Carbon 37 (1999) 165–180
Carbon materials for lithium-ion rechargeable batteries S. Flandrois a , *, B. Simon b a
Centre de Recherche Paul Pascal, Av. Albert-Schweitzer, 33600 Pessac, France b SAFT /Alcatel-Alsthom-Recherche, Route de Nozay, 91460 Marcoussis, France Received 3 June 1998; accepted 5 October 1998
Abstract The recent development of lithium rechargeable batteries results from the use of carbon materials as lithium reservoir at the negative electrode. Reversible intercalation, or insertion, of lithium into the carbon host lattice avoids the problem of lithium dendrite formation and provides large improvement in terms of cycleability and safety. This paper reviews the main achievements on performance and understanding of charge–discharge mechanisms, resulting from the tremendous activity devoted to these systems in the past few years. As a matter of fact, all carbon materials can be lithiated to a certain extent. However, the amount of lithium reversibly incorporated in the carbon lattice (the reversible capacity), the faradaic losses during the first charge–discharge cycle (the irreversible capacity), the profile of the voltage curves during charging and discharging, all depend on the structure, the texture and heteroatom content of the carbon material. In this paper, we successively examine the electrochemical behaviour of the main families of materials, namely, natural and synthetic graphites, graphitizable carbons, low-temperature and non-graphitizing carbons, and doped carbons. 1999 Elsevier Science Ltd. All rights reserved. Keywords: A. Electrodes; Intercalation; D. electrochemical properties
1. Introduction As the most powerful reducing element, lithium metal associated with strong oxydants (V2 O 5 , MnO 2 , LiNiO 2 , LiCoO 2 ,) leads to high voltage and high energy batteries that gained a deep interest from applications requiring higher and higher energy density for power sources. However, the well-known problem of dendritic shape of metallic lithium deposited during charge and the associated problems of safety and cycle life seemed to restrict it to primary cell applications. Hope arose again when Sony announced the commercialization  of lithium ion rechargeable batteries, where metallic lithium is replaced by a carbon host structure that can reversibly absorb and release lithium ions at low electrochemical potentials. These batteries actually present only a small decrease of energy density compared with parent Li metal batteries, along with major improvement of cycle life and safety. The movement was soon followed by all major battery makers, mainly japanese, that had in fact *Corresponding author. e-mail: [email protected]
explored the subject earlier . It also generated a large interest in the scientific community owing to the numerous problems, fundamental as well as practical, that this system has to overcome for proper operation. The smart solution of electrode morphology stabilization by using a host structure makes these batteries belong to the so called ‘rocking-chair’ battery family, where ions spontaneously exchange (discharge) from an intercalation structure to another one more oxydant, the reaction being forced in the reverse direction on charge. In the case of carbon-based lithium ion batteries, lithiated carbon is a powerful reducing agent (negative electrode) whereas a metal oxide constitutes the oxydant positive electrode. As the battery is assembled with profit in the discharged state where the active materials present low reactivity to the environment, it is the positive material that has to be in a lithiated state (LiCoO 2 , LiNiO 2 , LiMn 2 O 4 .) This concept was proposed as early as 1980 by M. Armand . At that time, carbons were known to be able to intercalate lithium by a chemical route [4,5], the maximum amount being achieved with graphite (LiC 6 ) but early attempts to use graphite as a negative host structure
0008-6223 / 99 / $ – see front matter 1999 Elsevier Science Ltd. All rights reserved. PII: S0008-6223( 98 )00290-5
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
for lithium ion failed because of strong reactivity with the electrolyte [6,7]. Only intercalation via solid polymer electrolyte at high temperature was reported as successful . Other less crystalline materials were found to be less sensitive to the electrolyte, but presented still modest performances [9,10]. After the claim from Sony, the large variability of low crystallinity carbon materials revealed to result in very different electrochemical behaviours owing to their differences in micro- or macrostructure. Crystalline materials then regained interest when it was demonstrated that they could be used in electrolytes with suitable solvent composition [11–13] mainly based on ethylene carbonate (EC). This papers reviews the main achievements towards the comprehension and the improvement of carbon performances for lithium ion reversible absorption.
optimum ionic conductivity in a wide temperature range . The cells were cycled galvanostatically at 20 mA / g (carbon) current density in 0–2 V potential range. The electromotive forces (E.M.F). of lithiated carbons were obtained after short-circuiting the cell for 48 h and removing lithium from the structure at 5 mA / g, with 2 h rest every 5 mA h / g. All capacities are expressed per gram of studied material. Impedance spectroscopy measurements with a Solartron 1250 equipment were made in the same coin cell with an additional reference electrode. The reference electrode consisted of a thin strip of insulated stainless steel foil plated with lithium, incorporated under the polypropylene gasket. The double layer capacitance, which can be considered as representative of the electrochemically active surface area for similar materials is obtained prior to cycling at low frequency where the carbon behaviour approached a purely capacitive response.
2. Experimental tests for carbon performance The charge–discharge performance of a selected carbon is generally evaluated with the help of small lithium / carbon cells. Thus, carbon is the cathode of the cells and Li intercalation is the discharge process, whereas it is the charge process in carbon / oxide cells (a confusion often found in the literature). A lithium foil is used as the anode and the carbon electrode is made up of a mixture of carbon powder, polymeric binder and often, carbon black (a few percent in weight). The role of carbon black is to increase the conductivity and the wettability of the electrode by the electrolyte (a lithium salt in a non-aqueous solvent). We now give examples of testing cells used in our laboratories. Test electrodes were fabricated on a weight basis of 90% carbon, 5% teflon and 5% acetylene black (YS). They were obtained by mixing carbon powders, teflon aqueous dispersion (Algoflon D60VB) and ethanol. The paste was malaxed and rolled to a foil. Electrodes (1 cm 2 ) were cut from this foil, dried, applied on a nickel grid (pressure equal to 2 t / cm 2 ) and dried at 1208C under vacuum before cell assembly. Electrodes as used in actual batteries were made with Polyvinylidenedifluoride PVDF (Solvay) in solution in N-Methyl-Pyrrolidone (NMP) by mixing with carbon powder and coating the slurry on a copper foil. The PVDF content of the dried electrode varied from 10 to 15%. The test cells consisted in 2025 coin cells with metallic lithium as counter electrode, and microporous polypropylene sheet (Celgard) as separator. Electrochemical experiments were performed with a MacPile potentiostat galvanostat. The electrolytes were based on organic carbonates (PC propylene carbonate, EC ethylene carbonate, DMC dimethylcarbonate) and lithium salts such as lithium hexafluorophosphate (LiPF 6 ) or lithium trifluorosulfonimide (LiTFSI). Examples of compositions are EC / DMC LiPF 6 1 M or LiTFSI 1 M and PC / EC / 3DMC LiPF 6 1 M, the latter being selected by Saft for its
3. Main characteristics of Li / C cells
3.1. Energy The replacement of metallic lithium (3600 mA h / g theoretical capacity) by a lithiated carbon host structure results in a decrease in energy density because of the presence of a matrix and the lower packing density of lithium. The basic demand for carbon is thus to allow high capacity for reversible lithium ion absorption at a potential close to that of lithium metal. This property appears to be mainly dependent on the carbon crystalline structure (Fig. 1). Owing to the importance of this subject for this review, it will be presented in more detail in the next sections with separate treatment of the main carbon families.
Fig. 1. Cell voltage as a function of capacity during Li removal for four carbon samples with different crystallinity, previously saturated in Li by short-circuiting the cell, as described in Section 2.
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
3.2. Kinetics Fast charge and discharge are also required for modern batteries. Lithium doping and undoping in carbon implies solid state reactions, with lower kinetics compared to a first kind electrode like metallic Li. Except for very thin electrodes of microbatteries, the use of carbon powders is mandatory to increase kinetics of lithium exchange to an acceptable level. The relevant parameters to ensure high kinetics thus are the exchange surface area and the lithium diffusivity inside the particles. However, the potentialities of the carbon materials are not always achieved at the rate used in actual batteries , as classical limitations of porous electrodes can occur. Performances of industrial electrodes at high current density depend on multiple parameters such as electrode thickness, binder nature and active material coverage, ionic conductivity in the porosity, or particle orientation of graphite flakes [16–18].
Fig. 2. First discharge (1) and charge (2) curves of a Li / natural graphite cell. The plateau at 0.8–0.9 V on the discharge curve is characteristic of passivating layer formation. The other plateaus for voltage below 0.3 V are signatures of the existence of defined stages (see Section 4).
3.3. Reversibility All types of carbon experience irreversible side reactions concentrated on the first electrochemical absorption of lithium ions. As the battery capacity, i.e. the quantity of lithium ions exchanged between the two electrodes, is initially in the positive electrode, this phenomenon leads to a definite loss of battery capacity and has to be minimized. It is now widely accepted that electrolyte instability participates largely in this irreversible process [11,17,19,20]. No electrolyte has been found that withstands the low electrochemical potential of metallic lithium or highly lithiated carbons. In the best cases, electrolytes eventually are reduced as the negative electrode potential lowers until the formation of an electronically insulating (passivating) layer that allows further lithium exchange. This process is clearly visible on the galvanostatic curves as an irreversible plateau at a potential close to 0.8 V versus Li 1 / Li, slightly dependent on the rate and the temperature (Fig. 2). As the properties of this layer are fundamental for the functioning of the negative electrode, it has been the subject of extensive work using various characterization techniques. Direct evidence of a layer presence has been obtained with TEM [21,22] and more recently, the building of the passivating layer at the carbon outer surface has been visualized by powerful advanced techniques such as STM or AFM [23–26]. Quartz crystal microbalance was also used to establish the presence of the layer  and its stability . Indeed, solubility of some layer components in the electrolyte has been proposed as a major mechanism for negative electrode self discharge [29,30]. Impedance spectroscopy also appears as a promising tool for exploring the electrical characteristics of the passivating layer [31,32]. The specific response of the passivating layer can clearly be visualized (Fig. 3) on thin
electrodes (30 mm thick). This configuration minimizes the problem of alternative current distribution in thick porous electrodes that generally leads to depleted arcs and hardly deconvoluted responses [33,34]. The electrical response of the passivating layer can be roughly modeled by a simple RC circuit in the higher frequency range, that remains apparent at high potential. By contrast, the charge transfer response at middle range frequency disappears when no lithium remains in the graphite. The electrical parameters associated to the first RC circuit are stable upon cycling (Fig. 4), showing great difference with metallic lithium behaviour . A major contribution to the knowledge of the layer composition and electrolyte reduction mechanism came
Fig. 3. Impedance spectra of a thin (30 mm) electrode of natural graphite in PC / EC / 3DMC1Li PF 6 1 M electrolyte: (h) in the fully lithiated state (down to 50 mV versus Li 1 / Li), (s) in the delithiated state (up to 2 V versus Li 1 / Li).
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
Fig. 4. Evolution of the electrical characteristics, (d) resistance R p and (h) capacitance Cp of the passivating layer formed on a thin electrode of natural graphite with the cycle number.
from the work of Aurbach and coworkers [36,37], who took advantage of their previous work on the passivating layer formed on metallic lithium . Mainly based on FTIR technique, their results show that cyclic carbonates like propylene carbonate or ethylene carbonate seem to be the best suited solvents, as they yield lithium alkyl carbonates and lithium carbonate with a good passivating ability. The anion of the lithium salt also participates in the reduction process. These results were later confirmed by other techniques like XPS [28,39]. Quantitative analysis was also proposed , that shows clearly the influence of the lithium salt. However, there still exists some controversy on the exact mechanisms of reaction, as LiOH was found as an underlying layer  by XPS, or CO evolution reported . Anyway, the formation of a passivating layer is general to all carbons working at low potential, and as this reaction consumes electrons, an irreversible capacity is unavoidably associated to this phenomenon. A direct way to minimize the associated irreversible capacity would be to decrease the exchange surface area. A relation between surface area and amount of faradaic losses has been indeed reported [11,42–44], when no additional source of irreversibility occurs. This is the case, for example, for petroleum cokes or graphites in a suitable electrolyte, as shown in Fig. 5, where the double layer capacitance is proportional to the electrochemically active surface area (the non-zero intercept with y axis is due to other losses non-related to the carbon surface, such as teflon reduction and carbon black contribution, supposed to be constant for every sample). The electrochemically active surface area of the samples should always be taken into account and passivation contribution subtracted from the overall capacity losses in the search for other possible mechanisms of irreversibility. As we will show in the next sections, other sources of irreversible losses at the first cycle are the presence of
Fig. 5. Irreversible losses of teflon bonded electrodes for (s) natural graphites, (h) synthetic graphites and (앳) petroleum cokes as a function of the double layer capacitance, Cdl , determined from impedance spectroscopy measurements. Electrolyte: LiTFSI 1 M in EC / DMC.
hetero-atoms or structural defects and the existence of exfoliation. Obviously, they depend on the structure, texture and composition of the carbon. The polymeric binder, necessary to provide sufficient mechanical properties to the electrode, can also be subjected to electrochemical reaction at low potential. Perhalogenated polymers are known to be reduced by Li amalgams . When teflon is used as a binder, the irreversible capacity increases with the content [46,47]. By contrast, when PVDF is used as the binder, no significant variation is observed (Fig. 6), in accordance with the non-reactivity reported in Ref. . Binders, especially filmogen ones, can undirectly affect the reversibility of the negative electrode by protecting it from solvent cointercalation .
Fig. 6. Effect of the binder on the irreversible capacity of a graphite electrode: the faradaic losses are proportional to the teflon content (h), whereas they are independent of the PVDF content (s).
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
4. Natural and synthetic graphites
4.1. Passivating layer formation and exfoliation As seen in the previous section, during the first discharge of a Li-carbon cell (carbon reduction), a passivating layer is formed on the surface of carbon particles, at voltages of about 0.8–0.9 V versus Li 1 / Li. This layer is electronically insulating, permeable to Li 1 ions but impermeable to other electrolyte components, and is mainly composed of lithium carbonate and lithium alkyl carbonates [11,21,22,49], as the electrolyte solvents generally used are alkyl carbonates. The process is irreversible and contributes to the so-called irreversible capacity. The use of graphite, a crystalline form of carbon, causes additional complications, especially with propylene carbonate based electrolytes [6,7]. In this case, solvated lithium ions are intercalated in the same voltage range (|1 V versus Li 1 / Li), giving rise to ternary graphite intercalation compounds . When the Li content is low (critical C / Li ratio typically .18), the ternary compounds Li(solv) y C n are thermodynamically favoured over the binary compounds LiC n . Further reduction causes the decomposition of the solvent molecules between the graphene layers, at voltages generally of the order of 0.5 V versus Li 1 / Li (Fig. 7). In the case of propylene carbonate (PC), for example, the reduction process leads to propylene gas formation which exfoliates the graphite matrix, resulting in a dramatic increase in the irreversible capacity due to the increase in specific surface area and to the loss of electrical contacts. As the amount of irreversible capacity must be minimized in practical cells, there was a search for electrolytes which prevent the formation of ternary graphite compounds. It has been shown [11,12] that ethylene carbonate
Fig. 7. Examples of high irreversible capacity observed for Madagascar natural graphite and graphitized mesocarbon microbeads. Electrolyte: LiPF 6 1 M in PC / EC / 3DMC. The plateau at about 0.5 V on the first discharge curve is characteristic of cointercalated solvent reduction.
(EC) based electrolytes generally reduce the irreversible capacity. As EC is a solid at room temperature, it is mixed with other alkyl carbonates such as diethyl carbonate (DEC) or dimethyl carbonate (DMC). However the beneficial effect of EC against exfoliation is not observed for all graphites, as shown in Fig. 8 which gives the first cycle for two different graphite samples with similar surface area. The electrolyte was 1 M LiTFSI in a 1:1 mixture of EC and DMC. One of the graphite samples exhibits an ‘exfoliation’ plateau at about 0.5 V and a large concomitant irreversible capacity. We will come back to these results in the next section. Addition of crown-ethers  or of inorganic additives, such as CO 2 , N 2 O, [52,53] may also hinder the intercalation of solvated lithium ions to some extent.
4.2. Hexagonal versus rhombohedral graphite: the effect of crystal structure on the electrochemical intercalation of lithium ions Graphites with high crystallinity (e.g. interlayer distance inferior to 0.336 nm) are synthesized by many producers in the world or can be obtained from different natural ores (e.g. from Madagascar, Germany, China, Sri Lanka, Brazil,....). The most common structure is the hexagonal Bernal structure , where the carbon layers are arranged in the....ABAB..... sequence with a shift of the B layers with respect to the A layers. It has been found [55a,b] that another crystalline form can exist with a rhombohedral structure. The stacking sequence is...ABCABC..., where the C layers are shifted with respect to the B layers. This form is never pure. It is always mixed with the hexagonal form in variable amount which can be increased up to 30–40% of rhombohedral content by mechanical grinding  or ultra-sonication . On heat treatment at temperatures above 20008C, the rhombohedral structure disappears
Fig. 8. The first cycle for two graphite samples of similar particle size. Electrolyte: LiTFSI 1 M in EC / DMC 1:1.
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
progressively, which shows that the hexagonal phase is thermodynamically more stable. Although graphites with different rhombohedral phase contents seem to be indistinguishable in terms of reversible lithium capacity , it is not the same for the irreversible capacity at the first cycle, as shown by B. Simon and coworkers [59,60]. These authors investigated the electrochemical performance of natural and synthetic graphite powder samples from different origins. The rhombohedral phase content was estimated by X-ray diffraction from the intensities of rhombohedral and hexagonal (101) reflections . Electrochemical experiments were performed on 2025 coin cells using 1 M LiPF 6 dissolved in EC / DMC or PC / EC / 3DMC mixtures as electrolytes. The results are given in Fig. 9. As the irreversible capacity due to passivation only is proportional to the active specific area determined from the double layer capacitance (see Section 3 of this paper), the losses at the first cycle per unit of double layer capacitance are plotted in this figure as a function of the rhombohedral phase content. If only passivation occurred, the result should be constant, with a value close to 90–100 mA h / F / g. Graphites with low rhombohedral content deviate strongly from this base line, particularly in the PC containing electrolyte. On the other hand, when the rhombohedral content is higher than 30%, graphite exfoliation does not occur, even with a high PC content, as shown in Fig. 10 for a natural graphite sample of 40% rhombohedral content. The physical mechanism responsible for this behaviour remains to be determined, and the observed correlation can possibly be an indirect one. As exfoliation implies penetration of large species far into the interlayer spaces, the intercalation of solvated ions is favoured only if the solvation energy compensates for the work needed for layer opening. In non-crystalline carbons such as low-HTT cokes, this work is high due to the small crystallite sizes and to the presence of interstitial carbons atoms (see
Fig. 9. The faradaic losses per unit of double layer capacitance for graphite samples of various rhombohedral content. Electrolyte solvent composition: (s) EC / DMC, (h) PC / EC / 3DMC.
Fig. 10. The first cycle for a natural graphite with 40% rhombohedral content in PC containing electrolytes. No exfoliation occurred even with 80% PC content (4PC / EC).
Section 5), thus preventing exfoliation. It is likely that layer opening in graphites with high rhombohedral phase content is harder than for pure hexagonal graphite because of the presence of phase boundaries and dislocations.
4.3. Reversible capacity and phase diagram One of the most important characteristics of graphite intercalation compounds (GICs) is the staging phenomenon, which corresponds to a periodical arrangement of intercalated layers within the graphite layer matrix. GICs are thus classified by a stage index, n, denoting the number of graphite layers which separate two successive intercalated layers. The staging phenomenon and the resulting phase diagram have been thoroughly investigated in the eighties , using X-ray and neutron diffraction on Ligraphite compounds which were chemically intercalated, mainly from the vapour phase. In addition to the richest compound of stage-1 with LiC 6 composition, the following phases have been clearly recognized: stage-2 (LiC 12 ), a dilute lattice-gas disordered stage-2 (Li–C 18 ), stage-3 (LiC 24 ) and stage-4 (whose composition was not well defined). In the conditions where unsolvated Li ions are electrochemically intercalated into graphite and once the passivating layer has been formed, reversible intercalation of lithium takes place. The theoretical reversible capacity of 372 mA h / g of graphite is generally approached if the charge–discharge regime is slow enough (typically lower than C / 20 rate where C is the capacity of the cell). The voltage curves exhibit several reversible plateaus in the voltage range 0–0.25 V versus Li 1 / Li, as shown in Fig. 11 for a Li-natural graphite cell. This curve was obtained in the following way. Graphite was first saturated in lithium by short-circuiting the cell for 48 h. Then, lithium was sequentially de-intercalated at 5 mA / g, with 2 h rest every 5 mA h / g. The voltage values reported on the
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
of the cell voltage and a smooth shift of the (001) diffraction peaks from the stage-3 positions to the stage-4 positions. This is probably due to the existence of randomly mixed sandwiches of stages-3 and 4 in the compound , but the occurrence of such a continuous transition between the two pure phases is not understood as yet. Finally, for low lithium content, the intercalated Li is randomly distributed throughout the graphite host in a dilute stage-1 phase. The voltage plateau at about 0.2 V generally corresponds to the transition between this phase and the stage-4 phase, although Ohzuku et al.  obtained evidence for the presence of a stage-8 compound.
Fig. 11. Potential plateaus observed for a graphite electrode during Li deintercalation in conditions close to the equilibrium (see text for detail).
5. Graphitizable carbons
curve are those measured at the end of each relaxation period. As it is well known, the existence of voltage plateaus corresponds to transitions between different single phases and the end points of the plateaus give the composition range of biphasic domains. In the case of GICs, the single phases are obviously compounds of defined stage. The electrochemical Li–C phase diagram has been determined using ex-situ or in-situ X-ray diffraction [12,13,63,64] and Raman spectroscopy . All the studies are not in complete agreement, concerning for example the number of plateaus clearly visible or the precise voltage value at each plateau. This could result from the nature of graphite samples and from the experimental conditions (electrolyte composition, current intensity, temperature...). Nevertheless, analysis of the data leads to the following conclusions. Transitions between stage-1 (LiC 6 ) and stage-2 (LiC 12 ), stage-2 and stage-2L (LiC 18 ), stage-2L and stage3, stage-4 and dilute stage-1 occur at about 0.09, 0.12, 0.14 and 0.20 V versus Li 1 / Li, respectively. The composition of stage-3 and stage-4 compounds are not always well defined. They are of the order of C 25 –C 30 Li for stage-3 and C 44 –C 50 for stage-4. The transition between these two stages seems to be continuous, with a progressive variation
Soft carbons are carbon materials whose structure evolves progressively toward the graphite structure when they are heat-treated at high temperature, up to 30008C. They are constituted of more or less misoriented crystallites whose sizes and crystalline order increase with the heat treatment temperature (HTT). At the beginning of the graphitization step (HTT|1200–13008C), the size of the ˚ both parallel and crystallites is of the order of 50 A, perpendicular to the layers (La and Lc , respectively), with ˚ The an average interplane distance d 002 close to 3.44 A. ˚ at HTT|20008C and reaches crystallite size is about 100 A ˚ with higher HTT’s, while d 002 apseveral hundreds A ˚ proaches the graphite value (3.354 A). Typical soft carbons are graphitizable cokes, ex-mesophase carbon fibers, vapor-grown carbon fibers or mesocarbon microbeads (MCMB). All these materials, heat-treated in the temperature domain where graphitization occurs (1300–30008C), exhibit common features for Li electrochemical intercalation and deintercalation [66–76]. Firstly, the faradaic losses at the first cycle (irreversible capacity) correlate well (Fig. 5) with the double layer capacitance (DLC) measured by impedance spectroscopy, whatever the electrolyte composition. As shown in Table 1 for a series of heat-treated
5.1. Li intercalation in heat-treated soft carbons
Table 1 Heat-treated petroleum cokes: electrochemical data and g parameter deduced from X-ray diffraction measurements a HTT (8C)
DLC (mF / g)
Irreversible losses (mA h / g)
Reversible capacity (mA h / g)
1300 1700 2000 2200 2400 2800
1800 1200 600 600 550 400
140 115 60 50 50 110
260 265 182 205 280 310
0 0.13 0.29 0.50 0.75 0.94
g5(3.442d 002 ) /(3.4423.354), see text, Section 5.2).
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
petroleum cokes, the irreversible losses decrease with HTT, except for the HTT-2800 sample. As the DLC decreases also with HTT, one can assume a decrease of active surface area resulting from structural reorganization (elimination of microporosity). As expected, the effect is greatest at the beginning of the graphitization process (from HTT51300 to 20008C). In highly graphitized samples (HTT528008C and above), some exfoliation occurs, as in graphite, producing an increase of irreversible losses. Secondly, the reversible capacity values are lower than for graphite and exhibit a minimum for samples treated at about 20008C (Table 1). Finally, this HTT value delimits two domains with different voltage behaviours. For soft carbons heat-treated above this temperature, the cell voltage exhibits the progressive appearance of plateaus, characteristic of diphasic domains (Fig. 12). On the other hand, for carbons heat-treated at 20008C and below, the voltage varies continuously, as expected for monophasic systems.
5.2. Dependence of the reversible capacities and of the occurrence of voltage plateaus on HTT The results obtained on all heat-treated soft carbons point to the existence of two domains of electrochemical behaviour according to whether the heat-treatment temperature is higher or lower than 20008C. In fact, since the sixties, when the graphitization of soft carbons was thoroughly investigated, it has been known that this value of HTT is a critical value for the behaviour of numerous physical and structural properties. It corresponds to a sharp increase in three-dimensional ordering, which is evidenced by the appearance of modulations on the (hk) bands of the X-ray spectra. The increase of reversible capacities above HTT-2000 can therefore be understood as resulting from increased crystallinity. The capacity increase for low HTT’s is unexpected. As we will see in the next section,
Fig. 12. Voltage profile observed during Li deintercalation in conditions close to equilibrium for petroleum cokes heat-treated at 22008C (XP 2200), 24008C (XP 2400) and 28008C (XP 2800).
high capacities have been observed for carbon materials prepared from organic precursors pyrolyzed at low temperatures (700–10008C: carbonization step). The proposed explanations, related to the presence of heteroatoms or to the filling of internal porosity, are not valid for soft carbons in the graphitization domain: (i) the amount of heteroatoms is quite small for soft carbons treated at 13008C and above; (ii) the internal porosity is negligible (He pycnometry gives density values close to those deduced from X-ray diffraction). Endo and coworkers  showed that the minimum capacity occurs when Lc , the crystallite thickness determined from the width of the (002) X-ray diffraction ˚ For Lc .100 A, ˚ a classical peak, is of the order of 100 A. intercalation process should occur, governed by the Lc value, as shown for example in the case of sulfuric acid ˚ the intercalation . On the other hand, for Lc ,100 A, authors assume the occurrence of a different process of doping and undoping, namely the formation of covalent Li 2 molecules which could act as a Li ion reservoir. In the opinion of the authors, such a process would be enhanced by decreasing the crystallite thickness. Another explanation was proposed by Dahn and coworkers [67,78]. Their model is based on Franklin’s model of graphitization , in which a pregraphitic carbon is seen as a mixture of organized and unorganized regions. Organized regions would be constituted of parallel carbon layers either registered in ABAB stacking as in graphite or turbostratic, i.e. with random shifts or rotations between them. The unorganized regions would consist of groups of tetrahedrally bonded carbon atoms or highly buckled graphitic sheets. Dahn and coworkers were able to determine the relative amounts of registered, turbostratic and unorganized layers in a given pregraphitic carbon from analysis of X-ray data. Then, assuming that the amount of Li which six carbon atoms can accommodate (x in Li x C 6 ) is 1 for registered layers, 0.9 for unorganized layers and 0.25 (HTT,22008C) or 0.00 (HTT.22008C) for turbostratic layers, they could fit the dependence of reversible capacities on HTT. Thus, turbostratic disorder would not play the same role according to whether HTT is higher or lower than 22008C. For HTT.22008C, the interlayer spaces or galleries between randomly stacked layers would not host lithium ions and these ‘blocked’ galleries would frustrate the formation of the regular sequence of full and empty galleries (staged phases) and therefore the appearance of voltage plateaus. An alternative explanation was recently proposed by the present authors , who showed that the dependence of capacities and voltage behaviour on HTT can be under´ stood within the framework of Mering’s model of graphiti´ zation . Mering and his coworkers have done extensive work on pregraphitic carbon X-ray diffraction. Based on these studies and on others, such as bromine intercalation and oxidation kinetics, a model was proposed for the graphitization of soft carbons, which has never been
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
refuted. Contrary to Franklin’s ideas, where graphitization was thought to be simply the transformation of disordered stacks of graphitic atomic layers into an ordered arrange´ ment, in Mering’s model the structure of each layer is disordered, contains defects and the fundamental process of graphitization lies in the transformation of elementary layer states. The rearrangement of the layers then appears only as a consequence of this internal transformation. ´ From a quantitative analysis of X-ray data, Mering assumed that each face of an elementary layer can be in either an a state or a b state. The b state is that of the surface of a perfect graphite, whereas the a state results from the presence of interstitial carbon atoms grafted onto the face of the layer, with a corresponding extra thickness. Graphitization thus consists of a cleaning of the face of each layer, converting it from the a to the b state. As this cleaning must occur independently on each face, the ordered arrangement is possible only between two adjacent bb faces. Thus, in a partially graphitized carbon, three types of elementary interlayer spaces are present: aa, ab and bb (graphite layer). Given g as the fraction of layer faces in the b state, the relative proportion of each type is: (12g)2 for aa, 2g (12g) for ab and g 2 for bb. The g parameter is given by the X-ray diagram. It has been found that, at the end of the carbonization process (HTT|1100– 12008C), where only the presence of aa spaces can be assumed, all soft carbons exhibit a mean interlayer dis˚ whereas it is 3.354 A ˚ in graphite tance d m close to 3.44 A, (bb spaces). Thus, for a partially graphitized carbon, the mean interlayer distance can be expressed as: d m 5 3.354g 2 1 [(3.354 1 3.44) / 2]2g(1 2 g) 1 3.44(1 2 g)2 or, equivalently: g5(3.442d m ) /(3.4423.354). The composition corresponding to the reversible capacities of heat-treated cokes is plotted in Fig. 13 as a function
of g. For each carbon sample, the composition (or the stoichiometric coefficient x in Lix C 6 ) must be the sum of the contributions of aa, ab and bb spaces, with x bb 51 (graphite): x 5 (1 2 g)2 x aa 1 2g(1 2 g)x ab 1 g 2 The solid line in Fig. 13 is the fit with x aa 50.75 and x ab 50.20. The agreement is remarkable. The lower value of x ab can be qualitatively understood from STM observations  on pyrocarbon heat-treated at 20008C. It has been shown that, in the ab state, interstitial atoms form ] ] clusters in which they are regularly arranged in a Œ7 3Œ7 R 1981 lattice, commensurate with the graphite lattice. Thus, the sites that Li normally fills, are already occupied by interstitial carbon atoms. Since Li cannot penetrate these clusters, there is a resulting decrease in x with respect to a random distribution of interstitials. It remains to quantitatively understand the values of x aa and x ab . ´ Mering’s model may also help us understand the appearance of voltage plateaus for HTT higher than 2 20008C. When g50.5, g 50.25 which means that, on average, one interlayer space out of four is of bb type and can be filled completely with lithium. For such a carbon sample, we can expect the formation of a stage-4 compound. This seems to be the situation observed for a coke heat-treated at 22008C: one plateau is visible for a voltage of about 0.18 V, a voltage which could correspond to the transition between stage-4 and dilute stage-1 . In the same way, the plateaus observed on the curve for the HTT-2400 sample lead to the formation of a stage-2. For this sample, g 2 ¯0.5 which means that, on average, one interlayer space out of two is bb.
6. Low-temperature carbons and non-graphitizable carbons
6.1. The materials: general features
Fig. 13. The reversible capacity of heat-treated cokes as a function of g parameter. Solid line is the fit with the expression x5(12g)2 x aa 12g(12g) x ab 1g 2 , where x aa 50.75 and x ab 50.20.
The pyrolytic conversion of organic compounds to carbon residues at temperatures below 10008C occurs by a complex chemistry, which involves a great number of parallel and sequential reactions and depends on the nature of the starting material . Nevertheless, after the primary step of carbonization (up to HTT500–6008C), all carbonaceous materials are made up of similar elemental bricks with different relative arrangements . The elemental unit or basic structural unit (BSU) contains planar aromatic structures consisting of about 10–20 rings, which are piled up more or less in parallel by groups of two to four. In the case of soft carbons, by increasing HTT, the USB’s form distorted columns which will coalesce during the graphitization step. In hard carbons, the presence of cross-linking groups, especially when the precursor
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
contains oxygen, hinders the formation of columns and their coalescence. Whatever the graphitization degree they can reach, carbonaceous materials heat-treated between 500 and 10008C exhibit common features. Firstly, they have a low crystallinity: X-ray diffraction spectra exhibit only a few broad peaks with weak intensity. The most intense peaks appear close to the position of (002), (100) and (110) peaks of graphitized samples. The mean distance calculated from the maximum of (002) peaks is larger than 3.44 ˚ In some cases, distances of more than 4.0 A ˚ have been A. reported; however, such broad peaks, whose full width at half maximum is of a few degrees, have to be corrected from Lorentz-polarisation, carbon diffusion and thermal factors. Second, these carbon materials have a large amount of micro- or nanoporosity observable by smallangle X-ray scattering . The porosity is also reflected in density measurements, which give values between 1.4 and 2.0 g / cm 3 . Third, they all contain heteroatoms left from the organic precursor, essentially hydrogen, nitrogen, oxygen and sulfur, depending on the precursor. The most part of hydrogen is lost at 10008C , oxygen and nitrogen at around 15008C  and sulfur not below 20008C . Finally, it must be noticed that, for both soft and hard carbons, the electrical conductivity increases by several orders of magnitude, when they are heat-treated from 500 to 10008C, and a model of disorder-induced non-metal–metal transition has been proposed to explain this behaviour .
6.2. Electrochemical insertion of lithium Intercalation is defined as the introduction of guest species into the interspace of host layers. Whereas it is justifiable to talk about ‘lithium intercalation’ in the case of soft carbons during the graphitization step, in the case of low-temperature carbons and non-graphitizing carbons, ˚ the Li storage cannot be where the USB size is 10–15 A, restricted to the interspace between aromatic molecules which constitute the USB’s. Clearly, the low crystallinity, the internal porosity, the presence of heteroatoms and of functional groups must have a strong influence on the reversible and irreversible capacities obtained by lithium insertion. The electrochemical behaviour of a large number of carbonaceous materials pyrolyzed at temperatures between 500 and 11008C has been investigated. They include (the list is not exhaustive): pitch cokes [89,90], petroleum cokes , ex-mesophase carbons in the form of spheres (mesocarbon microbeads, MCMB) [71,89], fibers or films , pyrolyzed polymers, such as poly ( p-phenylene) (PPP) [93,94], poly (acrylonitrile) (PAN) [95,96], poly (vinyl chloride) (PVC), poly (vinylidene fluoride) (PVDF) and poly (phenylene sulfide) (PPS) , phenolic resins [97–100], pyrolyzed sugars [100–102], pyrolyzed mixtures of pyrene and dimethyl-p-xylene glycol . They
Fig. 14. The first and second cycles obtained for a hard carbon, showing high reversible and irreversible capacities and large hysteresis in the voltage curves.
generally show similar electrochemical behaviour, an example of which is given in Fig. 14. The main features are: (1) high reversible capacities, which can be more than twice the graphite value, (ii) high irreversible capacities in the first cycle, (iii) large hysteresis in the voltage curves, i.e. Li 1 ions are inserted near 0 V versus Li 1 / Li and removed at about 1 V. Those features are more pronounced when the pyrolysis temperature is lower 1 Several different mechanisms have been proposed to explain this behaviour. As regards the high reversible capacities, Sato et al.  proposed that Li 2 covalent molecules would be intercalated in addition to the regular site occupation leading to LiC 6 . This assumption was based on 7 Li NMR measurements which showed two bands with chemical shifts of 10 and 0 ppm, attributed to regular intercalated Li and Li 2 molecules, respectively. Recent NMR measurements  gave also shown evidence for two different Li sites, but the band close to 0 ppm was interpreted as due to Li ions located at the edges of carbon layers. According to Xiang et al. , the role of the edges of carbon layers is essential, as they observed a linear relationship between the capacity of the plateau at about 1 V during charging and the total edge length per unit mass deduced from La and Lc measurements. The edge effect on the high capacity of carbonaceous materials heat-treated below 10008C could result from the presence of hydrogen atoms bonded at the periphery of the aromatic molecules which constitute the USB’s. Indeed, the capacity of various polymers pyrolysed below 10008C was shown to be correlated to the H / C atomic ratio [91,94,98]. This, in turn, could explain the large
It must be mentioned that some hard carbons heat-treated between 1000 and 12008C may exhibit voltage curves with significant capacity below 0.1 V. The existence of this low-voltage plateau is not well understood .
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
hysteresis in voltage observed between lithium insertion and removal. Zheng et al.  proposed that the lithium atoms may bind on hydrogen-terminated edges of hexagonal carbon fragments, causing a change in the carbon– carbon bond from sp 2 to sp 3 (the local geometry would be analogous to those of organolithium molecules). A simple model, which treats the bonding change in the host during lithium insertion and removal as an activated process, was shown to fit many aspects of the data qualitatively. On the other hand, some hard carbons can maintain capacities higher than 400 mA h / g, even though their hydrogen content has been considerably reduced by heat-treatment . In addition, they develop a pronounced low voltage plateau and little hysteresis . From X-ray diffraction and small angle X-ray scattering experiments , it was proposed that lithium can be adsorbed onto internal surfaces of nanopores formed by single-layer, bi-layer and tri-layer groups of graphene sheets which are arranged like a ‘house of cards’. This adsorption on the surface of nanopore walls is similar to the pore filling proposed by Sonobe et al.  and to the cavity filling proposed by Tokumitsu et al. . It also bears some resemblance with the multilayer model proposed by Yazami and Deschamps : each (a,b) face of an USB would be covered by two or three lithium layers. The origin of the high irreversible capacities observed during the first cycle has been studied by several groups. Part of the irreversible losses is due, as for graphitic carbons, to the formation of a passivating layer on the carbon surface. The losses will increase if the carbon particles have micropores which can be penetrated by the passivating layer. Using various analytical techniques such as Fourier transform infrared attenuated total reflectance (FTIR-ATR), secondary ion mass spectroscopy (SIMS) and X-ray photoelectron spectroscopy, Matsumura et al.  showed that a large portion of the irreversible capacity could come from the reaction of lithium with active sites in the bulk of the carbon electrode. These active sites can be hydroxylic groups or carbon radicals. It is well known that disordered carbons have many active sites associated with high concentrations of unpaired electron spin centers which give rise to surface complexes, mainly with oxygen but also with other elements . Kikuchi et al.  showed that the irreversible current peak observed in the cyclic voltammograms disappeared after their pitch-based carbon fiber electrodes were heated at 9808C under vacuum. The effect of different gas exposures on the irreversible capacities of sugar carbons was recently studied by Xing and Dahn . They showed that carbon samples exposed only to Ar or N 2 exhibit strongly reduced irreversible capacities compared to those observed after air exposure for several days. Exposures to CO 2 , O 2 , water and steam led to increases in irreversible capacity. In the case of water steam exposure, the voltage curve during the first discharge showed an irreversible plateau near 2V, that the authors assumed to
result from the reaction of Li with water adsorbed in the micropores of the sugar-based carbon: 2 1 Li 1 1 e 2 1 H 2 O → LiOH 1 ] H 2 . 2 From the standard chemical potentials of the species involved in this reaction, we have calculated the standard potential versus Li 1 / Li as equal to 2.1 V at 258C, in excellent agreement with the observed plateau voltage.
7. Doped carbons
7.1. Definition and methods of doping Carbons, and especially low-temperature carbons, are not pure carbon. As we have seen previously, they usually contain foreign atoms from a few ppm to a few per cent. But it is possible to introduce foreign atoms into the carbon lattice intentionally, which is the meaning of doping. Doping has been used extensively in the past , mainly to change the distribution of electrons between energy levels in the carbon, but also to affect the graphitization process (positive or negative ‘catalysis’) and to modify the chemical state of the surface of the carbon particles. The doping methods include: (i) co-deposition by CVD of carbon and foreign atoms (ex: boron, nitrogen), (ii) pyrolysis of organic molecules containing foreign atoms (ex: nitrogen, oxygen, silicon), (iii) chemical treatment of the carbon (ex: boron, phosphorus, oxygen, halogens). One of the main problems in characterizing the products is the actual location of the dopant atoms in the carbon network. They can be either substituted to carbon in aromatic rings, or located in interstitial sites, or, most often, chemically bound to the macromolecules.
7.2. Effect of doping on lithium electrochemical insertion 7.2.1. Doping with boron Boron is one of the few elements which are known with some certainty to enter substitutionally into the carbon lattice. By chemical reaction at, or near, thermodynamic equilibrium (for example, reaction of B 4 C and graphite at 23508C), it has been shown  that the thermodynamic solubility of boron in the carbon lattice cannot exceed 2.35 at. %. However, using chemical reactions which are not at equilibrium (for example, CVD from benzene and boron trichloride), carbons containing up to 25 at. % boron can be prepared , although it is not sure that all boron
Note that in Ref.  this equation is erroneously written with two electrons.
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
atoms are in substitution for the highest concentrations . Using the CVD method, Way and Dahn  prepared boron-substituted carbons, B z C 12z , with concentrations 0,z,0.17, from benzene and boron trichloride at 9008C, and measured the voltage and reversible capacity of Li / Li x (B z C 12z ) 6 cells. All cells for z.0 showed greater reversible capacities than for z50, with the increase in capacity, being roughly proportional to z. For z50.17, the reversible capacity corresponded to x51.16 (437 mA h / g) whereas for z50, x50.65. In addition, the cells with boronated carbons showed an increase in voltage of about 0.5 V for a given lithium concentration compared to that in the cell with z50. This behaviour can be understood qualitatively by taking into account the acceptor character of boron which has three valence electrons, one less than carbon. In a rigid-band model, the Fermi level is lowered, which allows more lithium (an electron donor) to be intercalated. In addition, the presence of boron strengthens the chemical bond between the intercalated Li and the boron–carbon host compared to the pure carbon host. As a result, the potential of the lithium in boron-substituted carbons is increased relative to unsubstituted carbon.
7.2.2. Doping with nitrogen Nitrogen is another element which can be incorporated substitutionally for carbon, either alone or together with boron . C x N compounds, with x ranging from 7.3 to 62, and BC x N compounds, with x52, 3, 7 and 10 have been prepared by CVD from C 2 H 2 and ammonia or by pyrolysis of nitrogen-containing organic molecules such as acetonitrile, acrylonitrile or pyridine, at temperatures between 400 and 10008C [119–124]. The electrochemical behaviour of these materials as electrodes of lithium cells appears to be somewhat contradictory. BCx N compounds have been found to exhibit reversible capacities of about 100 mA h / g only, and, in addition, these capacities decreased gradually with the increase in cycle number [119–121]. Weydanz et al.  found reversible capacity values of 250–300 mA h / g for their C x N samples. However, the irreversible capacity increased with the nitrogen content. A second effect of nitrogen was a shift of the cell capacity to lower voltages compared to pure carbon electrodes. It was proposed that, in these C x N materials, part of the nitrogen is substitutional for carbon (‘lattice nitrogen’), the rest being chemically bound to aromatic molecules (‘chemical nitrogen’). The irreversible capacity would be due to the reaction of lithium with the latter form of nitrogen. The lowering in voltage was explained from arguments similar to those used in the case of boron: in this case, nitrogen acts as a donor and its presence weakens the lithium-host bond compared with a pure carbon. Finally, the authors concluded that such nitrogen-containing carbons are not useful as anodes for Li-ion cells. A quite opposite conclusion was given recently by two Japanese groups [123,124]. C 7.3 N samples, synthesized by CVD from pyridine at 8008C, showed reversible capacities
values of more than 500 mA h / g, although 75% of the lithium was extracted at potentials higher than 0.5 V and, in addition, the irreversible capacity was high (218 mA h / g) . Similar results were obtained by the same authors with C 28 S synthesized by CVD from thiophene at 8008C. Nakajima et al.  synthesized carbon–nitrogen compounds, C 14 N to C 62 N, by thermal decomposition of acetonitrile at 800–11008C in the presence of nickel catalyst. The samples were obtained as filaments at 8008C and as particles at higher deposition temperatures. With an increase in the deposition temperature, the cycleability of electrochemical intercalation–deintercalation of lithium ions was improved (with reversible capacities of 400 mA h / g) and the profile of the charge–discharge curve approached that of natural graphite powder. The use of the nickel catalyst would enable to increase the nitrogen atoms incorporated substitutionally in the carbon network and to decrease the amount of pyridine-type nitrogen atoms existing at the edge of graphene layers .
7.2.3. Effect of oxidation Mild oxidation of carbon materials in air, oxygen, or other oxidizing substances has two main effects: opening of pores and formation of surface complexes. Obviously, the extent of these reactions, as concerns the pore shape and dimension, and the amount and nature of functional groups, depends on both the oxidation conditions and the surface characteristics of the starting carbon material. In the case of graphite, mild oxidation has been shown to improve performance in Li / Li x C cells. Peled et al.  found an increase in reversible capacity by 10–30% for two synthetic graphite powder samples partially burnt in air at about 6008C. Moreover, if only a few per cent (up to about six) of the graphite was burnt off, the irreversible capacity decreased by about 10–20%. The authors attributed the performance improvement to chemical bonding of the passivating layer to the surface carboxylic groups, and to accommodation of extra lithium at edge sites and inside nanochannels formed by oxidation, which were shown by STM to have an opening of a few nanometers. Ein-Eli and Koch  found similar performance improvement for a synthetic graphite sample chemically oxidized by the strong oxidative agents, ammonium peroxy sulfate and hot, concentrated nitric acid. In contrast to graphite, carbons made by pyrolyzing epoxy-novolak resins at 10008C were shown to have poor electrochemical performance after oxidation in air at temperatures from 300 to 6008C . Initially, these materials had a significant nanoporosity, but, due to the smallness of pore openings, the electrolyte could not penetrate the pores, so excellent behaviour was observed. As the samples were oxidized, the pores did not grow significantly in volume, as measured by SAX, but the size of their openings apparently did, to the point where the electrolyte could penetrate the pores, leading to irreversible electrolyte decomposition reactions during the first electro-
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
chemical reaction of lithium with the carbon, and hence large irreversible capacity, even for burnoffs as small as 5%. In addition, surface oxides, resulting from oxygen chemisorption, contribute to the irreversible capacity and lead to large voltage hysteresis. Related experiments are those of severe mechanical grindings of carbons and their effect on lithium insertion. It has been known for a long time that numerous active centers are produced in graphite by powdering . The structure defects resulting from grinding provide sites which have a considerable reactivity to oxygen and other elements. In addition, while mild grinding does not alter the interlayer spacing in graphite, severe grinding converts pristine graphite into a disordered material similar to some low-temperature carbons, a process which is the reverse of graphitization . Recently, the effect of severe grinding on lithium insertion in graphite and soft carbons has been studied . It was shown that the reversible capacity could be increased up to about 700 mA h / g, however with high irreversible capacities and large hysteresis behaviour (during charge, about 80% of the lithium is recovered for voltages higher than 0.5 V). To interpret the increase in capacity, the authors assumed that lithium atoms were adsorbed on surfaces of single-layer carbons, as in microporous carbons . As the voltage profiles differed greatly from those of microporous carbons, Xing et al.  examined the electrochemical behaviour of milled sugar carbons. The cells demonstrated large reversible capacities of more than 600 mA h / g, but also large hysteresis compared to that of unmilled sugar carbon / Li cells. It was shown that the milling created broken carbon– carbon bonds, which were highly reactive, and the airmilled samples contained substantial oxygen. The authors proposed that the mechanism for quasi-reversible lithium insertion in milled carbons may involve: (i) reactions of Li atoms at the edge of small graphene sheets, (ii) intercalation in cases where stacked layers remain, and (iii) reactions with surface functional groups where they exist.
7.2.4. Doping with other elements ( Si, P) Carbonaceous materials containing silicon or silicon and oxygen have been prepared [133–135] by pyrolysis of epoxy-silane composites, of silicon-containing polymers or by CVD of benzene and of silicon-containing precursors. According to the preparation conditions, the materials can be disordered carbons containing nanodispersed silicon or silicon carbide, or mixtures of disordered carbon phase and a glassy phase. Large reversible capacities up to 770 mA h / g have been obtained. However, the voltage profile generally developed large hysteresis and large irreversible capacities. The effect of phosphorus-doping on the electrochemical performance of petroleum green cokes  and hard carbons from epoxy resin  has been examined. Doping was realized by heating the carbonaceous materials with phosphoric acid. Doping levels up to 10 wt. % P were
obtained, with a small amount of oxygen (about 1 wt. %). It was shown that P-doped carbons exhibited improved reversible capacities and reduced irreversible losses. However the capacity increase occurred at potentials above 0.9 V versus Li 1 / Li. The voltage profile was similar to those reported for hydrogen-containing carbons, which seems to indicate a similar mechanism for Li insertion.
8. Conclusion Among the innumerable applications of carbon materials , the use of carbons as a lithium reservoir in rechargeable batteries is one of the most recent. It is also the most important application of carbon intercalation compounds. A number of studies have been recently devoted to the search for alternatives to the carbon electrode, especially since Idota’s work on oxide materials [139–141]. It has been shown that tin oxides , vanadates [143,144] and transition metal nitrides [145,146] may display larger reversible capacities compared to graphite, up to 900 mA h / g in some cases. However, they also show large irreversible capacities and the average potential versus lithium is markedly increased, which would result in a significant battery voltage drop, of the order of 1 V or even more. Therefore, at the present time, carbon is the material of choice for the negative electrode of lithium-ion batteries. Numerous carbon materials have been examined during the last decade, from crystalline graphites to strongly disordered carbons. Analysis of the published results shows that the best electrochemical performances are obtained at both ends of the structural evolution, for highly crystalline or highly disordered materials. Carbons of intermediate crystallinity (heat-treated in the temperature range 1600–21008C) have lower reversible capacities. Low-temperature carbons and hard carbons, doped or not, may display high reversible capacities, comparable to those of the above-mentioned oxides. However, they suffer from similar problems of voltage (or hysteresis) and irreversible capacities. Thus, at the moment, graphites, especially those enriched in rhombohedral phase, appear to be the best compromise: they have low irreversible capacities (,100 mA h / g) and reasonable reversible capacity values (of the order of 400 mA h / g or more, when using treatments such as mild oxidation); the electrode potential is almost constant during charge and discharge, and close to Li 1 / Li potential (|0.1–0.2 V higher, a value high enough to prevent Li metal deposition); in addition, graphites are low-cost materials. Nevertheless, improvement in performance can still be obtained. Carbon is the most versatile element of the periodic table, as concerns the nature of chemical bonds between carbon atoms themselves or with other elements, and the variety of structures, textures and particle shapes. Even though hundreds of carbon materials have been
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
examined in Li cells in the past few years, their number is limitless. New doping or preparation methods can be imagined. For example, samples of controlled porosity may be obtained from a new strategy that uses inorganic templates to direct the synthesis of carbons from polymeric precursors . New forms of carbon can appear, like fullerenes and nanotubes recently, although we do not believe the latter to be useful in Li cells because of their high surface area. Some coatings or dispersions on carbon particles may prove to be of interest, as for example composite electrodes made up of ultrafine silver particles supported on graphite which were shown to provide a higher volumetric specific capacity and longer cycle life than conventional graphite electrodes [148,149]. There is no doubt that carbon materials still have good prospects as negative electrodes in lithium rechargeable batteries.
Acknowledgements This review paper has been written in the light of the work performed under collaboration between authors laboratories during several years. The authors wish to thank P. Biensan, M. Broussely, A. de Guibert and P. Lerner for numerous fruitful discussions. Support from SAFT and Alcatel-Alsthom-Recherche is gratefully acknowledged.
References  Nagaura T, Tozawa K. Prog Batt Solar Cells 1990;9:209.  US Patent 4, 668, 595, 1987 (Asahi Kasei) and priority data (1985). EU Patent Application 0 357 001, 1989 (Sony) and priority data (1988) EU Patent Application 0 346 088, 1989 (Sharp) and priority data (1988) EU Patent Application 0 328 131, 1989 (Tohiba) and priority data (1988).  Armand M. In: Murphy DW, Broadland J, Steeles BCH, editors. Materials for advanced batteries. New York: Plenum Press, 1980. ´  Herold A. Bull Soc Chim 1955;187:999. ´ ´  Guerard D, Herold A. Carbon 1975;13:337.  Dey AN, Sullivan BP. J Electrochem Soc 1970;117:222.  Arakawa M, Yamaki J. J Electroanal Chem 1987;219:273.  Yazami R, Touzain P. J Power Sources 1985;56:81.  Kanno R, Takeda Y, Ichikawa T, Nakanishi K, Yamamoto O. J Power Sources 1989;26:534.  Mohri M, Yanagisawa N, Tajima Y, Tanaka H, Mitate T, Nakajima S, Yoshida M, Yoshimoto Y, Suzuki T, Wada H. J Power Sources 1989;26:545.  Fong R, von Sacken U, Dahn JR. J Electrochem Soc 1990;137:2009.  Ohzuku T, Iwakoshi Y, Sawai K. J Electrochem Soc 1993;140:2490.  Billaud D, Henry FX, Willmann P. Mater Res Bull 1993;28:477.  Perton F, Baudry S, Porcheron A. US Patent 5, 472, 809, 1995.
 Broussely M, Planchat JP, Rigobert G, Virey D, Sarre G. J Power Sources 1997;68:8.  Tran TD, Feikert JH, Song X, Kinoshita K. J Electrochem Soc 1995;142:3297.  Shu ZX, Mc Millan RS, Murray JJ. J Electrochem Soc 1993;140:922.  Liu Q, Zhang T, Bindra C, Fischer JE, Josefowicz JY. J Power Sources 1997;68:287.  Chusid O, Ein-Eli Y, Aurbach D. Extended Abstracts of 6th International Meeting on Lithium Batteries. Munster, Germany, 1992:16.  Simon B, Boeuve JP, Broussely M. Extended Abstracts of 6th International Meeting on Lithium Batteries. Munster, Germany, 1992:25.  Takei K, Kumai K, Kobayashi Y, Miyashiro H, Iwakori T, Uwai T, Ue H. J Power Sources 1995;54:171.  Naji A, Ghanbaja J, Humbert B, Willmann P, Billaud D. J Power Sources 1996;63:33.  Inaba M, Yoshida H, Ogumi Z, Abe T, Mizutani Y, Asano M. J Electrochem Soc 1995;142:20.  Inaba M, Siroma Z, Funakubi A, Ogumi Z. Langmuir 1996;12:1535.  Chu AC, Josefowicz JY, Farrington GC. J Electrochem Soc 1997;144:4161.  Hirazawa KA, Sato T, Asahina H, Yamaguchi S, Mori S. J Electrochem Soc 1997;144:L81.  Morita M, Ichimura T, Ishikawa M, Matsuda Y. J Power Sources 1997;68:253.  Aurbach D, Zaban A, Ein-Eli Y, Weissman I, Chusid O, Markowsky B, Levi M, Levi E, Schechter A, Granot E. J Power Sources 1997;68:91.  Simon B, Boeuve JP, Broussely M. J Power Sources 1993;43–44:65.  Jean M, Tranchant A, Messina R. J Electrochem Soc 1996;143:391.  Ryu YG, Pyun SI. J Electroanal Chem 1997;433:97.  Besenhard JO, Castella P, Wagner MW. Mater Sci Forum 1992;91–93:647.  Takami N, Satoh A, Hara M, Ohsaki T. J Electrochem Soc 1995;142:371.  Liu P, Wu H. J Power Sources 1995;56:81.  Aurbach D, Markovsky B, Schechter A, Ein-Eli Y. J Electrochem Soc 1996;143:3809.  Chusid O, Ein-Eli Y, Camueli Y, Babai M, Aurbach D. J Power Sources 1993;43–44:47.  Aurbach D, Ein-Eli Y, Chusid O, Carmeli Y, Babai M, Yamin H. J Electrochem Soc 1994;141:603.  Aurbach D, Daroux ML, Faguy PW, Yeager E. J Electrochem Soc 1987;134:1611.  Mori S, Asahina H, Suzuki H, Yonei A, Yokoto K. J Power Sources 1997;68:59.  Kanamura K, Shiraishi S, Tadezawa H, Takehara Z. Chem Mater 1997;9:1797.  Yoshida H, Fukunaga T, Hazama T, Terazaki M, Mizutani M, Yamachi M. J Power Sources 1997;68:311.  Schoderbock P, Boehm HP. Mater Sci Forum 1992;91– 93:683.  Simon B, Boeuve JP, Biensan P. Proceedings GFECI, 1994:87.  Ohta A, Koshina H, Okuno H, Murai H. J Power Sources 1995;54:6.
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180  Kavan L, Dousek F, Micka K. Solid State Ionics 1990;38:109.  Li G, Xue R, Chen L. Solid State Ionics 1996;90:221.  Nakagawa Y, Wang S, Matsumura Y, Yamaguchi C. Synth Metals 1997;85:1343.  Santhanam R, Noel M. J Power Sources 1996;63:1.  Naji A, Ghanbaja J, Willmann P, Billaud D. Carbon 1997;35:845.  Besenhard JO. Carbon 1976;14:111.  Shu ZX, Mc Millan RS, Mullay JJ. J Electrochem Soc 1993;140:L101.  Besenhard JO, Wagner MW, Winter M, Jannakoudakis AD, Jannakoudakis PD, Theodoridou E. J Power Sources 1993;43–44:413.  Besenhard JO, Winter M, Yang J, Biberacher W. J Power Sources 1995;54:228.  Bernal JD. Proc Roy Soc (London) 1924;A106:749. [55a] Lipson H, Stokes AR. Proc Roy Soc (London) 1942;A227:330. [55b] Jagodzinski H. Acta Crystallogr 1949;2:298.  Gasparoux H. Carbon 1967;5:441. ´  Flandrois S, Fevrier-Bouvier A, Biensan P, Simon B. French Patent 94-08291. ¨ MY, Koksbang R. J Electrochem Soc  Shi H, Barker J, Saıdi 1996;143:3466. ´  Simon B, Flandrois S, Fevrier-Bouvier A, Biensan P. Mol Cryst Liq Cryst 1998;310:333. ´  Flandrois S, Fevrier-Bouvier A, Biensan P, Simon B. U.S. Patent 5, 554, 464.  Gay R, Gasparoux H. In: Groupe Franc¸ais d’Etude des Carbones, editors. Les carbones. Paris: Masson, 1965:63.  Fischer JE. In: Legrand AP, Flandrois S, editors. Chemical physics of intercalation. New York: Plenum Press, 1987:59.  Dahn JR. Phys Rev B 1991;44:9170.  Billaud D, Henry FX, Lelaurain M, Willmann P. J Phys Chem Solids 1996;57:775.  Safran SA. In: Legrand AP, Flandrois S, editors. Chemical physics of intercalation. New York: Plenum Press, 1987:47.  Dahn JR, Fong R, Spoon MJ. Phys Rev B 1990;42:6424.  Dahn JR, Sleigh AK, Shi H, Reimers JN, Zhong Q, Way BM. Electrochim Acta 1993;38:1179.  Zheng T, Dahn JR. Phys Rev B 1996;53:3061.  Satoh A, Takami N, Ohsaki T. Solid State Ionics 1995;80:291.  Takami N, Satoh A, Hara M, Ohsaki T. J Electrochem Soc 1995:2564.  Mabuchi A, Tokumitsu K, Fujimoto H, Kasuh T. J Electrochem Soc 1995;142:1041.  Tatsumi K, Zaghib K, Sawada Y, Abe H, Ohsaki T. J Electrochem Soc 1995;142:1090.  Endo M, Nishimura Y, Takahashi T, Takeuchi K, Dresselhaus MS. J Phys Chem Solids 1996;57:725.  Tatsumi K, Akai T, Imamura T, Zaghib K, Iwashita N, Higushi S, Sawada Y. J Electrochem Soc 1923;1996:143.  Kostecki R, Tran T, Song X, Kinoshita K, McLarnon F. J Electrochem Soc 1997;144:3111. ´  Flandrois S, Fevrier-Bouvier A, Guerin K, Simon B, Biensan P. Mol Cryst Liq Cryst 1998;310:389.  Inagaki M, Iwashita N, Hishiyama Y. Mol Cryst Liq Cryst 1994;244:89.  Dahn JR, Sleigh AK, Shi H, Way BM, Weycanz WJ,
     
                    
Reimers JN, Zhong Q, von Sacken U. In: Pistoia G, editor. Lithium batteries, new materials and perspectives. Amsterdam: Elsevier, 1993:1. Franklin RE. Proc Roy Soc (London) 1951;A209:196. ´ Maire J, Mering J. In: Walker PC, editor. Chemistry and physics of carbon, vol. 6. New York: Marcel Dekker, 1970:125. Saadaoui H, Roux JC, Flandrois S, Nysten B. Carbon 1993;31:481. Fitzer E, Mueller K, Schaefer W. In: Walker PL, editor. Chemistry and physics of carbon, vol. 7. New York: Marcel Dekker, 1971:193. Oberlin A. In: Thrower PA, editor. Chemistry and physics of carbon, vol. 22. New York: Marcel Dekker, 1989:1. Hoinkis E. In: Thrower PA, editor. Chemistry and physics of carbon, vol. 25. New York: Marcel Dekker, 1997:71. Kinney CR. Proceedings of the 1st and 2nd International Conference on Carbon. New York: Pergamon, 1955:83. Marchand A, Zanchetta JV. Carbon 1966;3:483. Cerutti M, Uebersfeld J, Millet J, Parisot J. J Chim Phys 1960;57:907. ` P, Carmona F. In: Walker PL, Thrower PA, editors. Delhaes Chemistry and physics of carbon, vol. 17. New York: Marcel Dekker, 1981:89. Yazami R, Deschamps M. Mater Res Soc Symp Proc 1995;369:165. Mori Y, Iriyama T, Hashimoto T, Yamazaki S, Kawakami F, Shiroki H, Yamabe T. J Power Sources 1995;56:205. Zheng T, Liu Y, Fuller EW, Tseng S, von Sacken U, Dahn JR. J Electrochem Soc 1995;142:2581. Matsumura Y, Wang S, Mondori J. Carbon 1995;33:1457. Sato K, Noguchi M, Demachi A, Oki N, Endo M. Science 1994;264:556. Wang S, Zhang Y, Yang L, Liu Q. Solid State Ionics 1996;86–88:919. Verbrugge MW, Koch BJ. J Electrochem Soc 1996;143:24. Jung Y, Suh MC, Lee H, Kim M, Lee SI, Shim SC, Kwak J. J Electrochem Soc 1997;144:4279. Zheng T, Zhong Q, Dahn JR. J Electrochem Soc 1995;142:L211. Yata S, Kinoshita H, Komori M, Ando N, Kashiwamura T, Harada T, Tanaka K, Yamabe T. Synth Metals 1994;62:153. Xiang HQ, Fang SB, Jiang YY. J Electrochem Soc 1997;144:L187. Xing W, Xue JS, Zheng T, Gibaud A, Dahn JR. J Electrochem Soc 1996;143:3482. Xing W, Xue JS, Dahn JR. J Electrochem Soc 1996;143:3046. Xing W, Dahn JR. J Electrochem Soc 1997;144:1195. Tokumitsu K, Mabuchi A, Fujimoto H, Kasuh T. J Electrochem Soc 1996;143:2235. Tatsumi K, Kawamura T, Higuchi S, Hosotubo T, Nakajima H, Sawada Y. J Power Sources 1997;68:263. Yamazaki S, Hashimoto T, Iriyama T, Mori Y, Shiroki H, Tamura N. J Mol Struct 1998;441:165. Zheng T, Mc Kinnon WR, Dahn JR. J Electrochem Soc 1996;143:2137. Liu Y, Xue JS, Zheng T, Dahn JR. Carbon 1996;34:193. Zheng T, Xue JS, Dahn JR. Chem Mater 1996;8:389. Sonobe N, Ishikawa M, Iwasaki T. Extended Abstracts of the 35th Battery Symposium, Nov. 14–16, Paper 2B10. Nagoya, Japan, 1994:49
S. Flandrois, B. Simon / Carbon 37 (1999) 165 – 180
 Matsumura Y, Wang S, Mondori J. J Electrochem Soc 1995;142:2914.  Puri BR. In: Walker PL, editor. Chemistry and physics of carbon, vol. 6. New York: Marcel Dekker, 1970:191.  Kikuchi M, Ikezawa Y, Takamura T. J Electroanal Chem 1995;396:451.  Marchand A. In: Walker PL, editor. Chemistry and physics of carbon, vol. 7. New York: Marcel Dekker, 1971:155.  Lowell CE. J Am Ceram Soc 1967;50:142.  Kouvetakis J, Kaner RB, Sattler ML, Bartlett N. J Chem Soc Chem Commun 1986:1758.  Ottaviani B, Derre´ A, Grivei E, Mohamed Mahmoud OA, ` P. J Mater Chem Guimon MF, Flandrois S, Delhaes 1998;8:197.  Way BM, Dahn JR. J Electrochem Soc 1994;141:907.  Kawaguchi M. Adv Mater 1997;9:615.  Ishikawa M, Morita M, Hanada T, Matsuda Y, Kawaguchi M. Denki Kagaku 1994;62:897.  Kawaguchi M, Kawashima T, Nakajima T. Denki Kagaku 1993;61:1403.  Ishikawa M, Nakamura T, Morita M, Matsuda Y, Kawaguchi M. Denki Kagaku 1994;62:897.  Weydanz WJ, Way BM, Van Buuren T, Dahn JR. J Electrochem Soc 1994;141:900.  Ito S, Murata T, Hasegawa M, Bito Y, Toyoguchi Y. J Power Sources 1997;68:245.  Nakajima T, Koh M, Takashima M. Electrochim Acta 1998;43:883.  Nakajima T, Koh M. Carbon 1997;35:203.  Peled E, Menachem C, Bar-Tow D, Melman A. J Electrochem Soc 1996;143:L4.  Ein-Eli Y, Koch VR. J Electrochem Soc 1997;144:2968.  Xue JS, Dahn JR. J Electrochem Soc 1995;142:3668.  Harker H, Horsley JB, Robson D. Carbon 1971;9:1.  Tidjani M, Lachter J, Kabre TS, Bragg RH. Carbon 1986;24:447.
 Disma F, Aymard L, Dupont L, Tarascon JM. J Electrochem Soc 1996;143:3959.  Xing W, Dunlap RA, Dahn JR. J Electrochem Soc 1998;145:62.  Xue JS, Myrth K, Dahn JR. J Electrochem Soc 1995;142:2927.  Wilson AM, Dahn JR. J Electrochem Soc 1995;142:326.  Wilson AM, Zank G, Eguchi K, Xing W, Dahn JR. J Power Sources 1997;68:195.  Tran TD, Feiker JH, Song X, Kinoshita K. J Electrochem Soc 1995;142:3297.  Schoenfelder HH, Kitoh K, Nemoto H. J Power Sources 1997;68:258.  Flandrois S. In: Bernier P, Lefrant S, editors. Le Carbone ´ dans tous ses etats. Gordon and Breach Science Publishers, 1997:517.  Idota Y. Eur. Patent 0 567 149 A1, 1993.  Idota Y, Nishima N, Miyaki Y, Kubota T, Miyasaki T. Can. Pat. 2 134 053, 1994.  Idota Y, Matsufuji A, Maekawa Y, Miyasaki T. Science 1997;276:1395.  Courtney IA, Dahn JR. J Electrochem Soc 1997;144:2943.  Guyomard D, Sigala C, Le Gal La Salle A, Piffard Y. J Power Sources 1997;68:692.  Denis S, Baudrin E, Touboul M, Tarascon JM. J Electrochem Soc 1997;144:4099.  Nishijima M, Kagohashi T, Takeda Y, Imanishi M, Yamamoto O. J Power Sources 1997;68:510.  Shodai T, Okada S, Tobishima S, Yamaki J. J Power Sources 1997;68:514.  Winans RE, Carrado KA. J Power Sources 1995;54:11.  Momose H, Honbo H, Takeuchi S, Nishimura K, Horiba T, Muranaka Y, Kozono Y, Miyadera H. J Power Sources 1997;68:208.  Aragane J, Matsui K, Andoh H, Suzuki S, Fukuda H, Ikeya H, Kitaba K, Ishikawa R. J Power Sources 1997;68:13.