`and allowing only for the Li intercalation-deintercalation
`the cell is recharged the [002] Bragg peak converts back to
`reaction. This loss in capacity duringthefirst cycle is al-
`its original position indicating further that the interca-
`most the same in EC (28%) and in PC (21%), but the polar-
`lation process is reversible over this range of composition
`ization is twice as high in PC than in EC.
`and voltage. These resul