Abstract
was investigated to understand the effect of replacement of the cobalt by aluminum on the structural and electrochemical properties. In situ X-ray absorption spectroscopy (XAS) was performed, utilizing a novel in situ electrochemical cell, specifically designed for long-term X-ray experiments. The cell was cycled at a moderate rate through a typical Li-ion battery operating voltage range. 
XAS measurements were performed at different states of charge (SOC) during cycling, at the Ni, Co, and the Mn edges, revealing details about the response of the cathode to Li insertion and extraction processes. The extended X-ray absorption fine structure (EXAFS) region of the spectra revealed the changes of bond distance and coordination number of Ni, Co, and Mn absorbers as a function of the SOC of the material. The oxidation states of the transition metals in the system are 
, 
, and 
in the as-made material (fully discharged), while during charging the 
is oxidized to 
through an intermediate stage of 
, 
is oxidized toward 
, and Mn was found to be electrochemically inactive and remained as 
. The EXAFS results during cycling show that the Ni–O changes the most, followed by Co–O, and Mn–O varies the least. These measurements on this cathode material confirmed that the material retains its symmetry and good structural short-range order leading to the superior cycling reported earlier.
Export citation and abstract BibTeX RIS
Environmentally friendly and economical lithium battery electrode materials with higher capacity and good cycling stability are in high demand. With the emergence of portable telecommunications, computer equipment, and, ultimately, hybrid electric vehicles, there has been a great demand for less expensive batteries with longer lifetimes and smaller sizes and weights. 
has been the backbone for commercial use in Li-ion batteries since their inception by Sony due to its ease of production, stable electrochemical cycling, and acceptable specific capacity. The relatively high cost and toxicity of cobalt and the quest for materials with higher capacities has led to the study of possible alternatives, though many have limitations precluding their widespread use. 
is thermally unstable at high states of charge.1 Variants of 
exhibit good structural stability on overcharge but have low capacity and exhibit high solubility of 
in the electrolyte at elevated temperatures above 
.2 Layered systems such as 
have attracted attention as a cathode for lithium-ion batteries especially for hybrid electric vehicle applications.2 Compared to 
, this system proved to possess enhanced electrochemical performance at a lower projected cost.3 Other layered materials under consideration as cathode materials for Li-ion batteries include NMCs, which have the general formula 
(Ref. 3), where y is most commonly equal to 
. There have recently been attempts to further reduce the Co content (for example, 
).4 This has met with some success, but the rate capability of the electrodes tends to decrease as the Ni content is raised due to increased antisite mixing. However, there is a great deal of interest in the synthesis and characterization of layered compounds with reduced Co content.4–7
Recently, 
, 
(Ref. 8), and 
, 
(Ref. 9), compounds were studied as cathode materials. For low values of y, aluminum substitution appears to improve the rate capability and cycling stability compared to the baseline materials. Other studies on Al substitution of layered oxide structures suggest that the thermal abuse tolerance is improved.10 These results motivated us to investigate specific Al-substituted materials using in situ X-ray absorption spectroscopy (XAS) in order to fully characterize the structural properties of this material. An ultimate goal is to allow design of low Co content layered oxides with improved electrochemical properties.
There has been a continuous effort to study the detailed changes that occur in various electrode materials during the charge–discharge process. Real-time tracking of the structure and valence changes during delithiation, relithiation, and prolonged cycling of layered cathode materials may have predictive value for determining the intrinsic electrochemical characteristics and can be achieved by employing an in situ electrochemical cell. XAS, accompanied by simultaneous electrochemical measurements, provides valuable and important information about the relationship between the structure and the electrochemical properties of the electrode materials,11 information that is not always accessible utilizing ex situ experimental measurements.
In this investigation, we present a comprehensive transmission XAS study to investigate 
electrodes (as used in Li-ion batteries) at various states of charge. Information about the oxidation state of the investigated transition metal elements (in this case, Mn, Co, and Ni) and their electronic configuration was obtained from the X-ray absorption near-edge spectroscopy (XANES) region of the K-absorption edges, whereas the extended X-ray absorption fine structure (EXAFS) region was used to probe the structure around the X-ray absorbing atoms.
Experimental Procedures
powder was synthesized and electrodes were prepared as previously described.9 The specifications of the electrode are given in Table I, where the specific capacities were calculated based on the valence changes and the amounts of the elements in the compound. Phase purity was assessed by X-ray powder diffraction using a Phillips X'Pert diffractometer with an X'celerator detector and Cu 
radiation. The experiments were performed using a complete 
half-cell. A special electrochemical in situ XAS cell was used for these experimental investigations; for a complete detailed design and assembly, the reader is referred to Deb et al.12 A sheet of porous polypropylene membrane (Celgard 3400) was utilized as a separator and 1M 
in ethylene 
carbonate (1:1 volume, Merck, LP40) as the electrolyte. A Li metal foil cut into a circular disk (diameter 
) was used as the counter electrode. The charge–discharge cycling was performed at room temperature at the beam line with a potentiostat/galvanostat system (Princeton Applied Research, Model Versa) under constant current control. In situ XAS measurements performed during charging and discharging were carried out at various states of charge of the electrode. Charging and discharging were performed at a constant current density of 
between 1.0 and 
, which corresponds to about the C/25 rate, assuming that all of the Li could be extracted. The measurements performed at different points along the charging cycle are labeled in Fig. 1. The charging was paused during the XAS measurements to allow the electrode systems to stabilize. The spectra presented here were obtained during the charging process at different states of charge (to 
versus Li). For the measurements with the model compounds shown here for comparison, the samples were diluted with boron nitride (BN) (samples were mixed in a 1:10 ratio with BN using a mortar and pestle) and 
pellets were prepared. The sample pellets were then finally loaded on aluminum holders using Kapton adhesive foil on both sides of the sample.
Table I. Electrode properties and experimental conditions.
| Formula weight: 94.5815 |
|---|---|
 to  specific capacity |
|
 to  specific capacity |
|
| Total specific capacity |
|
| Active material weight |
|
| Electrode area |
|
| Active material loading |
|
| Total current |
|
Figure 1. Charging curve of the 
electrode. Labels shown A to O are the points where the measurements were performed during the charging cycle corresponding to 
(A), 0.96 (B), 0.92 (C), 0.88 (D), 0.84 (E), 0.80 (F), 0.76 (G), 0.67 (H), 0.55 (I), 0.49 (J), 0.38 (K), 0.31 (L), 0.27 (M), 0.23 (N) and 0.19 (O).
The XAS measurements were performed in the transmission mode at the bending magnet beamline station D of the DND-CAT (Sector 5), at the Advanced Photon Source, using a water cooled Si(111) double crystal monochromator, and the energy resolution of the monochromatic beam was determined to be 
. A beam size of about 
was used for the beam to pass easily through the in situ cell X-ray window resulting in an incident photon flux of 
. For both Mn and Ni K-absorption edges, the monochromator was scanned from 200 below to 
above the K-absorption edge, while the Co K-absorption edge data range was limited to a maximum of 
at the onset of the Ni K edge. The EXAFS data presented were analyzed using a combination of EXAFSPAK software package13 and ATHENA.14 The resulting 
function was weighted with 
to account for the damping of oscillations with increasing k. The radial structure functions presented here were obtained by Fourier transformation of 
using a k range of 
for Mn and Ni and 
for Co.
Results and Discussion
crystallizes in the R-3m space group and exhibits a well defined 
structure as does the parent material 
.9 Al substitution causes the a unit cell parameter to decrease by 0.1% and the c unit cell parameter to increase by 0.2% compared to the parent material. There is also an increase in the degree of antisite mixing (Ni in the 3a or lithium sites) from 6.6% to 7.3% and a slight increase in the lithium slab spacing, which results in improved lamellarity upon substitution with Al.9
In general, the shape of the K-edge XANES of the transition metal oxides provides unique information about the site symmetry and the nature of the bonding with surrounding ligands, while the threshold energy position of the absorption edge provides information about the oxidation state of the probed atom. To extract the information of the initial oxidation state of the transition metals in this material and elucidate the charge compensation mechanism in this system, Mn, Co, and Ni K-edge XAS experiments were carried out during the charge/discharge cycle at a constant current density of 
. The specific capacity was calculated from the elapsed time, current, and the mass of the active material in the cathode, assuming that all the current passed was due to the Li deintercalation. The comparison of the XANES region before charging [
(point A) in Fig. 1] with that of the model compounds is shown in Fig. 2. The comparison [Fig. 2a] reveals that the edge energy is essentially the same as that of 
, indicating that the oxidation state of Mn in this material at the start of the charging cycle 
is nearly all tetravalent 
. Figure 2b indicates that the XANES spectrum for the Co K edge of the sample is identical to that for 
, indicating that Co in the sample at the beginning of the charging is 
. Figure 2c shows the comparison of Ni K-edge XANES spectrum with that of the model compounds of nickel (II) oxide 
and layered 
. At 
, the XANES spectrum is similar to that of the nickel(II) oxide, indicating that the Ni in this system is divalent 
, whereas the XANES spectrum of the layered 
is similar to that observed at 
, confirming that Ni is in the trivalent state 
at 
.
Figure 2. Normalized XANES comparison for 
of (a) Mn K edge at 
with the model compound 
and 
; (b) Co K edge observed at 
with the model compound 
, and (c) Ni K edge at 
and 
with the model compounds nickel (II) oxide 
and 
, respectively.
XANES spectra for selected points during the charging process for Mn, Co, and Ni edges are shown in Fig. 3. For the Mn K-edge XANES [Fig. 3a], during charge, the edge position of the scans at different states of charge did not exhibit any significant edge shift to higher energies, though it does exhibit some changes in regard to the shape of the edge due to the change in the local environment of the Mn in the system as Li is deintercalated. The edge does not exhibit any rigid shift to higher energies, however, suggesting that the Mn oxidation state remains unchanged during the charging procedure; i.e., the 
atom is electrochemically inactive. The Co K edge [Fig. 3b] shows a progression of the entire pattern from lower energy to higher energy as a function of the decreased Li content (at the different charge states), indicating the oxidation of 
toward 
.
Figure 3. Normalized XANES spectra at different states of charge of (a) Mn K edge, (b) Co K edge, and (c) Ni K edge, respectively. The insets show the nature of the pre-edge peaks of the respective XAS spectra; (d) Comparison of the linear combination (LC) best fit of the 
and 
spectra (with a composition of 62%, of 
and 38%, of 
) with the 
spectra.
XANES spectra for the Ni K edge shown in Fig. 3c indicate that the Ni ion undergoes a two-stage change during the charging cycle, where the first stage ranges from 
to 
and the second stage is from 
to 
. Beyond 
(Fig. 4), the Ni K-edge XANES did not move significantly to a higher energy. This two-stage reaction observation may be attributed to the two one-electron reactions of the Ni ions (
and 
) during charging. Similar observations of the redox reaction have been reported previously by Koyama et al.15 using ab initio calculations for 
consisting of 
, 
, and 
in the ranges of 
, 
, and 
, respectively. During discharge, these processes are reversed electrochemically: Li is inserted in the lattice, 
is reduced to 
, 
is reduced to 
, and M–O (metal–oxygen) and M–M (metal–metal) bond lengths return to their original values. For the purposes of this discussion, the results during charge will be discussed hereafter.
Figure 4. Plot of the white line energy shift vs the x in 
, for the Ni K edge. The filled symbol (●) represents the data during the charge, while the empty symbol (o) represents the data during discharge.
Figure 5. Mn-K edge 
-weighted Fourier transform at selected Li content (x) during the charge cycle (k 
). M represents the transition metal atom.
Figure 6. Co–K edge 
-weighted Fourier transform at selected Li content (x) during the charge cycle (k 
). M represents the transition metal atom.
Figure 7. Ni–K edge 
-weighted Fourier transform at selected Li content (x) during the charge cycle (k 
). M represents the transition metal atom.
To make sure of the presence of the 
species at 
, we see that the absorption edges for Ni in the various states of charge cannot be represented by linear combinations of only 
and 
spectra. Figure 3d shows a comparison of the spectrum calculated as a linear combination of the two species of 
and 
spectra and the spectrum for 
. The important areas of mismatch are the region near 
(just before the absorption maximum), the absorption maximum, and regions near 8358 and 
(after the absorption maximum). The mismatch in the absorption maximum is an indication of the presence of 
instead of a combination of 
and 
. This evidence indicates that 
is an intermediate state for the Ni during oxidation/reduction cycling.
The preedge peaks of the respective XAS spectra are also shown in Fig. 3. These give additional information on the nature of the electronic states. For most transition metal elements, the preedge peaks that occur well below the main edge (
below) are assigned to transitions to empty states with d-like character 
transitions,16, 17 where n represents the initial number of d electrons and 
includes the excited electron in the final state, which includes the effect of the core hole. These 
transitions are directly allowed through a very weak quadrupole transition18 or allowed via an admixture of 3d and 4p states.19 The weak preedge in the Mn-absorption (Fig. 3a, peaks A and 
near 6542 and 
) is the formally electric dipole-forbidden transition of a 1s electron to an unoccupied 3d orbital of a high-spin 
ion, which is partially allowed because of the pure electric quadrupole coupling and/or the 3d–4p (or Mn 3d–O 2p) orbital mixing. The presence of weak preedge intensity is indicative of octahedral coordination as opposed to tetrahedral coordination which would have resulted in a stronger preedge intensity.20 The two peaks (in the preedge region) for Mn are discernable, as Mn is in the tetravalent state 
, and a single peak structure in the preedge region is characteristic of trivalent 
Mn compounds. The latter single peak characteristic for trivalent manganese systems can be interpreted as splitting of the 
and 
energy levels modified by the Jahn–Teller distortion, though this is not observed in the studied system.21
For the Co preedge (Fig. 3b), the Co 
transitions are not well structured here, but two peaks (C and 
) are still resolvable, as well as a weak feature (C and 
) above 
with indication of an energy splitting in between C and 
is assigned to the 
transition, which is dipole and symmetry forbidden, but occurs, albeit very weakly, due to the mixing with the oxygen p orbitals and quadrupole transitions and 
transition which is dipole forbidden, and hence much weaker than the symmetry- and dipole-allowed 
main transition. The two transitions present here indicate at least a partial high-spin configuration, since the peaks are not clearly split, which may be due to the delocalized electrons in the band structure or an overlapping mixture of low-spin 
, intermediate 
, and high-spin 
states. The shoulder (B) around 
is the shake-down process, which occurs at a lower energy, since the binding energy decreases due to the core-hole screening effect in Co by an electron transferred in from the ligand.22, 23
Since the intensity of the preedge peak 
(D, in Fig. 3e) can be used as an indicator for the geometry of octahedral complexes, the presence of very weak Ni preedge during the cycling signifies that Ni atoms are not highly symmetrical, and the weak preedge arises due to the possible distortion of the octahedral 3b site in the rhombohedral R-3m space group, which results in the removal of the inversion symmetry.
Figure 4 is a quantitative picture of how the redox process progresses. It shows a plot of the relative white line peak position as a function of amount of Li present (x), in contrast to the edge position, as defined by the energy at half of the edge step for the K edge, observed during charge and discharge. The Co and the Ni peak positions change by 
and 
, respectively, as the lithium content changes from the fully discharged state to the charged (
versus Li) state. Though there is still a controversy regarding charge compensation via oxygen in these systems, the results here show clearly the contribution of the Ni and the Co ion to the charge compensation process. In a similar system, Yoon et al.24 reported that oxygen oxidation was part of the charge compensation process. In contrast with Yoon et al. , who reported only small shifts in metal XANES above 50% state of charge, we observe continuous shifting of both the metal XANES and the metal–oxygen bond lengths. There is thus no need to invoke oxygen oxidation; however, our data cannot rule out the possibility that a minor component of the charge compensation results from oxygen oxidation.
The local structure of the Mn, Co, and Ni atoms in the system was obtained from EXAFS measurements. The backgrounds were subtracted by extrapolating a Victoreen-type function from the preedge region, and EXAFS oscillations 
were extracted using cubic spline baseline functions. In all cases, the Li contribution to the EXAFS was ignored because of the low backscattering ability of the lithium. The FT (Fourier transform) of the data at selected SOCs is shown in Fig. 5, 6 and 7. The first peak (
, labeled S) in the FT is attributed to the M–O interactions, while the second peak (
, labeled T) is due to the M–M interactions. During the cycling process, the Mn–O distances remain nearly unchanged in contrast to the gradual dwindling of the Co–O distances accompanied by significant decreases in the Ni–O distances (Fig. 8). This change can be related to the change in the valence state in the transition metals where in an octahedral crystal field, the d orbitals split into a triply degenerate orbital set 
and a higher-energy doubly degenerate orbital set 
, which leads to an outermost electronic configuration for 
as 
. On the other hand, the outermost configuration for 
can be represented as 
. In a simple molecular-orbital picture, the 
orbitals are metal–ligand antibonding while the 
orbitals are metal–ligand nonbonding. As a consequence, changes in 
occupancy have a much larger effect on M–O distance than do changes in 
occupancy. Hence during the oxidation reaction of the charging cycle, the energy change for 
(or 
) would be larger (than for Co) as the change takes place between the lower 
and the higher 
sets, inducing a larger change in ionic radius from 
to 
.
Figure 8. Metal-oxygen first shell coordination bond length changes during 
cell cycling. 
to 0.19 represent first charging cycle and 
to 1.0 represent first discharge cycle.
From Fig. 6 and 7, it can be seen that the amplitude of the Co–O peak remains almost unchanged while the Ni–O peak increases during charging. These observations can best be explained if we consider the simpler systems of 
and 
. For 
studied earlier,24 Ni is surrounded by oxygen atoms at two different distances and the distortion of the octahedral coordination is consistent with the presence of the Jahn–Teller effect due to the 
in the low-spin state (in addition to the distortion inherent in R-3m symmetry). On delithiation of 
, the increase of the Ni–O peak amplitude is ascribed to the change in the oxidation from 
to 
, and since 
is not a Jahn–Teller active ion, the gradual change of 
to 
during charging diminishes the effect of the Jahn–Teller distortion and results in an increased amplitude of the Ni–O peak. Here in this system based on the above observation, we can explain that during delithiation from 
to 
(when Ni is oxidized from 
to 
), a diminution of the Jahn–Teller effect results in increased amplitude of the Ni–O peak. On the contrary, from 
to 
(when Ni is oxidized from 
to 
), increase of the Jahn–Teller effect results in a decrease of amplitude of the Ni–O peak. For the 
delithiation there is no change in the amplitude of the Co–O peak due to the absence of the Jahn–Teller effect for 
. Thus similarly for this system, during charging, Co changes from 
to 
neither of which are Jahn–Teller active ions (except that low-spin 
will be Jahn–Teller active, albeit with a small distortion), and thus a change in amplitude of the Co–O peak is not observed.
Structural parameters of the Mn, Co, and Ni absorbers were analyzed based on a two shell model, where bond distances (R) and the Debye–Waller factors 
were left as free parameters and the coordination number for Mn–O, Co–O, and Ni–O was kept fixed to the crystallographic value of 6 (Fig. 8). During cycling, the change in the Ni–O bond length is the largest 
, followed by Co–O 
, and the Mn–O change is negligible 
. The Ni–O bond length at the start of charge 
is 
, while at 
it is 
, and finally at 
it is 
. These Ni–O distances are consistent with octahedrally coordinated systems where the 
bond length is about 
, 
exhibiting a static Jahn–Teller distortion shows an average 
bond at 
, and finally the 
bond length is 
.25–29 Thus these fit results confirm that the average oxidation state of Ni ion at the start of the charge is 
, whereas at 
it is 
, and finally at the end of the charge, Ni ion is close to 
. The significant change in the Ni–O bond length can be explained by the change in oxidation state of 
since the ionic radius of 
is larger than that of 
. For Co, during charging, the observed changes in the Co–O distance 
are attributed to the oxidation of 
to 
, and the predicted decrease in Co–O distance between 
and 
is 
.29 Using Ref. 30, it can be estimated that between 
and 50% of the Co is oxidized during the charging cycle. Finally, for the metal–metal (Mn–M, Co–M, and Ni–M) interaction fits, we have taken two approaches. In one approach, we have fitted the experimental profile, taking Co as the scatterer (Fig. 9), but because this scattering atom can either be Mn, Co, or Ni (though not Al because it is too light), we also did fits with the Mn and Ni atoms. This was done to understand the sensitivity of the data, taking Co as the scatterer. In this approach we found that if instead of Co it was fitted with Ni or Mn, the corresponding bond distances shown in Fig. 9 during the cycling change by 
for Ni and 
for Mn. In the other approach, which actually gives similar result as shown in Fig. 9, we fitted the M–M peak by taking into consideration the ratio of Mn, Co, and Ni present initially at the beginning of the charging cycle. Regardless of the absolute accuracy of the apparent M–M distance, the changes in distance (Fig. 9) should be reliable.
Figure 9. Metal-metal first shell coordination bond length changes during 
cell cycling. 
to 0.19 represent first charging cycle and 
to 1.0 represent first discharge cycle.
Conclusions
In situ XAS characterizations for Al-substituted 
have been performed during the first charge and discharge process, providing us with an excellent tool for analyzing the changes that occur when Li is cycled in and out of the layered lattice. During cycling, Mn is electrochemically inactive and remains tetravalent 
, whereas divalent nickel 
is oxidized mostly to tetravalent nickel 
, passing through an intermediate stage of trivalent nickel 
, and about 
of 
is converted to 
during charge. Bond length changes upon delithiation and relithiation of this system appear to be very reversible, which may explain the excellent cycling stability that has been previously observed.
Acknowledgments
Work was performed at the DND-CAT beam line, which is supported by the E. I. DuPont de Nemours and Co., the Dow Chemical Company, the U.S. National Science Foundation through Grant no. DMR-9304725 , and the State of Illinois through the Department of Commerce and the Board of Higher Education Grant no. IBHE HECA NWU 96. This work was supported by the Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Vehicle Technologies of the U.S. Department of Energy under Contract no. DE-AC02-05CH11231 .
University of Michigan assisted in meeting the publication costs of this article.