High Capacity Li-Rich Positive Electrode Materials with Reduced First

Article pubs.acs.org/cm

High Capacity Li-Rich Positive Electrode Materials with Reduced First-Cycle Irreversible Capacity Loss Ramesh Shunmugasundaram,† Rajalakshmi Senthil Arumugam,‡ and J. R. Dahn*,†,‡ †

Department of Chemistry and ‡Department of Physics and Atmospheric Science, Dalhousie University, Halifax, Nova Scotia, Canada B3H 4R2 S Supporting Information *

ABSTRACT: The first charge−discharge cycling behaviors of two sets of Li−Ni−Mn−Co type positive electrode materials were compared. The samples in each set have similar Ni−Mn−Co ratios but different Lito-total metal ratio (Li/M). The samples that were Li-rich with a Li[LixM(1−x)]O2 structure showed a typical 4.5 V “oxygen loss” plateau and a typical irreversible capacity loss near 25%. Surprisingly, other samples with lower Li/M ratios that still exhibited a 4.5 V “oxygen loss” plateau exhibited an irreversible capacity loss as low as 4.0% of their first charge capacity. XRD analysis revealed that all samples were single-phase layered oxides. A separate and a detailed XRD analysis combined with dQ/dV analysis showed that the reduced irreversible capacity loss was not caused by the admixture of a spinel phase. ICP-OES results and the oxidation state versus atomic occupancy rules suggested the presence of metal site vacancies in the pristine materials with low IRC, which were confirmed by densities measured with a helium pycnometer. The results presented here show that the small irreversible capacity is a consequence of (a) metal site vacancies, leading to Li[□qM(1−q)]O2 structures, where □ is a metal site vacancy, which leads to (b) no Li atoms in the transition metal layer. These materials still have Li/M > 1, so they are “Li-rich”, but they are “traditional layered materials” with no Li in the transition metal layer. This study identifies a new route for fabricating high capacity Li-rich positive electrode materials with small irreversible capacity loss.

■

reaction.11 A similar approach to acid leaching is the chemical treatment of the pristine Li-rich material using reagents such as Na2S2O8.12 In this method, one claim for the observed low IRC was the formation of a spinel phase on the surface of the active material. Blending Li-rich materials with different materials such as V2O513 or LiV3O814 that can intercalate lithium near 3.5 V reduces IRC but at the expense of lower energy density. Surface coating15 methods have been shown to be useful to reduce the IRC in an unknown manner and at the apparent expense of overall capacity. All the above strategies are based on postsynthesis chemical treatments of the pristine materials. Such methods not only require harsh chemical conditions but also increase cost. By contrast, this study focuses on making new materials with a typical synthesis procedure that is applicable for industrial scale-up. The mobility of the transition metal atoms during the first electrochemical cycling and the subsequent interlayer mixing between atoms in the TM layer and the Li layer seems to be the major cause of the IRC loss in the Li-rich materials.16 It is generally believed that lithium vacancies created during the first charge are occupied by the TM atoms through surface to bulk

INTRODUCTION Li-rich layered-oxide positive electrode materials exhibit specific capacity as large as 250 mAh/g,1−3 and thus they are potential candidates for high energy density applications such as electric vehicles.4 However, they suffer from three major issues, which hinder their commercial application. First, they show a large irreversible capacity loss (IRC)5 in the first charge−discharge cycle. Second, they display very poor rate capability.6 Finally, they exhibit a significant voltage fade7 during charge−discharge cycling and hence their energy density drops even though their specific capacity does not. The focus of this article is on finding Li-rich materials with reduced first cycle IRC. IRC loss in positive electrode materials is a measure of the proportion of Li ions that are unable to reintercalate back into the host structure during the first discharge at reasonable rates.8 Researchers generally believe that a large IRC loss in Li-rich materials is associated with the irreversible oxygen loss occurring around 4.5 V during the first charge.9,10 There have been several reports with a focus on reducing the oxygen release during the first charge. One general approach is to partially delithiate the pristine material using chemical methods in such a way to limit the oxygen loss while traversing along the 4.5 V plateau. Chemical delithiation on Li-rich materials has been attempted by using acid leaching but with a sacrifice in structural stability associated with a H+ /Li+ exchange © XXXX American Chemical Society

Received: September 25, 2014 Revised: January 10, 2015

A

DOI: 10.1021/cm504583y Chem. Mater. XXXX, XXX, XXX−XXX

Article

Chemistry of Materials diffusion,17 which appears to be connected to the IRC. Hence, mitigating TM diffusion and interlayer mixing could be the key to reducing IRC. As a proof of concept, van Bommel et al.5 and Ito et al.18 have reported stepwise first cycle charging protocol and were successful in reducing the IRC of Li[Li0.2Ni0.2Mn0.6]O2.5 A recent report from the Tarascon group showed a reduced IRC loss (only 10 to 12%) in Li 2 Ru 0.5 Sn 0.5 O 3 or Li1.333Ru0.333Sn0.333O2.19 Their assertion for the cause of the low IRC was that the Sn atoms make the M−O bonding “more flexible” and essentially minimize the oxygen release. Alternatively the interlayer mixing may have simply been prevented or hindered by using a larger ion (Sn4+) in the place of the 3d metals. However, such materials will have limited commercial value. In this work, Li-rich layered oxides were prepared from mixed transition metal carbonate precursors reacted with Li2CO3 at elevated temperatures. A systematic study of the impact of lithium content during synthesis on the irreversible capacity of materials made from Ni(II)0.167Mn(II)0.5Co(II)0.333CO3 and Ni(II)0.2Mn(II)0.5Co(II)0.3CO3 precursors and Li2CO3 was made. Through an initially serendipitous experiment, it was found that the samples made with less Li2CO3 than targeted to make samples with transition metals in their expected oxidation states in Li-rich layered oxides1,9 led to materials with IRC as low as 4.0%. Studies of these materials using X-ray diffraction, elemental analysis, He pycnometry, and electrochemical methods are reported here. The conclusion is that the materials with low IRC are “Li-rich”, in that the Li/M ratio is greater than 1, but they are “traditional layer compounds” with no Li atoms in the transition metal layer. Instead, there are vacancies in the transition metal layer, and the composition of these materials is written as Li[□qM1−q]O2, where □ is a metal atom vacancy.

■

in a 3:1 reagent grade HCl:HNO3 (aqua regia) which was then diluted to 50 mL prior to measurement. For each sample, elemental compositions were reported as mass fractions of Li, Mn, Ni, and Co relative to the total solution mass with a 2% relative error for each mass fraction. From these results, the atomic ratios of Li:Ni:Mn:Co were calculated. The accuracy of the atomic ratios was 2% in each element. A powder X-ray diffraction pattern for each sample was collected using a Siemens D5000 diffractometer equipped with a copper target X-ray tube and a diffracted beam monochromator. Data was collected between scattering angles of 10° and 90°. An X-ray pattern fitting program called “Rietica” was used for the Rietveld refinement.20 The density of the samples was measured using an AccuPyc II 1340 gas displacement pycnometer using He gas (ultra-high pure 99.999%). The regulator pressure of the helium tank, the purge fill pressure, and the cycle fill pressure were set at 22.000 psig, 19.500 psig, and 19.500 psig, respectively. Before measuring the density of each material, the pycnometer was calibrated with the supplied standard of known volume. Prior to each individual measurement, each sample was heated to about 400 °C, cooled to room temperature, and transferred to the pycnometer with very minimum exposure to air in order to desorb moisture prior to measurment. The BET surface area of the samples was measured using a Micromeritics Flowsorb II 2300 surface area analyzer using liquid nitrogen. Each sample was degassed at 200 °C for about 1 h to remove the residual moisture prior to the actual surface area measurement. Scanning electron microscope images were recorded using a Hitachi S4700 field emission scanning electron microscope with an accelerating voltage of 10 kV and an emission current of 15 μA. Working electrodes were made from the synthesized powders as described in detail by Marks et al.21 About 90 wt % of the positive electrode material was mixed with 5 wt % of Super C45 carbon black (commercially available from TIMCAL), 5 wt % of polyvinylidene difluoride (PVDF) binder (commercially available from ARKEMA), and the appropriate quantity of N-methyl pyrrolidone (NMP) solvent. A mixer (Mazerustar) was used to create a uniform slurry, and the freshly prepared slurry was spread into a film on Al foil using a notch bar spreader. After drying for at least 3 h at 120 °C to completely remove the NMP, the dried electrode in the form of a film was calendered at 200 bar. The compressed electrode sheet was punched into several circular disks, which were eventually used as working electrodes in the coin cells. Coin-cell assembly was carried out in an argon-filled glovebox. 1 M lithium hexafluorophosphate (LiPF6) in 1:2 ethylene carbonate (EC)/ diethyl carbonate (DEC) (commercially available from BASF) was used as the electrolyte. All the constructed electrochemical coin cells were galvanostatically cycled using a specific current of 10 mA/g at 30 °C using a computer-controlled charger system (Maccor 4000). The first charge−discharge cycle was between 4.8 and 2.0 V, and the subsequent cycles were between 4.6 and 2.8 V using the same specific current and temperature.

EXPERIMENTAL SECTION

The desired precursors Ni(II)0.167Mn(II)0.5Co(II)0.333CO3 and Ni(II)0.2Mn(II)0.5Co(II)0.3CO3 were synthesized using a coprecipitation method. An aqueous solution of 0.1 M NH4OH, which served as a coordinating agent for the transition metals, was used as the starting reaction medium. A 2 M aqueous solution comprised of the appropriate proportions of metal sulfate solutions and an equimolar solution of sodium carbonate (Na2CO3) were mixed together in a continuously stirred tank reactor (CSTR) at a flow rate of 0.33 mL/ min. The temperature was controlled at 60 °C, and the pH was controlled at 8 by adding appropriate amounts of acid (H2SO4) or base (NaOH). After the completion of the coprecipitation reaction, the resulting suspension was collected and washed several (5) times with distilled water. Then it was filtered, and the resulting wet precipitate was dried in an oven at 100−120 °C in air for about 12 h. To synthesize the Li-rich layered lithium−transition-metal oxide, the required amounts of Ni(II)0.167Mn(II)0.5Co(II)0.333CO3 or Ni(II)0.2Mn(II)0.5Co(II)0.3CO3 precursor and Li2CO3 were weighed accurately, mixed, and ground well using a mortar and pestle. A 5 wt % excess of Li2CO3 was used for compensating Li loss due to evaporation during high temperature sintering. The mixed powders were loaded in an alumina crucible and calcined in air to yield the product. The following heating and cooling profile was used during the firing: step 1, heating from room temperature to 400 °C at 10 °C per minute and hold for 2 h; step 2, heating from 400 to 900 °C at 10 °C per minute and hold for 12 h; and step 3, cooling down to room temperature at 2 °C per minute. The hold at 400 °C was maintained to facilitate the complete removal of CO2 from the transition metal carbonates. The elemental composition of the lithiated metal oxide powders was obtained using inductively coupled plasma optical emission spectroscopy (ICP-OES). Approximately 10 mg of each sample was dissolved

■

RESULTS AND DISCUSSION Table 1 shows the nominal (target) compositions of all the studied samples (A1 to A3 and B1 to B3) and their respective precursors as well as their first cycle IRC values. The details of the actual compositions observed from the ICP-OES technique and their implications will be presented and explained in the forthcoming discussions. From the nominal compositions given in Table 1, the total occupancy of the metal sites versus the oxygen sites is perfectly balanced in the LiMO2 structural motif. However, the total charge balance between cations and anions is not equal in samples A2, A3, B2, and B3 assuming Li+, Ni2+, Mn4+, Co3+, and O,2−24 because the Li/TM ratio was reduced gradually from the first member to the last member in each of the series. For example, the Li/TM ratios go lower and lower from A1 to A2 and to A3. In any event this particular change in the Li/TM ratio between the samples has a profound effect on B


Article

Chemistry of Materials

from the A and B series. The IRCs observed for the first member of each series (A1 and B1), which have the stoichiometrically balanced amount of Li, are above at least 20% of their first charge capacity as expected for a typical Lirich material such as Li[Li0.2Ni0.2Mn0.6]O2.5 Surprisingly and beneficially, the low-Li members (e.g., A2, B3) show IRCs as low as 6.5%. Figure 1 shows that the sloping region capacity of all the samples in each of the series is very similar. However, the first member in each series exhibits the highest plateau region capacity whereas the last member has the lowest in accordance with their respective Li contents. Thus, an overall lower Li content in the pristine structure, as in the case of A2 and A3 or B2 and B3, could result in a relatively smaller amount of oxygen release or oxygen oxidation during the ∼4.5 V plateau. In any event, adjusting the Li/M ratio dramatically affects the IRC. Figures 2 and S1 (Supporting Information) show the measured and calculated XRD patterns of the A and B series samples, respectively. The calculated patterns were made using the Rietveld method with appropriate crystallographic parameters. Table 2 shows all the structural parameters obtained from Rietveld refinement including lattice parameters, a and c. The first members of the two series (A1 and B1) were pure singlephase layered materials with the O3 structure, analogous to αNaFeO2. The Rietveld refinements have been made in the space group R3̅m, so the weak superstructure peaks between 20° to 30° shown in the inset panels in Figure 2 are not considered. The superstructure peaks are due to the √3ahex × √3ahex superlattice formation as a consequence of ordering (C2/m) of Li or vacancies with Ni, Mn, and Co ions in the TM layer. McCalla et al. have shown that the vacancies, Ni, and Mn order on this superlattice in Li[Ni1/6□1/6Mn2/3]O2,27 so the presence of Li in the transition metal layer is not required for this superstructure to form. Table 2 also shows that very little transition metal has been incorporated in the Li layer in any of the samples. The following results and discussion will focus only on the Aseries samples from now on, since the B-series’ results replicate the A series in every aspect. Sample A1 was predominantly single phase. The small quantities of impurities found in sample A3 that were indicated using arrows in the XRD patterns are given in Figure 2. First, there is a tiny peak at ∼15.7° in the XRD pattern of A3, and it may correspond to orthorhombic LiMnO2 that could have emerged during the synthesis because the JCDPS database shows that orthorhombic LiMnO2 has a strong Bragg peak at ∼15.7°.25 Second, there is a hump at 18.7° (clearly shown at the right panel in Figure 2) in the XRD pattern of A3. This may be evidence of nanodomains of a spinel phase. At this point, the following questions arise: (i) whether conventional XRD methods can be used to detect nanodomains of impurities; (ii) whether the hump at 18.7° in A3 is caused by nanodomains of spinel; and (iii) if so, whether spinel is the underlying cause for the observed low IRC. The spinel structure can accommodate lithium ions into its 16c sites during discharge provided the lower cutoff potential is below 2.8 V,26 so it could be the reason for the low IRC. To check whether the XRD experiments could detect small amounts of spinel admixture, nanosized spinel LiNi0.5Mn1.5O4 (sample S) was synthesized at low temperature. Supporting Information Figure S2 shows scanning electron micrograph images of the nanosized LiNi0.5Mn1.5O4 indicating that the primary particles were about 100 nm. The nanosized spinel was physically mixed with sample A1 in two different weight proportions (5 and 10 wt %), and XRD patterns were

Table 1. Sample, Nominal Composition of the Precursor, Nominal Composition of the Sample and % Irreversible Capacity Loss

sample

nominal precursor composition

nominal composition of the sample

IRC (% of 1st cycle charge capacity)

A1 A2 A3 A3Q B1 B2 B3

Ni0.166Mn0.5Co0.333 Ni0.166Mn0.5Co0.333 Ni0.166Mn0.5Co0.333 Ni0.166Mn0.5Co0.333 Ni0.2Mn0.5Co0.3 Ni0.2Mn0.5Co0.3 Ni0.2Mn0.5Co0.3

Li1.143Ni0.143Mn0.429Co0.285O2 Li1.067Ni0.156Mn0.467Co0.311O2 Li1.05Ni0.159Mn0.475Co0.316O2 Li1.05Ni0.159Mn0.475Co0.316O2 Li1.130Ni0.174Mn0.435Co0.261O2 Li1.067Ni0.187Mn0.467Co0.280O2 Li1.050Ni0.190Mn0.475Co0.285O2

22.5 7.5 6.5 6.9 22.0 9.5 7.5

± ± ± ± ± ± ±

0.5 0.5 0.5 0.5 0.5 0.5 0.5

the electrochemical behavior, especially on the IRC of the materials. Figure 1 shows the first cycle voltage−specific capacity profiles ranging between 2 and 4.8 V for all the three samples

Figure 1. First-cycle voltage vs specific capacity profiles of samples (a) A1, A2, and A3 and (b) B1, B2, and B3. The dotted lines represent their respective sister cells. C


Article


Figure 2. X-ray diffraction patterns of samples A1, A2, and A3. Black dotted, red, and green lines indicate the experimental pattern, calculated pattern (from Rietveld refinement), and their difference, respectively. Right panel: Enlarged X-ray diffraction patterns in the range of 17° to 20°.

spinel. For comparison, the XRD patterns of the sample A1 and the nanosized spinel (S) are also presented in Figure 3. The evolution of a new Bragg peak at 36.4° (shown using green arrows in Figure 3) is a clear marker for the presence of the nanosized spinel phase. With the increase in the fraction of the spinel phase from 5% to 10%, the peak intensity grows proportionally. Thus, the XRD patterns can detect the presence of nanosized spinel. Figure 4 shows the XRD patterns in the range of scattering angles from 35° to 39° for samples A1, A2, and A3 along with the physically mixed composites containing the intentionally added nanosized spinel. The XRD patterns of A2 and A3 do not show evidence for the spinel phase. Electrochemical methods are also sensitive to detecting the presence of the spinel phase. dQ/dV vs V plots corresponding to the first cycle were measured. Coin cells were made from the electrodes of the composites and electrochemically cycled from 2 to 4.8 V. Figure 5 shows dQ/dV vs V plots for samples A1, 95 wt % A1 + 5 wt % of spinel, and 90 wt % A1 + 10 wt % of spinel. The bottom panel shows an enlarged portion of the dQ/ dV vs V plot in the voltage range from 2.6 to 3 V. The oxidation peaks around 2.9 V and the concomitant reduction peaks around 2.75 V are caused by the spinel LiNi0.5Mn1.5O4. The dQ/dV peaks grow with the fraction of the spinel phase just as the growth of the 36.4° peak in the XRD patterns in Figure 4. Thus, electrochemical characterization proves to be quite powerful to identify any impure phase even in tiny amounts. Figure 6 shows dQ/dV vs V for samples A1, A2, and A3 along with A1 + 5% spinel and A1 + 10% spinel. Figure 6 shows that there is no evidence of spinel LiNi0.5Mn1.5O4 in samples A2 and A3. However, it reveals an extra pair of redox peaks in sample A3, which can be distinguished from the peaks corresponding to the LiNi0.5Mn1.5O4 spinel phase that appeared

Table 2. Structural Parameters Obtained from Rietveld Refinement of the XRD Patterns sample

a (Å)

c (Å)

z (O)

n Ni in Li sites

RB

A1 A2 A3 A3Q B1 B2 B3

2.845(8) 2.847(6) 2.848(9) 2.843(8) 2.846(7) 2.851(7) 2.851(1)

14.21(4) 14.23(2) 14.24(1) 14.23(1) 14.20(8) 14.23(9) 14.24(4)

0.259(3) 0.258(4) 0.254(7) 0.258(8) 0.258(9) 0.258(5) 0.255(9)

0.00(6) 0.01(6) 0.02(2) 0.03(3) 0.00(6) 0.01(5) 0.02(4)

2.58 2.75 4.14 2.41 2.20 3.12 3.23

measured. Figure 3 shows the XRD patterns of the composites 95 wt % A1 + 5 wt % of spinel and 90 wt % A1 + 10 wt % of

Figure 3. X-ray diffraction patterns of samples A1, A1 + 5% Spinel, A1 + 10% Spinel, and pure Spinel. “S” refers to nanosized spinel LiNi0.5Mn1.5O4. Green arrows indicate the presence of spinel “S”. D


Article


Figure 4. Enlarged X-ray diffraction patterns in the range of 35° to 39° for A1, A2, A3, A1 + 5% Spinel, A1 + 10% Spinel, and pure Spinel. “S” refers to nanosized LiNi0.5Mn1.5O4. Green arrows indicate the presence of spinel “S”.

between 2.4 and 2.9 V. Those two peaks, both during charge and discharge, are indicated using red arrows in Figure 6. If a feature representing a particular crystallographic phase can be identified in dQ/dV vs V, then there should be a corresponding signature in the XRD pattern as well. Hence the XRD patterns of the samples, particularly sample A3, were re-examined carefully. Figure 7 shows the XRD patterns of samples A1, A2, and A3 over the complete range between 10° to 70°, and an expanded view of the patterns between 14° and 20° is given separately in the bottom panels. The extra pair of redox peaks in dQ/dV vs V especially for sample A3 may correlate to either the appearance on an XRD peak at 15.7° or to the hump at 18.7°. The enlarged portion in the bottom left panel of Figure 7 shows that the 15.7° XRD peak can be seen in sample A2 as well whereas the bottom right panel shows that the hump at the 18.7° is absent in sample A2. The electrochemical characterization of the composite layered-spinel in ref 26 shows a pair of redox peaks closer to the ones observed here that correspond to LiNi0.425Co0.075Mn1.5O4. The redox peaks for orthorhombic LiMnO2 are at a higher voltage.25 Hence, the hump at 18.7° is most likely due to a spinel phase, whose composition remains unknown here. It is quite difficult to extract the composition of the impurities because their proportions are very low. Instead of trying to identify the composition of the impurities, steps were taken to eliminate the impurities from sample A3 and to check if the reduced IRC is in any way associated with the presence of the impurity phases. On the basis of suggestions in the work of McCalla et al.,27 a sample (A3Q) with composition exactly same as that of A3 was synthesized using the same precursor but was quenched from 900 °C instead of cooling slowly. Figure 8 shows the XRD

Figure 5. Top panel: dQ/dV vs V for samples A1, A1 + 5% Spinel, and A1 + 10% Spinel in the voltage range between 2 and 4.8 V. Bottom panel: The green bordered portion between 2.6 and 3 V has been enlarged and is shown separately for clarity.

Figure 6. dQ/dV vs V for samples A1, A2, A3, A1 + 5% Spinel ,and A1 + 10% Spinel. Red arrows indicate the presence of an impure phase in A3.

patterns of samples A3 and A3Q. A part of the patterns between 15° and 20° were enlarged and are shown separately in the bottom panels for clarity. The quenching strategy resulted in a single phase material and eliminated the impurities completely. The XRD patterns of A3 and A3Q are otherwise E


Article


Figure 7. Top panel: X-ray diffraction patterns of samples A1, A2, and A3. Bottom panel: The green bordered portion between (left) 14° and 17° and (right) 17° and 20° has been enlarged and is shown separately for clarity. Figure 8. Top panel: X-ray diffraction patterns of samples A3 and A3Q. Bottom panel: The black bordered portion between 15° and 20° has been enlarged and is shown separately for clarity.

identical, as expected, with the absence of the impurity peak at 15.7° and the hump at 18.7° as well. Sample A3Q was also tested electrochemically, and the first cycle data was extracted. Figure 9 shows the first cycle V vs Q profile of samples A3 and A3Q. The voltage profile of sample A3Q replicates that of sample A3. The IRC measured for sample A3Q (6.9%) is very close to that of sample A3 (6.5%). This unambiguously shows that the reduction in IRC is not an effect of an impurity phase including any spinel phase in sample A3. Figure 10 shows the comparison of the dQ/dV vs V plots of samples A3 and A3Q, and the bottom panel in Figure 10 shows an enlarged view between 2.2 to 3 V. As expected, the dQ/dV profiles of samples A3 and A3Q are very similar, but the pair of redox peaks corresponding to the extra phase(s) in sample A3 does not appear in sample A3Q. A few key points should be noted here regarding the low IRC caused by spinel phase incorporation, when it occurs. First of all, a spinel-layered composite reported by E. S. Lee et al. had 18% IRC even with 25 mol % spinel which is much larger than that of A2 and A3.26 Second, the extra discharge capacity observed below 2.5 V in A3 is very tiny and cannot be responsible for such small IRC. In other words if the lower cutoff voltage was set to 2.8 V, for example, low IRC would still be observed. Hence the contribution of the spinel impurity to low IRC in samples A2 and A3. B2 and B3 are negligible. Thus, the small IRC of samples A2, A3, B2, and B3 have nothing at all to do with the possibility of nanodomains of impurity phases. Instead the low IRC is inherent to Li-rich layered materials of such compositions. What is then responsible for the small IRC in samples A2, A3, B2, B3, and A3Q? First of all, the effect of surface area on

Figure 9. First-cycle voltage vs specific capacity profiles for samples A3 and A3Q.

IRC was analyzed. Table 3 shows the measured specific surface area (SSA) of all the samples from both A and B series. The surface area decreased as the lithium content decreased which is consistent with the literature.28 Figure 11 shows SEM images of the samples from the A series. Figure 11 shows that the particle sizes are larger for sample A1, and hence the higher SSAs for samples A2 and A3 are not unanticipated. F


Article


of our samples were prepared using the same methods to explore the impact of composition on IRC. A careful analysis using ICP-OES was made to determine the true chemical compositions of the samples. Table 4 shows the results of the ICP-OES analysis in comparison with the target composition. The ICP-OES compositions (Li:Ni:Mn:Co) were very close to the target compositions. As already mentioned the target compositions of samples A2, A3, and A3Q were not charge balanced between cations and anions. That is, when the Li, Ni, Mn, and Co ions are assumed to be in the 1+, 2+, 4+, and 3+ oxidation states, respectively, the total cationic charge is larger than 4. At the same time, XRD proves that the samples A2, A3, and A3Q are predominantly single phase layered LiMO2 type materials. These above two observations imply that there must be an excess of oxygen, leading to metal site vacancies, or the presence of either Co2+ or Mn3+ ions in the material. It is very unlikely to have Co or Mn in the +2 or +3 oxidation states, respectively, because they are not stable oxidation states in the studied materials.22−24 Moreover, in that case, the redox capacity during charge would have been higher than the observed values for samples A2, A3, and A3Q. Figure 13 shows the results of a calculation of first charge redox capacity (due to TM oxidation) and the total cationic charge, both plotted against the concentration of the assumed Mn3+ for the A3Q sample. The first charge redox capacity calculation assumes the oxidation of Ni2+ to Ni4+, Mn3+ (if any) to Mn4+, and Co3+ to Co3.6+.30 The calculated redox region capacity for sample A3Q is ∼150 mAh/g, assuming no Mn3+, and agrees well with that observed (145 mAh/g). Figure 13 shows that approximately 50% of Mn should be in the +3 oxidation state in order to have a total cationic charge equal to 4 in LiMO2 having cation ratios corresponding to sample A3Q. According to Figure 13, having 50% of Mn in the +3 oxidation state would have resulted in a first charge redox capacity equal to at least 200 mAh/g, which is not the case here. The only way to explain the observed results is if metal site vacancies exist. Transition metal vacancies have been shown to exist at compositions of Li-rich layered oxides by McCalla et al. when the amount of Li used in synthesis is smaller than normally used to balance the cation and anion charges,27 so their occurrence here is not unanticipated. The possibility for the presence of metal site vacancies is now explored. The metal atom ratios Li:Ni:Mn:Co from ICP-OES listed in Table 4 were taken to accurately describe each sample. The ratios are represented by the variables p′, a, b, and c, respectively. The values p′, a, b, and c in Table 4 have been scaled so that their sum is exactly 2.0. After heating, it is assumed that the final compound is Lip□qNixMnyCozO2, where

Figure 10. Top panel: dQ/dV vs V for samples A3 and A3Q (prepared by quenching). Bottom panel: The pink bordered portion between 2.2 and 3 V has been enlarged and is shown separately for clarity.

Table 3. BET Surface Area sample A1 A2 A3 A3Q GA1 B1 B2 B3

specific surface area (m2/g) 1.09 4.09 4.60 4.60 5.02 1.23 4.69 5.71

± ± ± ± ± ± ± ±

0.11 0.72 0.43 0.21 0.07 0.01 0.19 0.12

To see if high surface area caused the reduced IRC, sample A1 was ground until its SSA matched that of sample A3Q. Figure S3 (Supporting Information) shows the SEM images of the ground sample (GA1) and suggests that the surface area of the ground sample was increased (see Table 3). The first cycle IRC of GA1 was measured in the same way as done before grinding. Figure 12 shows the comparison of the first cycle behavior of samples A1 and GA1 and indicates that the ground sample showed the same IRC (22.5%) as the pristine one. This experiment suggests that the claimed low-IRC behavior has little dependence on the specific surface area at least in the compositions studied here. Recently M. G. Verde et al. have reported that carbonate-derived Li-rich material had better Coulombic efficiency than hydroxide or sol−gel counterparts, and that might be the case in the current study.29 However, all

p+q+x+y+z=2

(1)

and p + 2x + 4y + 3z = 4

(2)

These equations result from a filling of all metal sites by Li, Ni, Mn, Co, or vacancies (eq 1) and from charge balance, assuming Li+, Ni2+, Mn4+, and Co3+ (eq 2), respectively. The metal atom ratios as determined by ICP-OES must match those in Lip□qNixMnyCozO2, leading to the equations:

G

x = a(2 − q)/2

(3)

y = b(2 − q)/2

(4) DOI: 10.1021/cm504583y Chem. Mater. XXXX, XXX, XXX−XXX

Article


Figure 11. Scanning electron micrograph images of samples A1, A2, and A3.

is believed that a small of amount of Ni is in the 3+ oxidation state. The authors believe that Ni3+ and metal site vacancies cannot exist at the same time in these materials. Table 5 shows the calculated TM vacancy content based on the ICP-OES compositions and the oxidation state rules (Ni = 2+, Mn = 4+, and Co = 3+) as well as the composition of the formula units with the calculated vacancy contents. In Table 5, “q” represents the calculated TM vacancy concentration based on the above calculation. Table 5 also shows true density measurements obtained using a helium pycnometer. The densities were calculated from the XRD results assuming the vacancy content based on eq 7 and also based on zero vacancies. The results from both calculations are also listed in Table 5. In the latter case, Mn ions were assumed to be a mixture of Mn3+ and Mn4+ ions so that the total positive charge adds up to 4 in a formula unit. The calculated densities for the case assuming metal site vacancies agree well with the measured densities. The results in Table 5 suggest that very few, if any, Mn ions are Mn3+ in samples A2, A3, and A3Q. In order to show that the pycnometer measurements are reliable, the densities of commercial LiCoO2 and as-prepared Li2MnO3 were measured. Table 5 also shows their measured densities, and they are very close to their crystallographic densities calculated from their lattice constants. Therefore, samples A2, A3, and A3Q contain metal site vacancies, and it is believed that these vacancies are responsible for the low IRC in these materials. Most interesting about Table 5 is that all the materials that show low IRC can be described as Li[□qM1−q]O2 compositions. There need be no lithium atoms at all in the transition metals layers (indicated by “□qM1−q” within brackets). Rietveld

Figure 12. First-cycle charge−discharge voltage profiles of samples A1 and GA1 (grounded sample).

z = c(2 − q)/2

(5)

and p = p′(2 − q)/2

(6)

Equations 2−6 can be used to solve for q, the vacancy content of the resulting layered material. One obtains q = 2 − 8/A ,

where A = p′ + 2a + 4b + 3c

(7)

When the calculated value of q ≤ 0, it is believed that there are no metal site vacancies in the structure. However, instead, it H


Article


Table 4. Sample, Target Composition, Number of Moles of Lithium Required To Obtain a Formula Unit of Target Material, Number of Moles of Lithium Originally Added Including Extra 5 wt % for Compensating Li Loss To Obtain a Formula Unit of Target Composition, and Composition Observed from Chemical Analysis (ICP-OES) sample

target composition

no. of moles of Li required to obtain a formula unit of target composition

A1 A2 A3 A3Q B1 B2 B3

Li1.143Ni0.143Mn0.429Co0.285O2 Li1.067Ni0.156Mn0.467Co0.311O2 Li1.05Ni0.159Mn0.475Co0.316O2 Li1.05Ni0.159Mn0.475Co0.316O2 Li1.130Ni0.174Mn0.435Co0.261O2 Li1.067Ni0.187Mn0.467Co0.280O2 Li1.050Ni0.190Mn0.475Co0.285O2

1.143 1.067 1.05 1.05 1.13 1.067 1.05

no. of moles of Li originally added including extra 5 wt % for compensating Li loss to obtain a formula unit of target composition

ICP-OES Li:Ni:Mn:Co = p′:a:b:c

total charge of cations in a formula unit

1.200 1.120 1.103 1.103 1.187 1.120 1.103

1.126:0.147:0.429:0.298 1.049:0.161:0.466:0.325 1.013:0.167:0.484:0.336 1.017:0.166:0.482:0.335 1.155:0.163:0.430:0.251 1.054:0.187:0.470:0.289 1.046:0.190:0.472:0.292

4.030 4.207 4.290 4.283 3.954 4.175 4.191

Figure 13. Plot of % Mn in +3 oxidation state vs first charge sloping region capacity and total charge of the cations for A3Q.

Table 5. Sample, Calculated Vacancies Based on ICP-OES Compositions and Oxidation State Rules, Composition of Formula Units with Calculated Vacancies, Calculated Densities Based on ICP-OES Compositions and Calculated Vacancy Content, Densities Determined from a He Pycnometer, Calculated Densities Based on ICP-OES Compositions, and Measured Lattice Parameters Assuming No Vacancies

sample A1 A2 A3 A3Q B1 B2 B3 commercial LiCoO2 Li2MnO3 a

calculated metal site vacancy content based on ICP-OES compositions and Ni2+, Mn4+, Co3+ oxidation states (q) 0.015 0.100 0.136 0.132 −0.023b 0.084 0.091 -

composition of a formula unit with calculated metal site vacancy content

calculated density based on ICP-OES composition, calculated metal site vacancy content, and measured lattice parameters from XRD (g/cm3)

Li1.118□0.015Ni0.146Mn0.426Co0.296O2 Li0.998□0.098Ni0.153Mn0.443Co0.309O2 Li0.945□0.135Ni0.156Mn0.451Co0.313O2 Li0.950□0.132Ni0.155Mn0.450Co0.313O2 Li1.168□−0.023Ni0.165Mn0.435Co0.254O2 Li1.01□0.084Ni0.179Mn0.45Co0.277O2 Li0.998□0.091Ni0.181Mn0.451Co0.279O2

4.4552 4.5063 4.5255 4.5426 4.3782 4.5006 4.5096 -

-

-

± ± ± ± ± ± ±

0.0314 0.0323 0.0328 0.0328 0.0308 0.0320 0.0321

measured density from helium pycnometer (g/cm3) 4.4528 4.5304 4.6005 4.6025 4.3801 4.5118 4.5313 5.0569

± ± ± ± ± ± ± ±

0.0036 0.0082 0.0197 0.0128 0.0068 0.0165 0.0252 0.0098

3.9137 ± 0.0041

calculated density based on ICP-OES compositions and measured lattice parameters from XRD assuming no vacancies (g/cm3) 4.4764 ± 4.6567 ± 4.7381 ± 4.7504 ± 4.3966 ± 4.6279 ± 4.6491 ± 5.0576a

0.0316 0.0340 0.0352 0.0347 0.0305 0.0334 0.0337

3.888

The density was obtained from the JCPDS database. bImplies oxygen vacancies and so density calculated accordingly. I


Article


The metal compositions from ICP-OES results and the oxidation state versus atomic occupancy rules suggested that the compounds with low irreversible capacity contain metal site vacancies. This was further supported by true density measurements using a helium pycnometer. In fact, taking all the experimental evidence into account, the low IRC materials are best thought of as Li[□qM1−q]O2 with metal atom vacancies in the transition metal layer and no transition metals in the Li layer. As such, these are “traditional layered compounds” and should not be expected to have high IRC, even though they are “Li rich” (more Li atoms than transition metal atoms) and display a 4.5 V plateau. The metal site vacancies may provide avenues for enhanced atomic diffusion during and after the oxygen loss process which may contribute to the small irreversible capacity. Further studies are in progress to better understand the properties of these materials. In summary, all the materials with low IRC can be described in the following way: 1. Materials with low IRC have less than or equal to one Li atom per two oxygen atoms, so that there is no need for lithium atoms in the transition metal layer. It may be this “traditional layer structure” that leads to the low IRC. 2. Materials with low IRC have a 4.5 V plateau caused by the “oxygen-loss” plateau because there is insufficient transition metal redox capacity to charge compensate when all (or most) of the Li is extracted. 3. Materials with low IRC and a 4.5 V plateau have vacancies on the metal atom sites. These vacancies may play a role in diffusion to enable low IRC. In order for statements #1 and #2 above to be simultaneously true, there must be metal atom vacancies.

refinement showed no evidence for transition metals in the Li layers (Table 2), so the samples with low IRC appear to be “traditional layered materials” but with vacancies on the transition metal sites! Therefore, provided that lithium is extracted from and reinserted into the lithium layers, the IRC should be small even in the presence of the oxygen-loss process. In addition, the presence of the vacancies may help facilitate any needed atom motion to enable low IRC. However, one must realize that these would normally be thought of as “Lirich” materials because the lithium to transition metal ratio is still greater than 1. This suggests the origin of the high IRC in samples A1 and B1 is the presence of Li atoms in the transition metal layer and the lack of vacancies to enable needed atom motions to allow reintercalation of all the lithium. A relatively low IRC (∼10 to 12%) was reported by Sathiya et al. in their Li-excess material that exhibited oxygen oxidation/reduction and the associated 4.5 V plateau.19 Hence, it is very likely that, in the case of the materials with low IRC, where there is an extended 4.5 V plateau, oxygen oxidation/reduction plays a role in the charge−discharge process. Finally, Figure 14 shows the first cycle profiles of samples A1, A2, A3, and A3Q charged to the same delithiation limit. Once

■

ASSOCIATED CONTENT

S Supporting Information *

X-ray diffraction patterns of samples B1, B2, and B3. Scanning electron micrograph images of the as-prepared nanosized spinel LiNi0.5Mn1.5O4, samples A1 and GA1 (ground). This material is available free of charge via the Internet at http://pubs.acs.org.

■

Figure 14. First-cycle charge−discharge voltage profiles of A1, A2, A3, and A3Q during constant time (25 h) first charge protocol.

AUTHOR INFORMATION

Corresponding Author

*E-mail: jeff[email protected]. Tel.: 001-902-494-2991. Fax: 001902-494-5191.

again, A1 shows higher IRC whereas A2, A3 (∼4%), and A3Q show low IRC. Therefore, the increased IRC found for sample A1 is not caused by the fact that it has higher first-charge capacity (see Figure 1).

Notes

■

The authors declare no competing financial interest.

■

CONCLUSIONS Two sets of layered positive electrode materials with slightly different lithium to transition metal ratio were synthesized from two different precursors and electrochemically characterized with a focus on the first cycle behavior, especially the irreversible capacity loss. Surprisingly and beneficially, the samples having lower lithium to transition metal ratio exhibited irreversible capacity loss as small as 4.0% of the first charge capacity. Few impurities were found in the low-IRC materials, but quenching resulted in pure phase materials. Studies of dQ/ dV vs V confirmed that the impurities were removed by quenching and that quenching did not affect the low IRC. Careful studies showed that the low-IRC behavior did not originate from any impurity phases including nanodomains of spinel. Therefore, the small IRC of these samples is inherent to their composition and structure.

ACKNOWLEDGMENTS The authors thank NSERC and 3M Canada for funding this work under the auspices of the Industrial Research Chair program.

■

REFERENCES

(1) Lu, Z.; MacNeil, D. D.; Dahn, J. R. Electrochem. Solid-State Lett. 2001, 4, A191−A194. (2) Lu, Z.; Beaulieu, L. Y.; Donaberger, R. A.; Thomas, C. L.; Dahn, J. R. J. Electrochem. Soc. 2002, 149, A778−A791. (3) Thackeray, M. M.; Johnson, C. S.; Vaughey, J. T.; Li, N.; Hackney, S. A. J. Mater. Chem. 2005, 15, 2257−2267. (4) Aricò, A. S.; Bruce, P.; Scrosati, B.; Tarascon, J.-M.; van Schalkwijk, W. Nat. Mater. 2005, 4, 366−377. (5) van Bommel, A.; Krause, L. J.; Dahn, J. R. J. Electrochem. Soc. 2011, 158, A731−A735. J


Article

Chemistry of Materials (6) Hong, J.; Lim, H.-D.; Lee, M.; Kim, S.-W.; Kim, H.; Oh, S.-T.; Chung, G.-C.; Kang, K. Chem. Mater. 2012, 24, 2692−2697. (7) Zheng, J.; Gu, M.; Genc, A.; Xiao, J.; Xu, P.; Chen, X.; Zhu, Z.; Zhao, W.; Pullan, L.; Wang, C.; Zhang, J.-G. Nano Lett. 2014, 14, 2628−2635. (8) Mueller-Neuhaus, J. R.; Dunlap, R. A.; Dahn, J. R. J. Electrochem. Soc. 2000, 147, 3598−3605. (9) Lu, Z.; Dahn, J. R. J. Electrochem. Soc. 2002, 149, A815−A822. (10) Tran, N.; Croguennec, L.; Ménétrier, M.; Weill, F.; Biensan, P.; Jordy, C.; Delmas, C. Chem. Mater. 2008, 20, 4815−4825. (11) Kang, S.-H.; Johnson, C. S.; Vaughey, J. T.; Amine, K.; Thackeray, M. M. J. Electrochem. Soc. 2006, 153, A1186−A1192. (12) Zheng, J.; Deng, S.; Shi, Z.; Xu, H.; Xu, H.; Deng, Y.; Zhang, Z.; Chen, G. J. Power Sources 2013, 221, 108−113. (13) Gao, J.; Kim, J.; Manthiram, A. Electrochem. Commun. 2009, 11, 84−86. (14) Gao, J.; Manthiram, A. J. Power Sources 2009, 191, 644−647. (15) Wu, Y.; Murugan, A. V.; Manthiram, A. J. Electrochem. Soc. 2008, 155, A635−A641. (16) Fell, C. R.; Qian, D.; Carroll, K. J.; Chi, M.; Jones, J. L.; Meng, Y. S. Chem. Mater. 2013, 25, 1621−1629. (17) Koga, H.; Croguennec, L.; Ménétrier, M.; Mannessiez, P.; Weill, F.; Delmas, C. J. Power Sources 2013, 236, 250−258. (18) Ito, A.; Li, D.; Sato, Y.; Arao, M.; Watanabe, M.; Hatano, M.; Horie, H.; Ohsawa, Y. J. Power Sources 2010, 195, 567−573. (19) Sathiya, M.; Rousse, G.; Ramesha, K.; Laisa, C. P.; Vezin, H.; Sougrati, M. T.; Doublet, M.-L.; Foix, D.; Gonbeau, D.; Walker, W.; Prakash, A. S.; Ben Hassine, M.; Dupont, L.; Tarascon, J.-M. Nat. Mater. 2013, 12, 827−835. (20) Hill, R. J.; Howard, C. J. J. Appl. Crystallogr. 1985, 18, 173−180. (21) Marks, T.; Trussler, S.; Smith, A. J.; Xiong, D.; Dahn, J. R. J. Electrochem. Soc. 2011, 158, A51−A57. (22) Park, Y. J.; Hong, Y.-S.; Wu, X.; Kim, M. G.; Ryu, K. S.; Chang, S. H. J. Electrochem. Soc. 2004, 151, A720−A727. (23) Kim, J.-M.; Chung, H.-T. Electrochim. Acta 2004, 49, 937−944. (24) Oishi, M.; Fujimoto, T.; Takanashi, Y.; Orikasa, Y.; Kawamura, A.; Ina, T.; Yamashige, H.; Takamatsu, D.; Sato, K.; Murayama, H.; Tanida, H.; Arai, H.; Ishii, H.; Yogi, C.; Watanabe, I.; Ohta, T.; Mineshige, A.; Uchimoto, Y.; Ogumi, Z. J. Power Sources 2013, 222, 45−51. (25) Reimers, J. N.; Fuller, E. W.; Rossen, E.; Dahn, J. R. J. Electrochem. Soc. 1993, 140, 3396−3401. (26) Lee, E.-S.; Huq, A.; Chang, H.-Y.; Manthiram, A. Chem. Mater. 2012, 24, 600−612. (27) McCalla, E.; Rowe, A. W.; Camardese, J.; Dahn, J. R. Chem. Mater. 2013, 25, 2716−2721. (28) Wang, J.; Yuan, G.; Zhang, M.; Qiu, B.; Xia, Y.; Liu, Z. Electrochim. Acta 2012, 66, 61−66. (29) Verde, M. G.; Liu, H.; Carroll, K. J.; Baggetto, L.; Veith, G. M.; Meng, Y. S. ACS Appl. Mater. Interfaces 2014, 6, 18868−18877. (30) Arunkumar, T. A.; Wu, Y.; Manthiram, A. Chem. Mater. 2007, 19, 3067−3073.

K


High Capacity Li-Rich Positive Electrode Materials with Reduced First

Recommend Documents