Electrochimica Acta 47 (2002) 1217– 1223

www.elsevier.com/locate/electacta

Mechanism of self-discharge in graphite–lithium anode

Rachid Yazami *,1, Yvan F. Reynier
LEPMI INPG/CNRS 5631 -BP75, 38402 Saint Martin D’Hères, France

Received 23 May 2001; received in revised form 20 August 2001

Abstract

In order to evaluate the anode contribution to the lithium-ion battery self-discharge, three electrode coin cells composed of
metallic lithium as reference and counter electrode, organic liquid electrolyte and graphite composite working electrode were
constructed as test cells. They were first cycled for a dozen cycles and then stored in the full lithiated state of graphite, at 70 °C
for periods from 1 to 8 days. The capacity loss was determined during the first delithiation following storage. The latter was found
composed of two terms, a reversible and an irreversible one, where the relative amounts are storage time dependants.
Electrochemical impedance spectroscopy (EIS) was used to investigate the changes in the cell interfacial characteristics. A model
involving the formation of an absorbed electron-ion–electrolyte complex on the graphite surface is proposed as the mechanism
of the reversible and irreversible capacity losses. It is also suggested that precipitation/dissolution reactions are taking place at the
solid electrolyte interphase (SEI). Precipitation occurs with insoluble inorganic species such as, LiF and Li2CO3, whereas
dissolution may concern the organic and/or polymer part of the SEI. The continuous growth of the inorganic (and most resistive)
part of the SEI with the subsequent electrode isolation is proposed as the major mechanism of the electrode end of life. © 2002
Elsevier Science Ltd. All rights reserved.

Keywords: Graphite– lithium; Self-discharge; Thermal ageing; SEI; Impedance; Lithium-ion batteries; Electrode; Isolation

1. Introduction will increase too, which culminates in the battery end of

life.
A primary or a secondary battery at full or partial The major factors contributing to the battery self-dis-
initial state of charge (SOC) will naturally lose its charge can be listed as:
charge capacity when kept for a period of time at a “ Internal electron leakage coming from the electrolyte
certain temperature. This phenomenon is referred to as partial electronic conductivity, or other internal
battery self-discharge, in which the kinetics will vary shorts.
depending on the battery chemistry, electrode composi-
“ External electron leakage resulting from the poor
tion and design, electrolyte formulation and impurities,
isolating properties of the battery seals or gasket or
and on the storage temperature. Primary batteries by
from external finite resistance between the leads (ex-
definition will experience only irreversible capacity
losses; secondary batteries may have a part of the lost ternal conductors, humidity…).
capacity restored following recharge. Such restored ca- “ Electrode/electrolyte reactions such as, anode corro-
pacity is the reversible part of the capacity loss, in sion or cathode reduction by the electrolyte or
contrast to the non-recovered one after recharge, which impurities.
then is the irreversible capacity loss. With increased “ Partial dissolution of the electrodes active material.
storage temperature and time, the irreversible part “ Electrode passivation by decomposition products
(insoluble species or adsorbed gases).
“ Electrode mechanical disintegration or isolation
* Corresponding author. Present address: Division of Engineering from current collectors.
and Applied Science, California Institute of Technology 138-78, “ Internal pressure built up and electrolyte leakage.
Pasadena, CA 91125, USA; tel.: + 1-626-3954496; fax: +1-626-
7956132.
Each of the factors listed may contribute more or less
E-mail address: [email protected] (R. Yazami). to the overall capacity loss. This makes it difficult to
1
ISE Member. predict the cell shelf life with good accuracy. However,

0013-4686/02/$ - see front matter © 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 1 3 - 4 6 8 6 ( 0 1 ) 0 0 8 2 7 - 1
1218 R. Yazami, Y.F. Reynier / Electrochimica Acta 47 (2002) 1217–1223

a proper choice of the active (electrodes, electrolyte) cro-porous separator provided by Tonen (Japan) was
and inactive (separator, gasket, hardware) cell compo- used.
nents together with advanced cell design and sealing Cells were sealed in the dry box, and then transferred
technology will reduce the effect of some factors, such outside for electrochemical measurements. They were
as internal or external leakage and mechanical integrity. subjected to 10–15 first cycles under C/5 rate, between
This is particularly true in the case of lithium ion 1.5 and 0 V versus Li to determine the initial capacity
batteries, which constitute the most advanced systems Qrev(0). The final operation consisted of a cell discharge
offered in the consumer market and designed to operate down to 0 mV versus Li, for which the graphite compo-
high-tech devices such as cellular phones and laptop sition is close to LiC6 (full lithiation). Cells were then
computers for the most. stored at open-circuit at 709 1 °C for periods of 1–4
Using prototype lithium ion batteries, a recent study and 8 days, then were charged to 1.5 V to determine the
carried out by the SAFT group [1] showed that the remaining capacity Qrem(t). This was followed by ten
major contribution to self-discharge is anode corrosion. cycles to determine the new reversible capacity, Qrev(t).
However, more specific anode study requires the use of EIS measurements were performed at end of the first
Li/electrolyte/graphite half-cells with internal reference. cycles at the charged state and after thermal storage.
We have conducted such a study and published prelim- The applied signal amplitude was 5 mV and the fre-
inary results based on combined galvanostatic cycling quency varied between 100 kHz and 10 mHz. Two cells
and electrochemical impedance spectroscopy (EIS) were constructed for each experiment to test for
measurements [2]. We then suggested a mechanism reproducibility.
involving continuous electrical isolation of the graphite
grains as result of increased thickness of the inorganic
component of the solid electrolyte interphase (SEI), the 3. Results and discussion
latter being the reaction products of PF− 6 and solvent
reduction by the active lithium stored in Lix C6. 3.1. Gal6anostatic cycling
In order to determine more precisely the time evolu-
tion of the reversible and irreversible capacity losses To illustrate the changes in the discharge– charge
together with that of the Lix C6/electrolyte interface cycle curves before and after storage, we drew in Fig.
characteristics, we have conducted a study on a larger 1a the last cycle before storage and the first cycle after
number of identical cells that have been stored at 70 °C 4 storage days in Fig. 1b. Differences are clearly seen
for a period of time up to 8 days. on both time and voltage axes, which are related to
capacity loss and cell polarization, respectively. The
total capacity loss determined from the time difference
2. Experimental between the last charge before storage and the first one
after storage was found in the 40–41% range. Increased
Two or three electrode coin cells were constructed cell polarization may result from changes in the elec-
using an in-house technique [3]. The working electrode trodes (lithium or graphite)/electrolyte interfacial char-
was prepared by spreading a mixture of graphite, acety- acteristics and/or from increased internal ohmic
lene black and PVDF in acetone on polyethylene (PE) resistance or from changes in the lithium transport into
substrate, followed by calendering and drying in air at the graphite bulk structure. This will be detailed in the
ambient temperature, then at 60 °C for few hours. The next section on the ISE measurements.
electrode weight composition was 75% graphite (SFG6 Noteworthy is the changes in the curves profile as the
from Timcal), 10% acetylene black (Y70, PCUK) and typical stage transition voltage plateaus which appeared
15% PVDF (Ato-Fina (Elf Atochem) co-polymer, grade in Fig. 1a become less well defined in Fig. 1b. In the
c2801). The electrode was then separated from its PE latter, the discharge branch took more time than the
substrate and disks of 16 mm on diameter were cut and charge one, which means that more lithium was re-in-
dried in vacuum at 80 °C for 5 h. Dried disc elec- tercalated than first de-intercalated. This time difference
trodes were introduced in the argon filled dry box and almost coincides with an unusual— although slight—
weighed. The electrolyte used in this study is 1 M LiPF6 curvature inversion in the final step of intercalation
in EC –DMC (1:1 vol.) mixture. The H2O content was curve as marked with an arrow in Fig. 1b. Such inver-
analyzed by Karl– Fischer titration technique and sion disappeared in the following cycles. Therefore,
found below 10 ppm. The HF content was estimated excess intercalated lithium probably involves a different
below 20 ppm from 19F NMR measurement. Coin cells intercalation mechanism characteristic of thermally
(CR-2430) used lithium disks as counter electrode. For aged electrode.
three electrode cells, a lithium chip was pressed on a Fig. 2a–e show the cell cycle profile before and after
fine Cu grid to which a Cu fine wire was attached and storage for 1–4 and 8 days, respectively. The initial
used for external voltage measurements. A PE:PP mi- reversible capacity stabilizes around 3559 15 mAh
 R. Yazami, Y.F. Reynier / Electrochimica Acta 47 (2002) 1217–1223 1219

g − 1. After storage, the charge curve shows a ‘V’ shape “ Re, electrolyte resistance.
(dot line on Fig. 2a) indicating the capacity loss and its “ Rf, Cf corresponding to the SEI passivation film.
total or partial recovery. “ Charge transfer and diffusion circuit, with double
Whereas, after 1 and 2 days most of the initial layer capacitance Cdl in parallel with the charge
capacity was recovered, irreversible losses appeared transfer resistance Rct and a Warburg type diffusion
only after 3 storage days. The extent of the latter was impedance W.
significantly increased after 8 days as the capacity Figs. 7 and 8 show the evolution with the storage time
dropped to around Qrev(8 days) = 150 mAh g − 1 com- of (Re, Rf, and Rct) and that of (Cf and Cdl), respectively.
pared with the initial Qrev(0) =350 mAh g − 1. Fig. 3 The electrolyte resistance increased gradually with time
summarizes the capacity losses with storage time and and in a stepwise mode. It increased by 100% after 4 days
indicates the relative reversible and irreversible parts. then almost stabilized. The film resistance increased by
Fig. 4 depicts the log Qrem(t)/Qrev(0) versus time nearly 400% after 3 days, then it decreased after 4
curve. It shows a linear behavior, indicating a first additional days. A similar behavior is observed with the
order kinetic law: charge transfer resistance. Concomitantly the film and
the double layer capacitance gradually decreased.
Qrem(t)= Qrev(0) exp(−ut)
Although the fitting parameters of the equivalent
The discharge kinetic constant was found u =0.17 circuit are approximate, the general trends of the resis-
per day. This translates into a cell half-life close to 4 tance and capacitance led us to the following
days at 70 °C. considerations.
“ The low frequency Warburg branch associated with
3.2. Impedance measurements finite lithium diffusion was not much affected in slope
and frequency after storage as shown in Fig. 5.
Typical Cole– Cole impedance spectra before and Therefore, diffusion should have a low effect on cell
after 2 days storage are shown in Fig. 5. The equivalent polarization shown in Fig. 1. Instead, increased
circuit that gives a good fit to these curves is drawn in equivalent circuit resistances should play the major
Fig. 6. It consists of three elements in series: role.

Fig. 1. Discharge – charge curves profile before (a) and after (b) 4 days of storage. The arrow in curve (b) indicates an unusual change in the
curvature, which may be the signature of the adsorbed complex re-dissociation.
1220 R. Yazami, Y.F. Reynier / Electrochimica Acta 47 (2002) 1217–1223

Fig. 2. Evolution of the capacity before and after 70 °C storage; (a) 1 day; (b) 2 days; (c) 3 days; (d) 4 days; (e) 8 days.

“ The appearance of a maximum in the Rf and Rct As shown in Fig. 9, X-ray diffraction (XRD) per-
suggests two reverse effects, one leading to increased formed on the graphite electrode before and after stor-
resistances, such new species precipitation to form the age indicated no significant structural changes, in
SEI, and another decreasing the resistances as a likely particular, no solvent co-intercalation or graphite exfo-
result of a partial SEI dissolution. liation could be detected. Therefore, the capacity loss is
“ Decreased capacitance should be associated with de- mainly due to surface effects.
creased active surface area of the electrode. This may
result from increased amounts of the isolating com-
ponents of the SEI, i.e. the inorganic part composed
of LiF and Li2CO3, which, respectively, results from
PF− 6 and EC (and/or DMC) reduction by active
lithium.
“ The increase of the electrolyte resistance may result
either from the decreased active surface accessible to
the ions or to increased resistivity due to the elec-
trolyte consumption (PF− 6 and EC:DMC), or to
increased concentration of soluble organic species
which may hinder the ions motion. Fig. 3. Determination of the self discharge coefficient.
 R. Yazami, Y.F. Reynier / Electrochimica Acta 47 (2002) 1217–1223 1221

transfer to the latter. Then an adsorbed chemical com-

plex should form involving the edge layers dangling
carbon, the electron and the electrolyte molecules. The
electron is then shared between the carbon and the
electrolyte through the lithium ion. This mechanism is
sketched in Fig. 10. The so formed electron-ion–elec-
trolyte complex is metastable.
On one hand it is stable in the sense that during the
electrode de-lithiation (cell charge) after storage, the
electron would not be used to accompany the lithium
Fig. 4. Evolution of reversible and irreversible capacity losses after
de-intercalation and therefore would account for the
1–8 days 70 °C storage.
capacity loss. During the following discharge (re-lithia-
tion) operation, the complex will dissociate back to its
components as the electrode polarization becomes large
enough to compensate the complex stabilizing energy.

Fig. 5. Typical EIS spectra before and after 2 days at 70 °C storage.

Fig. 7. Evolution of the electrolyte, film, and charge transfer resis-

tance after 1 – 8 days 70 °C storage.

Fig. 6. Half cells equivalent circuit.

3.3. Re6ersible and irre6ersible capacity losses: the

metastable adsorbed complex model

Assuming the electronic conductivity of the elec-

trolyte negligible compared with the ionic one, we
exclude the internal electron leakage as the source of
the reversible capacity loss. Instead we suggest the
formation of a metastable electron-ion– solvent com-
plex that adsorbs on the graphite layer edges. Interca-
lated lithium ions tend to diffuse to the graphene edges
driven by the chemical potential created by adsorbed
PF−6 and solvent molecules. It may then enter in an Fig. 8. Evolution of the film and double layer capacitance after 1 –8
interaction with the electrolyte without full electron days 70 °C storage.
1222 R. Yazami, Y.F. Reynier / Electrochimica Acta 47 (2002) 1217–1223

Fig. 9. XRD patterns of delithiated graphite electrode before and after aging for 8 days at 70 °C.

Fig. 10. Schematic representation of the mechanism of the electron-ion – solvent complex formation.

Fig. 11. Reversible capacity loss; the electron-ion – electrolyte complex dissociates to its elements.

Adsorbed lithium is then re-intercalated and might be by an arrow in Fig. 1b. Fig. 11 schematically shows the
used during the following de-lithiation as a part of the mechanism of the reversible lithium recovery from the
new electrode reversible capacity. Such mechanism complex.
should be at the origin of the unusual change in the On the other hand, the complex is unstable since
curvature at the end of the first discharge curve showed under thermal activation, a part of the adsorbed elec-
 R. Yazami, Y.F. Reynier / Electrochimica Acta 47 (2002) 1217–1223 1223

Fig. 12. Mechanism of irreversible capacity loss, the electron-ion – electrolyte complex recombines and forms additional SEI.

tron would transfer to the electrolyte (reduction of PF− 6 , ity loss that is found of two types, reversible and
EC, DMC, impurities…), which would result in LiF, irreversible. The differences between these losses lies in
Li2CO3, and other lithium alkyl-carbonates or polymer the extend of reaction with the electrolyte i.e. surface
formation [4,5]. Such chemical reduction is irreversible adsorption or full electron transfer, respectively. To deal
in nature and would account for the irreversible capacity with the mechanisms of reversible and irreversible capac-
loss. Fig. 12 illustrates the irreversible capacity genera- ity losses, we have proposed a model involving the
tion through full electron transfer to the electrolyte and formation of a metastable electron-ion–electrolyte com-
the latter subsequent decomposition. plex. This was supported by combined results of the
The SEI then tends to increase in thickness, which XRD and EIS.
leads to a progressive insulation of the active electrode When the electron in the complex is fully transferred
grains from each other or from the current collector. The to the electrolyte molecule(s), the SEI increases in resis-
result is an apparent loss of electrode capacity. Apparent tance as a result of increasing the thickness of the
in the sense that it does not come from the electrode insoluble and lower conductive inorganic layer. The
crystal structure modification but from its inactivation organic and polymeric parts of the SEI may dissolve in
due to ion and/or electron insulation. As consequence of the electrolyte, which then decreases in conductivity.
active surface reduction, the equivalent circuit resistances Combined effects i.e. isolation of active material and
including the ohmic one from the electrolyte tend to increase in the cell internal resistance, lead to the battery
increase and the capacitance tend to decrease. end of life. We are presently conducting work on the
A similar ‘trapped solvated electron’ model was pro- effect of stacking pressure applied to the cell on the
posed by Rahner [6] to deal with the irreversible capacity electrode isolation. Preliminary results show that apply-
loss in metallic lithium. However, our model diverges ing pressure suppresses or delays the appearance of
from that proposed by the Saft group [1] who assumed irreversible capacity.
a shuttle mechanism of dissolved species to explain the
self-discharge in lithium ion cells. Though their model
could be acceptable in a complete cell, it cannot apply References
in the case of a half-cell, which contains two ‘anode-type’
electrodes; lithiated graphite and metallic lithium having [1] M. Broussely, S. Herreyre, P. Biensan, P. Kasztejna, R.S.
very close potentials. Staniewicz, Proceedings of the 10 IMLB, Como, Italy, 28 May –3
June, 2000, J. Power Sources 97 – 98 (2001) 13.
[2] R. Yazami, in: C. Julien, Z. Stoynov (Eds.), Materials for
Lithium-Ion Batteries, Kluwer Academic Publishers, Dordrecht,
4. Conclusions The Netherlands, 2000, pp. 105 – 159.
[3] A. Martinent, B. Legorrec, C. Montella, R. Yazami, Proceedings
During thermal storage of lithiated graphite electrode, of the 10 IMLB, Como, Italy, 28 May – 3 June, 2000, J. Power
intercalated lithium will diffuse from the van der Waals Sources 97 – 98 (2001) 83.
[4] S. Geniès, R. Yazami, J. Garden, Synth. Met. 93 (1998) 77.
spacing between the graphene layers towards their exter- [5] D. Aurbach, B. Markousky, A. Shechter, Y. Ein-Eli, J. Cohen, J.
nal edges, driven by its reactivity with the electrolyte Electrochem. Soc. 143 (1996) 3809.
and/or impurities. This accounts for the electrode capac- [6] D. Rahner, J. Power Sources 82 (1999) 358.