Solid-state chemistry of lithium power sources†
`
`Peter G. Bruce*
`School of Chemistry, University of St. Andrews, St. Andrews, Fife, UK KY16 9ST
`
`This article describes the solid-state chemistry of inter-
`calation compounds that underpins a revolutionary new
`rechargeable lithium battery which has recently achieved
`phenomenal commercial success. The battery can store more
`than twice the energy compared with conventional alter-
`natives of the same size and mass and holds the key to the
`future improvement of consumer electronic products (e.g.
`mobile telephones), electric vehicles and implantable medi-
`cal devices (e.g. the artificial heart). Attention is focused on
`those lithium intercalation compounds that are useful as
`positive electrodes in rechargeable lithium batteries. The
`basic operation of the cell is summarised briefly and the
`structure/property relationships are developed that are
`important for the solid-state chemist when attempting to
`design and synthesise new lithium intercalation compounds
`capable of operating as positive electrodes. Finally, the
`structure, electronic structure and intercalation chemistry of
`several important positive intercalation electrodes are dis-
`cussed including some which show considerable promise for
`applications in future generations of rechargeable lithium
`batteries.
`
`Introduction
`Why should chemists be interested in rechargeable lithium
`batteries? For better or worse, chemical research must now find
`expression in its benefit to industrial competitiveness or the
`quality of life. This is not necessarily in conflict with its more
`timeless raison d’ˆetre, i.e. the advancement of knowledge. In
`many areas of chemistry it is possible, although often hard, to be
`interesting and useful at the same time. Some of the most
`exciting fundamental scientific developments in solid-state
`chemistry within the last decade, e.g. zeolites for catalysis, have
`occurred against a background of attempting to advance
`technology and reduce energy costs.
`The overwhelming need for lightweight and compact sources
`of portable electricity has resulted in a massive international
`effort into the development of radically new rechargeable
`batteries. This has led recently to the first successful com-
`mercialisation of a rechargeable lithium battery.1,2 The device is
`a triumph of solid-state chemistry and electrochemistry; the
`relevance of electrochemistry to lithium batteries is obvious,
`however if it were not for crucial advances in the solid-state
`chemistry of intercalation solids, rechargeable lithium batteries
`would not have achieved the success that they have. In this
`article I aim to highlight the key challenge presented to the
`solid-state chemist, namely, the development of new inter-
`calation compounds as positive electrodes for rechargeable
`lithium batteries.
`
`Rechargeable lithium batteries
`The considerable technological impetus in this area comes from
`three main sources, consumer electronics (e.g. mobile tele-
`phones), electric vehicles and implantable medical devices (e.g.
`the artificial heart).
`The introduction by Sony in 1990 of the world’s first
`commercially successful rechargeable lithium battery rep-
`resented a revolution in the power source industry.1–3 It has
`
`been likened by some to the semiconductor revolution which
`saw the replacement of the valve by the transistor in the 1940s
`and 50s. The new cell can store more than twice the energy
`compared with conventional rechargeable batteries of the same
`size and mass, an achievement which is remarkable in an
`industry that traditionally measures improvements in a few
`percentage points. The Sony cell represents but a first step on
`the road to greatly improved sources of portable electrical
`power, there is much scope and indeed need for further
`development.2
`The essential elements of the Sony cell are shown schemat-
`ically in Fig. 1(a). The cell is constructed in the discharged state
`and consists of a positive electrode composed of a thin layer of
`powdered LiCoO2 mounted on aluminium foil, a negative
`electrode formed from a thin layer of powdered graphite, or
`certain other carbons, mounted on a copper foil. The two
`electrodes are separated by a porous plastic film soaked
`typically in LiPF6 dissolved in a mixture of dimethyl carbonate
`and ethylene carbonate. Charging the cell involves diffusion of
`lithium ions within the LiCoO2 particles towards their interface
`with the electrolyte, the lithium ions then cross the electrolyte
`and are intercalated between the carbon layers in the graphite
`electrode. Charge balance requires that the equivalent number
`of electrons must pass around the external circuit. Discharge
`reverses the process moving lithium out of the graphite and
`reforming LiCoO2. This cell may be contrasted with earlier
`designs which employed lithium metal rather than graphite as
`the negative electrode, Fig. 1(b). The reactivity of lithium metal
`and the difficulty of plating and stripping it with high efficiency,
`have resulted in concerns over safety and the performance of
`this design. The Sony cell, which is known variously as a
`rocking-chair, swing or LION cell, to distinguish it from the
`lithium metal design, is a true expression of solid-state
`intercalation chemistry involving, as it does, the flow of lithium
`
`Negative
`LixC6
`
`Negative
`Li
`
`(a)
`
`Positive
`LiCoO2
`
`(b)
`
`Positive
`LiCoO2
`
`Charge
`Li+
`
`Li+
`
`Discharge
`
`Charge
`Li+
`
`Li+
`
`Discharge
`
`Fig. 1 Schematic representation of (a) the Sony rocking-chair cell and (b) a
`rechargeable lithium cell with a lithium metal negative electrode. The
`maximum lithium content in ordered graphite is LiC6.
`
`Chem. Commun., 1997
`
`1817
`
`SONY EXHIBIT 1015
`
`Page 1 of 8
`
`

`
`ions between two intercalation hosts. The performance ad-
`vantage over more conventional rechargeable batteries such as
`nickel–cadmium lies largely in the voltage. The Sony cell has an
`average potential of 3.6 V which is almost three times that of
`nickel–cadmium so that three conventional cells can be replaced
`by only one lithium cell!1,3
`A number of reviews have appeared which deal with
`rechargeable lithium batteries including graphite and other
`carbon-based negative electrodes.4–7 Although the Sony cell
`currently uses a liquid electrolyte, the future will see replace-
`ment of this by a solid polymer leading to all-solid-state
`rechargeable lithium batteries.8
`
`Relating solid-state chemistry to positive intercalation
`electrodes
`For our purposes, intercalation or insertion solids may be
`defined as hosts into which atoms (or more commonly ions +
`electrons) may be inserted or removed without a major
`disruption of the structure.9–12 Molecular intercalation has been
`reviewed by O’Hare.13
`In order to use our knowledge of solid-state chemistry to
`design and synthesise new intercalation compounds that will
`exhibit improved performance as positive electrodes for new
`generations of rechargeable lithium batteries, we must first
`understand what properties the intercalation solid should
`possess and how this relates to the structure and composition of
`the solid. The key structure/property relationships are listed in
`Table 1. As can be seen, a single compound must possess
`fourteen, often quite distinct, attributes and this serves to
`illustrate how challenging it is to develop solids which will be
`technologically competitive! Let us consider some of these
`attributes and their relationship to the solid-state chemistry in
`more detail.
`Firstly, positive electrodes must be intercalation compounds;
`only such compounds avoid the large kinetic barriers of defect
`diffusion and nucleation and growth associated with the
`majority of solid-state reactions. Also, as prepared, they should
`contain lithium, i.e. be in the fully discharged state since
`rocking chair cells are assembled in this state.
`The voltage of the cell is a crucial parameter and this should
`be large. If we consider a rocking-chair cell, such as graphite/
`LiCoO2, then the voltage between the two electrodes is related
`to the work the cell can deliver on transferring electrons around
`an external circuit and to the free energy change on transferring
`lithium from one intercalation electrode to the other. The free
`energy change associated with the transfer of one mole of
`lithium between the two intercalation electrodes is equivalent to
`the difference in the chemical potential of lithium in the two
`electrodes [eqn. (1)],14,15
`
`(1)
`
`
`(m Liint - m Li
`graph )
`V = -
`
`
`nF
`int and mLi
`where V is the open circuit voltage of the cell, mLi
`graph the
`chemical potentials of lithium in the positive intercalation and
`graphite electrodes respectively, n = 1 (since one e2 is
`transferred for each lithium) and F is Faraday’s constant. On
`discharge, lithium is transferred from a state of high mLi
`graph (high
`energy) in the negative graphite electrode to one of low mLi
`int (low
`energy) in the positive intercalation electrode; as a result work
`can be done by the cell. According to eqn. (1), in order to ensure
`a large cell voltage we must select positive intercalation
`electrodes which possess a low mLi
`int. On discharging a rocking-
`chair cell the lithium content and hence chemical potential and
`voltage, will change in each electrode (mLi = mo
`Li + RTlna,
`where a is the activity of lithium). It is convenient and normal
`practice therefore to measure the potential of a positive
`intercalation electrode against lithium metal (which has an
`invariant mLi) and to use the latter’s chemical potential as the
`standard state for lithium in the positive electrode. Lithium is
`rarely in the form of Li atoms in the intercalation compounds of
`
`1818
`
`Chem. Commun., 1997
`
`Table 1 Criteria for intercalation compounds as positive electrodes
`
`5
`6
`7
`
`1 Must be an intercalation host for lithium
`Low Fermi level and Li+ site energy ? high open-circuit voltage
`2
`Electrode potential varies little with lithium content ? cell voltage
`3
`varies little with state of charge
`4 Capable of accommodating large quantities of lithium per formula
`unit ? high capacity
`Low formula mass ? high gravimetric energy density
`Low molar volume ? high volumetric energy density
`Sustain high rates of lithium intercalation and deintercalation ?
`high cell discharge/charge rates
`8 Highly reversible lithium intercalation ? many charge–discharge
`cycles
`9 Avoid co-intercalation of solvent
`10
`Stable in contact with candidate electrolytes
`11 Adequate electronic conductivity
`12
`Low cost
`13
`Easily fabricated into electrode
`14
`Evironmentally friendly
`
`interest here but instead exists as Li+ ions along with their
`charge-compensating electrons, the latter located in the d levels
`of the transition-metal ion. As a result, it is helpful when
`considering structure–property relationships to separate mLi
`int
`according to eqn. (2)
`
`int
`
`int + me2
`int = mLi+
`mLi
`(2)
`int and me2
`where mLi+
`nt represent respectively the chemical poten-
`tials of Li+ ions and electrons.‡ The ion and electron chemical
`potentials include both energy and entropy terms. In developing
`the gross structure–property relationships the entropy terms
`may be neglected since over most of the composition range they
`are small, vary little and make a similar contribution to each
`intercalation compound. The entropy of electrons in a band is
`negligible.
`This torrid trip through some simple thermodynamics has
`brought us to the conclusion, important for the solid-state
`chemist, that the potential of a positive intercalation electrode,
`and hence the voltage of the cell, will depend on the energy of
`the electrons and the Li+ ions in the host. On inserting electrons
`into an intercalation host they will enter at the Fermi level, EF
`and this is the electron energy of importance [eqn. (3)].§¶
`EF = me
`(3)
`The site energy for Li+ is the major factor determining the ion
`contribution to the overall energy. Mutual repulsion between
`the Li+ ions is usually of secondary importance. Hence
`maximising the positive electrode potential reduces to design-
`ing intercalation compounds with a low Fermi level and a high
`stability (low energy) for Li+ in its sites. It is important to note
`that the electrode potential of graphite or other carbon
`electrodes is some +10 to +800 mV vs. lithium, therefore for
`rocking chair cells we must select intercalation electrodes with
`somewhat larger positive potentials (i.e. lower Li chemical
`potentials) than is necessary for cells with lithium metal
`electrodes.
`Let us consider further the problem of engineering a low
`Fermi level. The lowest Fermi level which can be achieved for
`an intercalation compound is determined by the energy
`corresponding to the top of the valence band. In oxides the
`valence band is largely of oxygen 2p parentage and lies
`significantly below the top of the 3p valence band in the
`corresponding sulphides. As a result, Fermi levels in oxides can
`be more than 2 eV lower, resulting in potentials of between 4
`and 5 V vs. the Li+/Li couple. Therefore, the focus is on oxides
`rather than chalcogenides. For any given transition-metal oxide
`the Fermi level is set by the position of the cation d levels. The
`lowest d levels are associated with those ions from the centre or
`right of the first transition series, i.e. Cr, Fe, Mn, Co, Ni; these
`all exhibit +4 oxidation states corresponding to d levels which
`lie close to (usually just above) the top of the oxygen valence
`
`Page 2 of 8
`
`Page 2 of 8
`
`

`
`band. Li1 2 xCr2O4, for example, when combined with a lithium
`negative electrode gives an open circuit potential of 5 V
`associated with the Cr4+/3+ mixed-valence state.17
`As lithium is inserted into the intercalation host forming a
`continuous range of solid solutions, the electrons fill up the
`band while the mutual repulsions between the Li+ ions rises.
`The net effect is that the potential will decrease somewhat with
`increasing Li content; whereas in many systems it turns out that
`this can be rationalised entirely by considering only the ion
`repulsions,14 in some cases both ion and electron interactions
`are important.18 If, on the other hand, intercalation induces the
`formation of a new phase and both phases have fixed
`composition then insertion is simply accompanied by the
`conversion of one phase to the other. This in turn is associated
`with a constant change of chemical potential and therefore a
`constant voltage as a function of the degree of intercalation
`(state-of-charge).
`Before leaving the topic of electrode potentials it is worth
`noting that the relationship between the cell voltage and the
`lithium chemical potentials in the electrodes can be derived in
`another way which serves to highlight a frequently misunder-
`stood fact. It is often stated that the voltage of a cell is
`determined by the difference between the Fermi levels of the
`two electrodes, this is incorrect. The voltage is equal to the
`difference in the Fermi levels between the electrodes only when
`they are in a cell i.e. in mutual contact with an electrolyte. Such
`Fermi levels are not those of the isolated electrodes. Selecting a
`lithium intercalation electrode based on its Fermi level will not
`yield the desired cell voltage.•
`One of the most important factors governing the performance
`of a cell is the energy which can be stored per unit weight and
`volume, i.e. the gravimetric and volumetric energy density. For
`electric vehicles, batteries are required which provide a high
`gravimetric energy density whereas for implantable medical
`devices volumetric energy is more important. Since energy is
`stored in the form of lithium in the intercalation electrodes the
`gravimetric and volumetric energy density of the positive
`electrode is of great importance and is given by eqn. (4),
`(4)
`E = VeQ
`where E is the volumetric or gravimetric energy density, Ve is
`the electrode potential and Q is the charge stored per unit mass
`or volume of the intercalation compound. Energy densities are
`conventionally expressed in terms of W h kg21 or W h l21
`whereas Q is expressed in terms of mA h g21 or A h l21. The
`energy density depends therefore on two separate factors, the
`electrode potential and the charge. Potential has been dealt with
`above. The charge stored is equivalent to the amount of lithium
`that can be accommodated within the intercalation host. It is
`important for the solid-state chemist to design intercalation
`hosts which can reversibly insert a large amount of lithium per
`formula unit. If the capacity to store lithium is high, then a high
`gravimetric energy density will be assured provided the molar
`mass of the intercalation host is small. A high volumetric energy
`density will be obtained provided the molar volume is small.
`We see again that oxides will be preferable to chalcogenides in
`order to achieve a high gravimetric energy density and that we
`wish to choose intercalation hosts that contain the active redox
`couple and little else. Considering both Ve and Q, a metal oxide
`of the formula LiMO2 from which all the lithium may be
`reversibly deintercalated and in which M is a light element, in a
`high oxidation state (M4+/3+ couple) and with a close-packed
`structure would seem to be optimal. It is no coincidence that the
`positive electrode in the Sony cell is LiCoO2.
`The rate at which a cell can be discharged or charged is an
`important parameter and is limited by the rate of lithium
`intercalation or deintercalation. Electroneutrality demands that
`the magnitude of the flux of Li+ ions and their charge
`compensating electrons into and out from the host must be the
`same. It is the coupled diffusion of Li+ and electrons down a
`concentration gradient that is important, rather than the
`
`Fig. 2 The crystal structure of LiCoO2 or LiNiO2. Dark and light octahedra
`represent respectively the sites for the transition-metal and lithium ions.
`
`individual diffusivities (mobilities) of the ions and electrons
`alone. A coupled diffusion coefficient may be defined, it is
`generally known as the chemical diffusion coefficient, ˜DLi, and
`is a measure of the intrinsic rate capability of the electrode.
`Several other factors are also of crucial importance. The
`chosen electrode must be stable in contact with the electrolyte of
`the cell. The surfaces of transition-metal oxides provide an ideal
`environment for catalytic decomposition of many electrolytes
`and this can present problems. Indeed metal oxide electrodes are
`frequently used for electrocatalysis in other applications! The
`starting materials used to prepare the intercalation electrode
`should be of low cost as should the preparation method.
`However, above all, the intercalation electrode must be
`environmentally benign.
`
`Intercalation electrodes
`Many intercalation compounds have been studied as possible
`positive electrodes for rechargeable lithium batteries. Limita-
`tions of space demands that attention is concentrated on the
`three compounds of main interest. The author apologises to
`those whose work could not be included.
`
`LiCoO2 and LiNiO2
`19–21
`The structures of LiCoO2 and LiNiO2 are identical and shown in
`Fig. 2. The oxide ions adopt a cubic close-packed arrangement
`with Co3+ or Ni3+ ions occupying octahedral sites between
`adjacent oxide ion layers. Only alternate sheets of octahedral
`sites are occupied by the transition metal ions with the
`remaining sheets being occupied by Li+. The structure is
`slightly distorted giving rise to rhombohedral symmetry, space
`–
`group R3
`m.
`The electronic structure of intercalation electrodes is an
`important factor which as we have seen has a strong influence
`on the voltage and can also induce structured distortions. Too
`little attention has been paid to the rˆole of electronic structure in
`the past. The electronic structure of LiCoO2 may be derived as
`0) in
`6eg
`follows. Co3+ (3d6) adopts a low-spin configuration (t2g
`an octahedral oxygen environment. The t2g orbitals are
`energetically close to the oxygen 2p orbitals and those 2p
`orbitals of p symmetry will overlap strongly with the Co t2g
`orbitals. As a result there is considerable covalent mixing and
`consequently a high degree of delocalisation over the CoO6
`octahedra. The Co–O–Co angle is close to 90° and hence
`neighbouring cobalt ions overlap with mutually orthogonal 2p
`orbitals on the bridging oxygens. This inhibits long-range Co–
`O–Co interactions, thus attenuating band formation due to
`metal–oxygen interactions. Spectroscopic data suggests that the
`electronic structure of LiCoO2 more closely approximates to a
`localised Co(3d6) electronic configuration than to a band. Full
`occupancy of the t2g p* orbitals which are separated energet-
`ically from the eg s* orbitals ensures semiconductor behaviour
`for LiCoO2. The extraction of lithium from between the oxide
`ion layers is facile and a chemical diffusion coefficient for the
`
`Chem. Commun., 1997
`
`1819
`
`Page 3 of 8
`
`Page 3 of 8
`
`

`
`first believed, instead several minor structural transformations
`occur.28 Within the composition range 0.9 < x < 1 in LixCoO2,
`a single hexagonal phase exists, whereas within the range 0.78
`< x < 0.9, two hexagonal phases coexist giving way to a
`region, 0.51 < x < 0.78, within which the second hexagonal
`phase exists alone. On further removal of lithium a monoclinic
`distortion of this second phase is observed for compositions in
`the narrow range 0.46 < x < 0.51. Extraction of lithium from
`x = 1 to x = 0.5 is accompanied by an expansion in the c lattice
`parameter from approximately 14.1 to 14.5 Å. The loss of
`lithium from between the otherwise weakly van der Waals’
`bonded oxide layers is the origin of this expansion. Decrease in
`the a lattice parameter associated with oxidation of Co3+ to the
`smaller Co4+ has been highlighted above. Further deintercala-
`tion below x = 0.46 results in reversion to a hexagonal phase
`until the composition Li0.22CoO2 is reached whereupon a
`second monoclinic phase appears coexisting with the hexagonal
`phase until the composition Li0.18CoO2 is obtained, at which
`point the monoclinic phase exists alone. Below x = 0.15
`another hexagonal phase appears corresponding to the fully
`delithiated CoO2.28 The monoclinic distortion at around x = 0.5
`has been ascribed to lithium ordering,27 however a complete
`rationale for the structural phase changes which occur on
`deintercalation has not yet been presented; nevertheless, there is
`no doubting the complexity of the structural chemistry. Until the
`recent studies by Amatucci, Tarascon and Klein,28 it had been
`generally believed that the weak van der Waals’ bonding
`exhibited by adjacent oxide ion layers would result in instability
`for a layered CoO2 compound. These authors have shown this to
`be incorrect and have demonstrated that the structure of CoO2 is
`not based on cubic close-packing but hexagonal close-packing
`(CdI2 structure). This is the same structure commonly adopted
`by the layered chalogenides, e.g. TiS2.28 Two factors may play
`a rˆole in stabilising CoO2 with adjacent oxide layers. First, the
`low-lying d5 configuration of Co4+ ensures very significant
`covalent mixing with the neighbouring oxygen 2p orbitals
`resulting in a high degree of delocalisation. It is not therefore
`realistic to view the compound as possessing adjacent O22
`layers. Second, as proposed before, a cooperative displacement
`of the Co4+ ions results in a ferroelectric distortion establishing
`dipoles which will oppose the repulsive interactions between
`the oxide ion layers thus assisting stabilisation of the struc-
`ture.22
`The structural chemistry of LixNiO2 is also complex.27,29,30
`Solid solutions with rhombohedral symmetry exist within the
`composition ranges 0.85 < x < 1 and 0.32 < x < 0.43.
`Between x = 0.50 and 0.75 a single monoclinic phase is
`observed. Over the range x = 0–0.32 a mixture of two
`rhombohedral phases is evident, one of these being NiO2. Two-
`phase mixtures are also observed between each of the other
`single-phase ranges. The monoclinic distortion is believed to be
`associated with a superlattice arising from the ordering of
`lithium ions.24,27,29 Arguments similar to those used for CoO2
`may be employed to explain the stability of NiO2 particularly
`the high degree of Ni–O covalency. It is interesting to note that
`NiO2 retains the cubic close packed structure of LiNiO2 (i.e. the
`CdCl2 structure type).29
`In contrast to LiCoO2, LiNiO2 has proved impossible to
`prepare as a stoichiometric material. Instead compounds with
`the formula Li1 2 dNi1 +dO2 are obtained. The closest approach
`to stoichiometry that has been possible so far corresponds to a
`d of approximately 0.02 and may be prepared by reacting Li2O2
`with NiO in oxygen at 700 °C.31 Non-stoichiometry is
`associated with lithium deficiency and the presence of Ni2+ ions
`in the lithium layers. These nickel ions serve to pin the oxide
`layers together reducing the mobility of the lithium ions,
`however compounds approaching the stoichiometric composi-
`tion do exhibit a similar c lattice expansion on deintercalation to
`that observed for LiCoO2. There is evidence that at high degrees
`of deintercalation further nickel ions may migrate from the Ni to
`the Li layers. Work by Delmas et al. has shown that the quantity
`
`(a)
`
`1.0
`5.0
`
`x In LixNiO2
`0.5
`
`0.0
`
`4.5
`
`4.0
`
`3.5
`
`3.0
`
`2.5
`
`0.0
`
`0.2
`
`0.6
`0.4
`x In LixCoO2
`
`0.8
`
`1.0
`
`(b)
`
`0
`
`50
`
`150
`200
`100
`Q / mA h g–1
`
`250
`
`300
`
`5.5
`
`5.0
`
`4.5
`
`4.0
`
`3.5
`
`3.0
`
`2.5
`
`/ V
`
`E
`
`Fig. 3 Open-circuit voltage vs. Li+/Li couple as a function of lithium content
`in (a) LixCoO2, reproduced from ref. 28 and (b) LixNiO2, based on a figure
`presented in ref. 29. With the permission of The Electrochemical Society.
`
`coupled (Li+ + e2) diffusion of 5 3 1028 cm2 s21 at Li0.65CoO2
`has been measured.22 Such extraction is accompanied by
`oxidation of Co3+ to Co4+ (low-spin d5). The associated
`contraction of the a axis of the unit cell (which lies in the oxide
`layers) signals a shorter Co···Co distance, reduced from 2.83 Å
`in the case of LiCoO2 to 2.81 Å for Li0.9CoO2. The strongly
`covalent Co–O interactions, will, due to the nephaluxetic effect,
`produce expanded antibonding d orbitals. Together the orbital
`expansion and shorter Co···Co distance conspire to make
`possible direct overlap of cobalt t2g orbitals across a shared
`octahedral edge resulting in a narrow d band and hole
`conduction in the mixed valence Co4+/3+ system. The critical
`Co···Co distance for direct t2g–t2g interactions is 2.82 Å,
`consistent with the switch from insulator to metallic behaviour
`on extracting a relatively small amount (10%) of lithium from
`LiCoO2. Experimental evidence verifies that extraction of small
`quantities of lithium from LiCoO2 does indeed induce metallic
`behaviour.23 Of course we cannot rule out the possibility of
`Co4+ d levels being located just below the top of the oxygen 2p
`band and introducing holes in it which would also lead to
`metallic behaviour; further studies are required to resolve this.
`In any case the low lying Co4+/3+ couple gives rise to a high
`potential of > 4 V vs. the Li+/Li couple, Fig. 3.
`LiNiO2 adopts a low-spin d7 configuration with fully
`occupied t2g levels and one electron in the eg orbitals. Full
`occupancy of the t2g levels implies that we must focus our
`attention on the single electron in the eg orbitals to understand
`the conduction process. The eg orbitals will overlap with oxygen
`2p orbitals of s symmetry giving rise to strong interactions but
`again neighbouring nickel ions form a 90° Ni–O–Ni geometry
`and therefore overlap is with mutually orthogonal 2p orbitals on
`the bridging oxygen. This inhibits delocalisation extending
`beyond the NiO6 octahedron and again attenuates band
`formation. As a result the eg electron remains in a localised
`atomic orbital on nickel although these are strongly antibonding
`associated with significant Ni–O covalency. A localised low-
`spin d7 configuration is Jahn–Teller active. There is some
`evidence for a dynamic Jahn–Teller distortion from EXAFS
`studies but no cooperative, static, distortion of the structure has
`been detected.24 Deintercalation is again facile with the
`chemical diffusion coefficient for Li reaching a maximum of 2
`3 1027 cm2 s21, amongst the highest observed in such
`materials.25 Extraction of lithium is associated with oxidation of
`Ni3+ to Ni4+. Since the eg orbitals point towards the coordinating
`oxygens there is little opportunity for direct nickel–nickel
`interaction in contrast to the situation with the t2g orbitals in the
`cobalt compound. As a result, measurements of electronic
`transport on LixNiO2 exhibit a small but definable activation
`energy for electronic transport associated with small polaron
`hopping in the mixed-valence Ni4+/3+ state.26 The location of
`the Fermi level in the higher energy eg orbitals, compared with
`the lower energy t2g orbitals in the analagous cobalt compound,
`results in an average voltage several hundred mV lower for
`LixNiO2, Fig. 3(b).19,27,28
`Considering now the structural changes that accompany
`lithium deintercalation from LiCoO2 in more detail; this process
`does not correspond to a simple continuous solid solution as was
`
`1820
`
`Chem. Commun., 1997
`
`Page 4 of 8
`
`Page 4 of 8
`
`

`
`of lithium extracted on first charging lithium nickel oxide
`cannot be reinserted.24 In fact these authors claim that the
`capacity loss on the first cycle is a more significant factor than
`any loss of capacity due to continuous disordering of nickel on
`extended cycling. Even for relatively stoichiometric materials,
`reintercalation up to the composition Li0.85NiO2 is easy but
`beyond this limit proves very difficult. The mechanism they
`propose to explain this capacity loss is based on oxidation of the
`nickel ions in the lithium layers at high degrees of deintercala-
`tion, subsequent structural collapse around the Ni3+ ions and the
`relatively high charge of these ions mitigates against Li+ ions
`reoccupying the sites around nickel. The best electrochemical
`performance is obtained from materials that are as close as
`possible to the stoichiometric composition.
`Comparing the properties of the LiCoO2 and LiNiO2
`compounds as positive electrodes for rechargeable lithium
`batteries, both exhibit high voltages and facile lithium diffusion
`rates despite the complex structural changes. The fact that these
`structural changes are small and in some cases, at least, only
`associated with lithum ordering, explains the preservation of
`fast cycling kinetics. Complete removal of lithium from LiNiO2
`corresponds to a charge storage capacity of 275 mA h g21. In
`practical cells based on liquid electrolytes cycling can occur
`over the approximate composition range 0.3 < x < 0.9
`corresponding to 150–160 mA h g21 depending on the rate of
`charge/discharge. In the case of LiCoO2, deintercalation to
`Li0.5CoO2 is possible with almost complete reinsertion of
`lithium, corresponding to a more modest 130 mA h g21 but at
`a slightly higher potential than the nickel electrode. As we now
`know the limits of deintercalation for this compound are not
`imposed by structural collapse per se but by the instability of the
`highly charged, and therefore high voltage, oxide in contact
`with the liquid electrolytes used in the commercial cells.21 It is
`the somewhat more modest voltage of LiNiO2 which gives rise
`to the greater capacity. LiNiO2 is also attractive because it is
`cheaper than the cobalt alternative. However, LiCoO2 has the
`significant advantage that it is easier to synthesise as a highly
`stoichiometric material and its structure is robust with respect to
`cycling, at least over the limited composition range stated
`above, with no evidence for cobalt disordering into the lithium
`layers. It is this that led to its selection by Sony for the first
`generation rechargeable lithium batteries.
`
`Li(NiAl)O2 and Li(NiCo)O2
`In an attempt to improve on the properties of LiNiO2 and take
`advantage of the lower cost of this material, partial substitution
`of nickel by other ions has been investigated. The work of
`Ohzuku et al. has shown that LiAl0.25Ni0.75O2 can deliver a
`capacity of ca. 150 mA h g21 and with excellent capacity
`retention on cycling.32 A complete range of solid solutions
`Li(Ni1 2 yCoy)O2 may be prepared. It has been shown that
`partial replacement of Ni by Co yields a material which is
`cheaper than LiCoO2 but with many of its advantages.33 This
`solid solution will be used in future commercial rocking chair
`cells.
`
`LiMn2O4 and related spinels
`Manganese is approximately 1% of the cost of cobalt and
`significantly more environmentally benign than either cobalt or
`nickel. These are major advantages and explain the intense
`interest in developing manganese oxide based positive elec-
`trodes. The use of LiMn2O4 spinels as positive electrode
`materials was first reported in 1983.34–36 The attractive features
`of LiMn2O4 led in 1996 to the announcement by Nippon Moli,
`Japan of the first commercial rechargeable lithium battery in
`which LiCoO2 is replaced by the lithium manganese oxide
`spinel.37
`LiMn2O4 is quite