Abstract
High temperature 
1.10; 
0.03, 0.06, and 0.10) phases were synthesized by solid-state chemistry. Their characterization by X-ray diffraction and galvanostatic measurements shows that 3% of Ni ions substituted for Co in the 
lattice suppress the two-phase domain, related to the semiconductor-to-metal transition, that is observed at the beginning of the charge process in 
These ions, trapped in the lattice, prevent the phase separation. On the other hand, more than 10% of Ni ions need to be substituted for Co in order to inhibit the monoclinic distortion due to a lithium/vacancy ordering in the interslab for 
Besides, a 
ratio 
strictly higher than one in 
leads, as in the case of the unsubstituted 
phase, to the disappearance of all the phase transitions upon deintercalation. 
magic angle spinning nuclear magnetic resonance measurements show that 
ions are the only paramagnetic species in the 
phases while in the overlithiated 
phases, 
and intermediate spin 
are present. This suggests the existence of structural defects associated with O vacancies which are responsible for the suppression of the electronic delocalization and of the lithium/vacancy ordering upon lithium deintercalation. © 2002 The Electrochemical Society. All rights reserved.
Export citation and abstract BibTeX RIS
is the most widely used positive electrode material in commercial Li-ion batteries.1
2 The high-temperature form of 
crystallizes in the trigonal system (space group: 
with the ideal layered 
structure in which 
and 
octahedra share their edges and are stacked alternately along the 
axis direction with an AB CA BC oxygen packing.3 Numerous investigations have reported the phase changes upon lithium deintercalation in 
with the existence of a two-phase domain for 
due to a macroscopic semiconductor-to-metal transition and the presence of a monoclinic distortion for 
due to the establishment of an interslab Li/vacancy ordering.1
4
5
6
7 However, since the earliest work, in an effort to improve its electrochemical performances, substitution of numerous metallic elements for cobalt has been attempted (Cr,8 Al,9
10 Mg,11
12 Fe,13 or Mn14
15). However, only the derivatives of the mixed 
oxide, for which a complete solid solution exists between the two end members 
and 
seem to be an interesting candidate from the electrochemical point of view.5
16
17
In this context, in 1993 Reimers et al. studied the effect of Ni doping in 
They showed that 2% of the nickel ions substituted for cobalt suppress the occurrence of the monoclinic distortion in 
concluding that phase transitions are very sensitive to impurity levels and could be used as a quality control check on the material.18 More recently, we observed that both phase transitions upon deintercalation can be suppressed as well in the unsubstituted lithium cobalt oxide by using nominal Li/Co ratios strictly higher than 1.0 in a high-temperature synthesis.19
In this general context, we performed a study on the combined effects of Ni doping and Li overstoichiometry in 
We synthesized 
1.10; 
0.03, 0.06, and 0.10) materials and characterized them using X-ray diffraction (XRD), galvanostatic intercalation/deintercalation tests, and 
magic angle spinning nuclear magnetic resonance (MAS NMR) spectroscopy.
Experimental
The 
1.10; 
0.03, 0.06, and 0.10) materials were prepared by calcination at 600°C under 
for 12 h of 
(Rho⁁ne Poulenc Rectapur % min 99), 
[calcination at 450°C for 12 h under 
of 
Carlo Erba % min 99], and NiO (calcination at 450°C for 12 h under 
of 
Carlo Erba % min 99); 
denotes the nominal 
ratio of the mixture; two compositions were considered in this study, 
and 1.10. Two successive heat-treatments at 900°C for 24 h with an intermediate grinding were done after the calcination.
Electrochemical measurements were carried out at room temperature (22°C) with 
-propylene carbonate 
cells. The positive electrode consisted of a mixture of 88% by weight active material, 2% polytetrafluorethylene (PTFE), and 10% carbon black. The first series of cells, assembled in an argon-filled dry box, was charged at 
the second series, for long-range cycling, was cycled at 
C/20 rate). The 
deintercalated materials for XRD characterization were recovered in an argon-filled dry box, washed in dimethyl carbonate (DMC) and dried under vacuum.
The XRD patterns of the 
starting materials were recorded using a Siemens D5000 powder diffractometer using the Cu Kα radiation and a graphite diffracted beam monochromator. Rietveld refinements were performed using the Fullprof program with a pseudo-Voigt fitting function.20 XRD patterns of the 
deintercalated materials were recorded on a Philips PW1820 powder diffractometer using the Cu Kα radiation, in a special airtight holder under argon atmosphere in order to prevent any reaction with air moisture.
MAS NMR spectra were recorded on a Bruker MSL200 spectrometer at 77.7 MHz, with a standard 4 mm Bruker MAS probe. The samples were mixed with dry silica (typically 50% by weight), in order to facilitate the spinning and improve the field homogeneity, since they may exhibit metallic or paramagnetic properties. The mixture was placed into a 4 mm diam zirconia rotor in the dry box. Spinning speeds 
of 10 and 15 kHz were used. For all phases, a Hahn echo sequence 
was utilized in order to facilitate the phasing of all the signals and of their spinning side bands and to ensure the observation of possibly very wide signals which would be lost during the receiver dead time in single pulse experiments. The 90° pulse duration used 
was equal to 3.05 μs. In order to synchronize the spin echo with the first rotational echo, 
was fixed to the rotor period 
With such a Hahn echo sequence, phasing can be done easily, with only zero order correction if required, and no baseline correction is needed. A 200 kHz spectral width was used, and the recycle time 
is long enough to avoid 
saturation effects. The isotropic shifts reported in parts per million are relative to an external sample of 1 M LiCl solution in water.
Since the purpose of this study is to emphasize the combined effect of Ni doping and Li overstoichiometry on the properties of 
phases, all the results reported in this work were obtained for the same batch for each 
material. Furthermore, in order to control the reproducibility of the experimental results, two different batches of each material were characterized.
Results and Discussion
X-ray diffraction study. Rietveld refinements.—
All the starting 
materials were characterized by XRD. In all cases, pure phases are obtained without any trace of synthesis reactants or parasitic phases. All materials exhibit narrow diffraction lines whatever the percentage of nickel ions in the structure. They crystallize in the trigonal system (space group: 
, with the ideal 
layered structure of 
Since the XRD patterns are very similar and close to those reported in the literature for the 
system,5
16
17
21 only that of the 
material is shown in Fig. 1.
Figure 1. Observed and calculated XRD profiles for 
filled circles: observed; solid line, calculated; lower trace, difference plot; bars, reflections.
All the 
samples were characterized by Rietveld refinement of powder XRD data. As the morphology of the crystallites could influence the line intensity ratios on diffraction patterns, all the Rietveld samples were prepared in such a way that preferred orientation was minimized.22 The refinement was led, in a first step, using a strictly 2D model. Global occupancies for the 3a 
3b (Li), and 6c (O) sites were fixed to 1.0. Due to Co absorption, the release of the occupancy values of the 3a and 3b sites always leads to a decrease of the Co/Ni (3a) site occupancy, no significant change in the occupancy value for the Li 3b site being observed. The results are presented in Table I for the 
phase and are in agreement with the structural data reported for pure 
phases by Rougier et al.21 In our 
phases, the hypothesis of a fraction of Co and/or Ni ions in the lithium 3b site is rejected since the fits invariably lead to a negative occupancy. However, due to the impossibility of quantifying Co absorption in the Bragg-Brentano geometry, no conclusion could be drawn about the presence of Li ions in the 3a crystallographic site of Co/Ni. Nevertheless, in agreement with the previous neutron diffraction studies on pure 
one can consider the amount of 
ions in the Co/Ni 3a site as negligible.3
19 The very good fitting of the profile of the XRD lines and the small intensity of the line in the difference pattern emphasize the quality of the Rietveld refinement.
Table I.
Rietveld refinement of the  XRD spectrum. | ||||||
|---|---|---|---|---|---|---|
| ||||||
| Wyckoff positions | ||||||
| Atom | Site |
 |
 |
| Occupancy |
 |
| Li(1) | 3b | 0 | 0 | 1/2 | 1.0 |
 |
| Co(1) | 3a | 0 | 0 | 0 | 0.94 |
 |
| Ni(1) | 3a | 0 | 0 | 0 | 0.06 |
 |
| O | 6c | 0 | 0 |
 | 1.0 | 0.05(16) |
Profile function: pseudo-Voigt       
| ||||||
Conventional Rietveld factors:    14 refined parameters. | ||||||
The standard deviations were multiplied by the Scor parameter  to correct for local correlations. | ||||||
The structural parameters of the various 
materials are presented in Table II. Both the 
and the 
parameters increase smoothly with Ni doping due to the difference in size between the trivalent nickel and cobalt ions 
23). Furthermore, in all materials, the 
ratio close to five 
for a cubic lattice) confirms the 2D character of the structure of the Ni-doped 
phases,21
24 in good agreement with the high and constant value of the oxygen 
position 
and the consequent difference in the slab 
and interslab space 
thicknesses 
for 
phases).25
Table II.
Hexagonal cell parameters and oxygen  position from refinement of XRD patterns for the various  phases. | ||||
|---|---|---|---|---|
| Sample |
 (Å) |
 (Å) |
a |
 |
b | 2.81506(16) | 14.0516(8) | 0.2603(4) | 4.99 |
 | 2.8177(1) | 14.056(1) | 0.2582(9) | 4.99 |
 | 2.8176(2) | 14.062(1) | 0.2597(5) | 4.99 |
 | 2.8182(2) | 14.067(1) | 0.2584(9) | 4.99 |
b | 2.81605(18) | 14.0511(8) | 0.2593(6) | 4.99 |
 | 2.8172(1) | 14.058(1) | 0.2597(7) | 4.99 |
a 6c oxygen  position. | ||||
| b Ref. 19. | ||||
Electrochemical Study
Lithium-stoichiometric 
phases (y0.0, 0.03, 0.06, and 0.10).—
Figure 2 shows the first galvanostatic charge curve of 
cells 
0.03, 0.06, and 0.10) at low rate 
C/160 rate (160 h are required to exchange one equivalent of electron)].
Figure 2. First galvanostatic charge of 
0.03, 0.06, and 0.10) electrochemical cells 
.
As already reported, the charge curve obtained with the pure 
material exhibits both the voltage plateau at ca. 3.93 V corresponding to the biphasic domain for 
due to a semiconductor-to-metal transition and the particular feature associated to the monoclinic distortion expected for 
due to the interslab lithium/vacancy ordering.4
7
A quite different behavior is observed for the Ni-doped materials. In the first part of the oxidation of the 
phases, the preferential oxidation of Ni ions compared to Co ones leads to the appearance of a lower potential step, all the more pronounced when the nickel amount increases. This difference in potential is due to the lower oxidizing power of the 
redox couple compared to the 
one.17
26 Moreover, the voltage plateau at the beginning of the charge process is less marked for the Ni-doped materials, even with only 3% of the doping element in the 
lattice, suggesting that the two-phase domain does not exist anymore in these materials. This is fully confirmed by the XRD patterns of electrochemically deintercalated 
materials; as shown in Fig. 3 and 4, a solid solution is obtained for 
in the 
system. One can see in Fig. 2 that the monoclinic distortion for 
exists in all the 
0.03, 0.06, and 0.10) systems studied here. This result conflicts with the conclusion by Reimers et al. who showed in a former study that 2% of the Ni ions substituted for cobalt are sufficient to suppress the monoclinic distortion at room temperature.18 Besides, one can see on the inset of Fig. 2 that this order/disorder transition tends to disappear for Ni contents higher than 10%.
Figure 3. XRD patterns of the electrochemically deintercalated 
samples 
.
Figure 4. Expansion of the (003) Bragg reflection of the XRD patterns of the electrochemically deintercalated 
samples 
.
Lithium overstoichiometric phases (y0.0 and 0.03).—
Contrary to the case of the Li stoichiometric materials, as shown in Fig. 5, the first galvanostatic charge curves of 
cells 
and 0.03) at a low rate 
C/180 charging rate) have a quite similar profile, whatever the value of 
The feature corresponding to the monoclinic distortion for 
and the voltage plateau at the beginning of the charge have disappeared both for the Ni-doped and for the pure 
phases. The unidirectional increase of the potential during lithium deintercalation suggests the occurrence of a monophasic reaction throughout the charge process for the 
materials. An XRD study of electrochemically deintercalated 
materials carried out in a former work had confirmed such an observation.19 The same behavior is expected for the 
system, although no XRD characterization was performed on deintercalated materials.
Figure 5. First galvanostatic charge of 
0.03) electrochemical cells 
.
Cycling tests.—
1.10; 
0.03, 0.06, and 0.10) cells were cycled between 2.7 and 4.15 V at 
(C/20 rate) as shown in Fig. 6. The reversible capacity for the first and tenth cycle is also indicated on the figure. The purpose of this study is to characterize the structural transitions and not to obtain large reversible capacities, a relatively low high-voltage cutoff was used.
Figure 6. Cycling behavior of the various 
1.10; 
0.03, 0.06, and 0.10) electrochemical cells 
C/20 rate) between 2.7 and 4.15 V.
As already reported elsewhere, the cycling behavior of the Co-rich 
phases is very satisfactory and close to that of pure 
16
24 The cycling curves, on Fig. 6, show that the reversibility of the deintercalation/intercalation process is good, as the reversible capacities for the first and tenth cycles, shown in Fig. 6, are quite high in the voltage window used. Note that the slight decrease in potential caused by 3% doping of Ni leads to a particularly good cycling capacity for the cutoff voltage used (4.15 V). Furthermore, one can note that polarization is very small in all cases and that the irreversible capacity at the first cycle is especially low compared to that of the 
-type compounds.27
The derivative curves 
of the first and tenth cycles of the 
cells presented just above are reported in Fig. 7. One can note the peak at the lower potential due to the preferential oxidation of 
ions compared to 
ones at the beginning of the charging process. Furthermore, as already evidenced from the slow rate charging curves (Fig. 2), the trace of the monoclinic distortion around 4.15 V is easily observable in all the cycling curves from the Li stoichiometric 
phases and still exists after ten cycles, confirming the persistence of an order/disorder transition for 
in these materials. On the contrary, it does not appear for lithium overstoichiometric materials.
Figure 7.
derivatives curves of the first (solid line) and 10th (dashed line) cycles of the various 
1.10; 
0.03, 0.06, and 0.10) electrochemical cells 
between 2.7 and 4.15 V.
MAS NMR Study
MAS NMR spectroscopy measurements were carried out on the various 
starting materials. Figures 8 and 9 show the 15 kHz 
MAS NMR spectra of the lithium-stoichiometric 
materials 
0.03, and 0.06) and of the Li overstoichiometric 
phase.
Figure 8. 15 kHz 
MAS NMR spectra for the various 
phases 
1.10; 
0.03, 0.06), 
side bands).
Figure 9. Expansion of the central signal of 15 kHz 
MAS NMR spectra for the various 
phases 
1.10; 
0.03, 0.06).
As already reported, 
exhibits a single narrow signal close to 0 ppm due to the exclusive presence of diamagnetic low spin 
ions in the structure, so that no hyperfine interaction is expected.28
29 For lithium stoichiometric Ni-doped materials, three new signals at 110, 
and 
are observed. These signals have been attributed by Marichal et al. to the presence of paramagnetic low spin 
ions with one single electron in the 
orbital 
The transfer of hyperfine interaction between the single 
electron toward the surrounding lithium ions can be done via the overlap with the oxygen orbitals with a 90° (first neighbor) or 180° geometry (second neighbor). The 110 ppm signal is assigned to Li ions with one 
ions as the first neighbor while the 
and 
signals result from the presence of Li ions with one and two 
ions as the second neighbor, respectively.26
29 As expected, the magnitude of the 
ions signals is all the more pronounced since the percentage of Ni ions in the material is large.
Furthermore, one can observe in Fig. 9 that the central signal of the 
material is quite different from those of the lithium stoichiometric Ni-doped phases, showing characteristic shoulders centered approximately at 3 and 
These features are similar to those observed in the case of the 
phase (see Fig. 11), resulting from the presence of a structural defect leading to oxygen vacancies in the material. In 
these signals, initially attributed to the presence of divalent cobalt ions, are in fact due to 
ions in the vicinity of paramagnetic trivalent cobalt ions in a quite unusual spin state.19
30 Indeed, this structural defect, induced by a nominal Li excess during the synthesis, leads to the presence of intermediate spinning 
ions (labeled 
which are responsible for three characteristic first/second neighbor effect NMR signals at 185 ppm (one 
first neighbor) and 
and 
(one and two 
second neighbor, respectively). Furthermore, the two shoulders on the 0 ppm signal (3 and 
are considered to be due to 
ions in an oxygen environment perturbed by the oxygen vacancy, leading to various mechanisms of transfer of the hyperfine interaction from the 
ion close to the oxygen vacancy. These shoulders are clearly apparent in the spectra of the 
19 However, since the 
and 
signals must get mixed up with the 
and 
signals and because the first spinning side band must hide the possible presence of a signal at 185 ppm, one cannot clearly observe these signals on the 15 kHz MAS spectra.
Figure 11. Expansion of the central signal of 10 kHz 
MAS NMR spectra for the various 
phases 
1.10; 
0.03).
As concerns the first-neighbor effect of 
ions in 
the first spinning side band at 
is clearly superposed on another band in the case of the lithium overstoichiometric material (Fig. 8). Therefore, in order to better observe the 150/250 ppm range, we have performed MAS NMR measurements using a 10 kHz rotation speed to shift the spinning side band without modifying the isotropic shifts. The NMR spectra, shown in Fig. 10 and compared to that of the 
phase clearly show the signal associated with 
at 185 ppm for the 
material. The expansion of the central signal of the various compounds (Fig. 11) confirms the presence of both 
and 
ions in the lithium overstoichiometric 
phase, the contribution of each type of paramagnetic ion adding in the 
range and leading to a loss in the central signal definition.
Figure 10. 10 kHz 
MAS NMR spectra for the various 
phases 
1.10; 
0.03), 
side bands).
Conclusions
The galvanostatic measurements and XRD results allow us to point out some conclusions on the influence of Ni doping and Li nonstoichiometry in 
This study clearly allows us to separate the relative effects of Ni-doping and lithium overstoichiometry which was not considered by Reimers et al.18 We show here that 3% of Ni ions substituted for Co do not make the monoclinic distortion disappear.
Before any conclusion, we have to consider what are the driving forces for the phase transitions in the pure 
system. On the one hand, it is now totally admitted that the biphasic domain for 
results from a brutal change in the electronic properties of the 
system as it undergoes a semiconductor-to-metal transition at the macroscopic scale.7 On the other hand, the monoclinic distortion for 
is considered to have a structural origin related to the establishment of a Li/vacancy ordering in the interslab layer.4 Indeed, a recent TEM study clearly showed, for the first time, the Li/vacancy ordering in the monoclinic 
phase.31 In the 
system 
the presence of structural defects (oxygen vacancy associated to 
ions) induces enough perturbations in the slab to prevent any Li/vacancy ordering in the interslab space.30 From the electronic properties point of view, even if the semiconductor-to-metal transition is observed upon lithium deintercalation, this transition is not associated with a phase separation since the defects are trapped in the lattice and cannot move at room temperature.
In the case of Ni-doped materials synthesized without any lithium excess, the experimental results show first that 6% of Ni ions substituted for Co do not create sufficient perturbations in the 
slab ordering to prevent the lithium/vacancy ordering in the insterlab and the resulting monoclinic distortion because of the close ionic radii of the 
and 
ions.23 When 10% of Ni are substituted for Co, the effect becomes significant and the trace of the monoclinic distortion becomes weak. As a possible explanation of the discrepancy between our and Reimers's results on that point, we suggest that the Ni-substituted samples studied by this author may have been slightly Li-overstoichiometric.
Second, from the electronic properties point of view, the Ni ions are trapped in the lattice and are an obstacle to electron delocalization. Therefore, the expected phase separation associated with the semiconductor-to-metal transition cannot occur; a solid solution domain is observed upon lithium deintercalation. The behavior resulting from the presence of the nickel ions is very similar to that due to the structural defects associated with lithium overstoichiometry in the 
phases.19
30
When both type of defects are simultaneously present (Ni ions and defects related to lithium overstoichiometry) as shown by NMR measurements conducted on the 
phase, all the phase transitions disappear.
Finally, the NMR measurements show that the doping effects of nickel and lithium are cumulative and independent. Indeed, in the 
phase, both 
ions (associated to an oxygen vacancy) and 
ions coexist in the diamagnetic 
lattice.
Acknowledgments
The authors thank I. Saadoune, A. Audemer, L. Gautier, and J. Scoyer for fruitful discussions, and Umicore, ANRT, and Région Aquitaine for financial support.
Institut de Chimie de la Matière Condensée de Bordeaux assisted in meeting the publication costs of this article.