Ni - ACS Publications

Subscriber access provided by University of Sunderland

Article

The Phase Evolution and Degradation Modes of R3#m LiNi CoAlO Electrodes Cycled Near Complete Delithiation x

1-y-z

y

z

2

Nicholas V. Faenza, Nathalie Pereira, David M. Halat, Julija Vinckeviciute, Lejandro Bruce, Maxwell D. Radin, Pinaki Mukherjee, Fadwa Badway, Anna Halajko, Frederic Cosandey, Clare P. Grey, Anton Van der Ven, and Glenn G. Amatucci Chem. Mater., Just Accepted Manuscript • DOI: 10.1021/acs.chemmater.8b02720 • Publication Date (Web): 09 Oct 2018 Downloaded from http://pubs.acs.org on October 13, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Chemistry of Materials

The Phase Evolution and Degradation Modes of R𝟑m LixNi1-y-zCoyAlzO2 Electrodes Cycled Near Complete Delithiation Nicholas V. Faenza1, Nathalie Pereira1, David M. Halat2, Julija Vinckeviciute3, Lejandro Bruce1, Maxwell D. Radin3, Pinaki Mukherjee4, Fadwa Badway1, Anna Halajko1, Frederic Cosandey4, Clare P. Grey2, Anton Van der Ven3, and Glenn G. Amatucci1 1Energy

Storage Research Group, Department of Materials Science and Engineering, Rutgers University, North Brunswick, New Jersey 08902, United States 2Department of Chemistry, University of Cambridge, Cambridge CB2 1EW, United Kingdom 3Materials Department, University of California Santa Barbara, Santa Barbara, California 93106-5050, United States 4Department of Materials Science and Engineering, Rutgers University, Piscataway, New Jersey 08854, United States Abstract: The practical utilization of energy densities near the theoretical limit for R3m layered oxide positive electrode materials is dependent on the stability of the electrochemical performance of these materials at or near full delithiation. In order to develop new chemistries and novel approaches towards the improvement of the electrochemical performance of these materials at such high states of charge, a robust understanding of the failure mechanisms limiting current materials is necessary. Thorough analysis of LixCo1-yAlyO2 and LixNi1-yAlyO2 as well as LixNi0.8Co0.2O2 and LixNi0.8Co0.15Al0.05O2 (1 ≥ x ≥ 0 and 0.2 ≥ y ≥ 0) enabled the identification of key relationships between the transition metal chemistry of the electrode, its structural stability, and cycling characteristics at or near complete delithiation (4.75 V). Extensive characterization of these materials was achieved by a multitude of physical and electrochemical techniques to investigate the relative importance of surface vs. bulk phenomena. The resulting insights derived from these analyses highlight the importance of the intrinsic structural and mechanical stability of the electrode when highly delithiated and establish guidelines for identifying positive electrode materials with improved high state of charge performance. Particularly important is the contrasting electrochemical impact of Al substitution into LiCoO2- and LiNiO2-based materials, which is shown to likely arise from the enhanced propensity for Al ions to migrate to the tetrahedral site in Co-rich compounds at high states of delithiation.

Introduction: To enhance the energy density of lithium-ion batteries and reduce the gap between the theoretical and practically obtained capacities for layered oxide (R3m) positive electrodes, the upper limit of the operating potential window needs to be increased.1 Presently, commercial layered oxide electrode materials are limited to ~4.2 - 4.3 V, but complete delithiation often occurs at potentials greater than 4.7 V.1–5 This results in only ~60% utilization of the stored energy, which is insufficient to meet the future energy demands of electric vehicles and portable devices.1,6 It has been over two decades since it was first demonstrated that Li could be completely delithiated from R3m LiCoO2 and LiNiO2 and reinserted with

ACS Paragon Plus Environment

Chemistry of Materials 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

>95% efficiency in the first cycle without structural exfoliation or collapse.2 While an abundance of excellent work focused on enabling cycle life at or near full delithiation beyond one cycle has been completed, there are still substantial obstacles preventing the practical application of these materials cycled within this high capacity window. Amongst the main issues restricting the use of layered oxide materials at the highest states of charge are the structural instability of these materials,5,7,8 catalytic parasitic reactions with the electrolyte,9–11 oxygen evolution from the host structure,9,12–14 transition metal dissolution,11,15–18 and thermal stability issues.14,19 To a significant degree, the aforementioned phenomena have been found to be interrelate and can result in significantly reduced electrochemical capacity and increased capacity fading.20–23 LiCoO2 was not only the first layered oxide material to be commercially used as a positive electrode in lithium-ion batteries, but also established the foundation from which current layered positive electrodes are based.1 The structural transitions in LiCoO2 and in the isostructural compound, LiNiO2, during delithiation are well documented, and are thoroughly described elsewhere.2,3,28,4,5,7,8,24–27 Both LiCoO2 and LiNiO2 have very similar phase transitions as a function of delithiation, although the specific lithiation values at which these transitions occur differ because of the different electronic states of Ni and Co.4,29 The main structural differences between LiCoO2 and LiNiO2 occur because of the Jahn-Teller distortion of NiO6 octahedra, and from the propensity for Ni to be reduced from Ni3+ to Ni2+ and subsequently migrate to the Li layer.7,25,30 Fully delithiated LiCoO2 has been shown to form a stable CoO2 compound, having an O1 structure in the hexagonal lattice. This structure arises from the shifting of the basal planes as the Li separating these planes is completely removed.2,24 Formation of the O1 structure with AB oxygen packing and face-sharing octahedra has been experimentally and theoretically shown to be more stable than the O3 structure with AB CA BC oxygen packing and edge-sharing octahedra for completely delithiated LiCoO2.2,24 While the O1 structure is also thermodynamically favorable for LiNiO2, Croguennec et al. demonstrated that it can be experimentally difficult to observe the O1 phase for LiNiO2 because an excess of nickel ions in the Li layer is sufficient to prevent the sliding of basal planes.7,25 The higher energy density, reduced cost and lower toxicity of Ni relative to Co, has made LiNiO2 a highly desirable positive electrode material.6,31,32 Unfortunately, the poor cycle stability, safety characteristics, and capacity retention of LiNiO2 has prevented its commercial utilization.22,33–36 As a result, there has been much effort to enhance the structural stability and electrochemical properties of LiNiO2 through partial substitution of Co or Al ions for Ni. Both Co and Al dopants are effective in limiting the amount of Ni ions present in the Li layer, which causes a significant improvement in the electrochemical performance.19,22,43–45,30,35,37–42 Rougier et al. demonstrated that when Co partially replaces Ni in LiNiO2, 𝐼𝐼𝐼 3+ the smaller ionic radius of Co3+ (𝑟𝐼𝐼𝐼 𝐶𝑜 = 0.53 Å) relative to Ni (𝑟𝑁𝑖 = 0.56 Å) contracts the transition metal slab. The tighter slab substantially reduces the stability of divalent Ni (𝑟𝐼𝐼 𝑁𝑖 = 0.68 Å) within the transition metal-oxygen layer, and subsequently limits the amount of Ni2+ on the Li sites (hereto referred as transition metal mixing). This mechanism occurs because of the necessity to maintain charge neutrality in synthesized materials, which requires a divalent Ni ion in the transition metal slab for every divalent Ni ion in the Li layer.37 A similar mechanism has been shown for Al-substituted nickelates, except to a greater extent because of the increased ionicity of the Al-O bonds, which efficiently contracts the transition metal octahedra.44,45 However, Al-substituted materials often still have transition metal mixing because of phase segregation into Al-rich and Al-deficient regions, and because of the tendency for Al to reside in the tetrahedral site.34,39,42,44 Finally, a recent computational study suggests that the reduced transition metal mixing and increased structural stability of Co or Al-substituted LiNiO2 may also be


Page 2 of 55

Page 3 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


attributed to the partial elimination of the 180o Ni-O-Ni “superexchange interaction” energy benefit that occurs when Ni ions reside in the Li sites.44 In addition to reducing transition metal mixing, Co and Al substitution in LiNiO2 increases the average voltage of the Li insertion and extraction processes and smooths voltage plateaus arising from phase transitions.22,35,37,45 The degree to which the voltage profile changes scales with the amount of metal additive incorporated into the host material.37,45 Furthermore, metal substitution can have a profound impact on stabilizing the rate and capacity retention of LiNiO2. The enhanced electrochemical performance of mixed metal layered oxide materials can be partially attributed to the reduction of transition metal mixing, which restricts lithium ion diffusivity, and to the smoothing of phase transitions, which alleviates some of the internal stresses on the host structure.22,34,45 However, it should be noted that Al is electrochemically inactive and can lower the obtained capacity, by limiting the amount of lithium available for extraction.34,35,40,45 Al substitution has also been theoretically predicted and experimentally shown to limit the Li+ diffusivity in layered transition metal oxides.33,46 In contrast, Kang and Ceder suggested that Al substitution lowers the activation barrier for Li hopping by lowering the electron density around the oxygen atoms, and thus minimizing the electron screening that reduces the electrostatic interactions between Li and the transition metals.46 As a result of the beneficial properties obtained by substituting low cost LiNiO2 with Co and Al, LiNi0.8Co0.15Al0.05O2 (NCA) has emerged as a leading positive electrode material.1,6,43 By using a mixture of multiple metal ions, NCA shows improved structural stability and capacity retention relative to the baseline Ni and Co layered oxide materials, and has become one of the most highly utilized positive electrode materials in present day Li-ion batteries.29 While there is much published work on the structural and electrochemical properties of Al-substituted LiCoO2 and LiNiO2 materials in the potential window up to 4.3 V, there is a dearth of results for these materials near full delithiation (approaching 4.7 V). Moreover, there is an inherent shortcoming in that these materials are often made in-house and generally only a few compositions are studied at once. As a result, there are large variations in material quality and characteristics between groups, which makes it difficult to draw broad conclusions about the effect of the metal chemistry on the material’s properties. The work herein investigates the structural and chemical stability of both the bulk and surface of a variety of highly delithiated Al-substituted LiCoO2 and LiNiO2 materials, as well as Co-substituted LiNiO2 and finally, NCA. LiCoO2 and LiNiO2 are also included as benchmarks. When possible, synthesized samples and commercially available materials are compared to demonstrate sample quality and the effect of different surface area/crystallite size. All materials investigated herein are introduced and organized into three groups, based on the samples chemistries encompassing LiCo1-yAlyO2, LiNi1-yAlyO2 (0 ≤ y ≤ 0.2), and LiNi0.8M0.2O2 (M = Al and/or Co), referred to as group 1, 2 and 3, respectively. Detailed electrochemical and structural characterization as well as the cycling performance and transition metal dissolution results are then systematically described for each group of materials enabling the identification of trends between the electrodes’ transition metal chemistry and the electrochemical and structural properties of the layered oxide materials. Furthermore, a summary of important characterization results for each material at high states of charge, presented in Table II, provides further insight into the critical properties determining the electrode’s cycling stability near full delithiation. Enhanced understanding of the mechanisms for capacity degradation and the influence of the transition metal chemistry on a wide range of layered oxide compositions was ascertained from the combination of extensive experimental physical, structural, and



electrochemical characterization techniques, and also supported by density functional theory (DFT) calculations. Experimental Section: a. Materials Preparation: Commercial NCA (NAT1050, TODA America) and LiCoO2 (Seimi Chemical Co. Ltd, Japan) materials were used as received as model materials in all characterization and electrochemistry experiments. All materials synthesized in-house were fabricated as follows. Lithium acetate (Aldrich), aluminum acetate basic hydrate (Alfa Aesar), nickel (II) acetate tetrahydrate (Aldrich), and cobalt (II) acetate (Aldrich) were mixed in stoichiometric amounts based on the precursors’ metal content and dissolved in deionized water. While continuously stirring, the water was boiled off and the resulting material was dried for at least 16 hours at 60oC in a dry room (dew point of -35 to -40oC), and then ground into a fine powder with a mortar and pestle. Mixtures were then annealed at 750oC (for Ni-rich compositions) or 850oC (for Co-rich compositions) for 4 hours and cooled gradually to room temperature, all in a flowing O2 gas environment. These temperatures were experimentally identified to produce the highest quality layered structures with respect to phase purity and cation ordering. Post-annealed materials were quickly transferred into the dry room, where they were ground with a mortar and pestle, and then annealed a second time, using conditions identical to the first, before being ground again. To protect the phase purity and surface structure of these layered oxide materials, which has been shown to degrade upon exposure to ambient air, all materials were stored in a dryroom (dew point of -35 to -40oC) or Ar-filled glove box (< 0.1 ppm of H2O and O2).23,47 Free-standing electrodes were fabricated in the dry room using the Bellcore method.48 Casting slurries consisted of a mixture of the active material, poly(vinylidene fluoride-co-hexafluoropropylene) (PVDFHFP, Kynar 2801, Elf Atochem), carbon black (Super P (SP), MMM), propylene carbonate (≥ 99.7%, Aldrich), and acetone (≥ 99.9%, Aldrich). The homogeneous slurries were cast, allowed to air dry, and then the propylene carbonate plasticizer was extracted by soaking the electrode in anhydrous diethyl ether (99.8%, Aldrich). The resulting self-standing electrodes composed of 79.9 wt.% active material, 7.0 wt.% SP, and 13.1 wt.% PVDF-HFP were subsequently dried at 120oC under vacuum for a minimum of 10 hours, and stored in an Ar-atmosphere glovebox to avoid atmospheric exposure. Binder-free powder electrodes comprised of a mixture of 97.5 wt.% active material and 2.5 wt.% SP were used for all NMR and TEM studies. Both the active material and carbon additive (SP) were dried at 120oC for at least 10 hours prior to mixing. b. Electrochemical Characterization: Coin cells with Al-clad positive bases (2032, Hohsen Corp.) were assembled in an Ar-filled glovebox (< 0.1 ppm of H2O and O2) using a Li metal (FMC Lithium) negative electrode and Whatman GF/D glass fiber separators saturated with a 1M LiPF6 ethylene carbonate: dimethyl carbonate (EC:DMC) (1:1 volume ratio) electrolyte (BASF) (< 20 ppm H2O). Electrochemical characterization was performed with a VMP3 and MacPile (Bio-Logic Science Instruments), and Series 4000 (Maccor) battery cycling systems at either 24 or 60oC. All cells were made in duplicates to ensure reproducibility. For evaluation of cycling performance, cells were galvanostatically charged at room temperature at 20 mA/g (per g of active material) to Vmax (4.5 or 4.75 V), held under potentiostatic conditions until the current dropped below 10


Page 4 of 55

Page 5 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


mA/g, and then discharged at 10 mA/g to 2.75 V. Differential charge cells, used to extract dQ/dV data, were charged and discharged at 5 mA/g between 4.75 - 2.75 V at room temperature, while samples to study dissolution by inductively coupled plasma - optical emission spectroscopy (ICP-OES) were charged at 25 mA/g to 4.75 V, and held at a constant potential of 4.75 V for 10 hours, all at 60oC. Potentiostatic intermittent titration technique (PITT) was used for a multitude of electrochemical applications. In all cases, a 10 mV step with a current cutoff of 1 mA/g was employed at room temperature. All Li+ diffusivity calculations were based on a previously used methodology, where the Li+ diffusivity (DLi) can be calculated from the linear portion of a single-phase region in a PITT experiment using Equation 1.28,49 Equation 1:

𝑫𝑳𝒊 = ―

𝒅𝐥𝐧 (𝐈)𝟒𝑳𝟐 𝒅𝒕

𝝅𝟐

The Li+ diffusivity is dependent only on the change in the natural log (ln) of current (I) with time and the diffusivity length (L). Confirmation that the current response was linear at each calculated Li+ diffusivity point was completed by plotting ln(I) and individually analyzing the results at each potential. To provide a comprehensive approximation of the Li+ diffusivity over the entire lithiation range studied, both singleand two-phase reactions were considered. It should be noted though that Equation 1 was originally derived by Weppner et al. for single-phase diffusion only and its application in two-phase reactions is not rigorous, only providing a measure of the rate limiting process during a two-phase reaction in the form of an effective diffusion coefficient.49 In the case of two-phase reactions, the Li+ diffusivity results were only taken from the data points after the non-Cottrellian region when the current response became linear. Since changes in the diffusivity length can cause shifts in the calculated Li+ diffusivity by greater than an order of magnitude, the focus of the diffusivity calculations was on the variation in the Li+ diffusivity as a function of material and lithiation rather than calculating the precise magnitude of the diffusivity. To easily associate the changes in the Li+ diffusivity between samples to the differences in the chemistries and structures of the materials, the diffusivity length (L) was assumed to be 100 nm for all materials, which is approximately the average radius of the synthesized materials. Reliable comparisons between all fabricated materials can be made since the synthesis procedure for all materials as well as the particle sizes and morphologies, as determined by field emission scanning electron microscopy (FESEM), were similar. c. Physical Characterization: FESEM images were obtained on a Zeiss microscope. Prior to being put under vacuum, samples were coated with approximately 20 nm of Au to prevent charging, and subsequently exposed to the ambient atmosphere for less than 1 minute. Nanometer resolution X-ray energy dispersive spectroscopy (EDS) mapping and high angle annular dark-field (HAADF) transmission electron microscopy (TEM) imaging were performed either in a FEI Talos or a FEI Titan Themis microscope. Both microscopes were fitted with four super EDS Bruker detectors. A Bruker D8 Advance diffractometer (Cu Kα, λ = 1.5406 Å) was used for all X-ray diffraction (XRD) characterization. Samples of positive electrode materials were prepared for XRD by carefully sprinkling the powder onto a glass slide with X-ray transparent grease in a manner to minimize preferential orientation. XRD scans of pristine powders for each material were conducted over a 2Θ range of 15-70o at a scan rate of 0.08o 2Θ/min. Analysis of cycled electrodes was conducted by disassembling and retrieving the sample electrode in an Ar-filled glovebox. The electrode was then washed with DMC to remove any



residual electrolyte, and allowed to dry in the glovebox. To minimize atmospheric exposure, the dried samples were sealed onto glass slides using Kapton film and X-ray transparent grease in an Ar-filled glovebox. A scan rate of 2o 2Θ/min over a range of 15-70o 2Θ was employed to minimize reactions with the ambient atmosphere since the Kapton film was not truly hermetic. In addition, XRD experiments were performed in-situ using a custom-made cell with a Be window and a carbon coated Al mesh positive electrode current collector.50 The cell consisted of the same active components as the previously discussed coin cells. A Kapton film lining protected the Be window from oxidation at high potentials. During in-situ experiments, the cell was charged galvanostatically at C/50 to 4.75 V, and then held potentiostatically until full delithiation was achieved. Slow electrochemical charge of the cell ensured that the change of the lithium content in the positive electrode during each 1 h-scan at the rate of ~1o 2Θ/min was x ≤ 0.02 in LixMO2. TOPAS software (version 5, Bruker AXS) was used for all Rietveld refinement. A MMC 274 Nexus multi-mode calorimeter with a high temperature coin cell module (Netzsch) was utilized to perform operando isothermal microcalorimetry experiments. Samples were 2032 coin cells (previously described) oriented with the positive electrode flush against the sample heat flux sensor. To eliminate the thermal contribution from heat sources other than the electrodes, a reference cell consisting of all metal cell parts, separator and the electrolyte was used. The reference cell was placed on the reference heat sensor, which is located directly opposite the sample sensor. The measured heat flux is the difference in the heat flux between the sample and reference sensors. A chamber temperature of 30oC was used for all isothermal microcalorimetry experiments and cells were cycled at a constant current (25 mA/g) between 4.75 V and 2.75 V. The influence of the irreversible heat and the heat from Li concentration gradients was minimized by using low currents for cycling. Furthermore, the impact of the heat capacity of the cell is negligible because the microcalorimeter was operated under isothermal conditions. Due to the absence of a prolonged time at very high potentials, little heat contribution is expected from parasitic electrode-electrolyte reactions. Thus, the entropic changes to the positive electrode material during cycling should be the main factor dictating the measured heat flux. To characterize transition metal dissolution, inductively coupled plasma - optical emission spectroscopy (ICP-OES) analysis was completed by Galbraith Laboratories, Inc. (Knoxville, TN). All ICP-OES coin cells were assembled identically as previously described except a Cu negative current collector was added to prevent transition metal leaching from the stainless steel cell body. The positive electrode was protected by the cell’s Al cladding. Cells were charged to 4.75 V at 25 mA/g, and held at 4.75 V for 10 h, while in a 60oC incubator. After charging, cells were quickly disassembled in the Ar-filled glovebox. The Li negative electrode and adjacent separator were retrieved and soaked in an acidic (28 vol.% HCl) solution where the negative electrode was fully dissolved. The separator was then removed and the solution was tested by ICP-OES. Total amounts of Co, and Ni dissolution were calculated by extrapolating the measured concentration to the entire sample volume. The percentage of dissolved Co, and Ni was determined by dividing the mass of the measured dissolution products by the initial amount of each metal present in the sample. To determine the precise stoichiometry of our in-house fabricated materials, select as-synthesized materials were also tested by ICP-OES. To account for any possible Ni or Co contamination from the stainless steel cell body with an Al-clad base or from any of the inactive components of the cell, two baseline cells comprising all cell components except for the positive electrode were fabricated and tested identically to the other samples. In the two baseline samples the amount of Ni


Page 6 of 55

Page 7 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


and Co on the negative electrode was below the instrument’s detection limit, signifying that the cell components have negligible contribution to the measured Ni and Co on the remaining samples. 27Al

solid-state nuclear magnetic resonance (NMR) spectra were obtained at 11.7 T on a Bruker Avance III 500 MHz spectrometer (University of Cambridge) operating at a Larmor frequency of 130.32 MHz for 27Al. Samples were prepared by first scraping the binder-free electrodes inside an Ar-filled glovebox, and then sealing the recovered powder into an airtight 1.3 mm zirconia rotor without atmospheric exposure. Measurements were performed at room temperature using a double-resonance 1.3 mm Bruker probe at a magic-angle spinning (MAS) rate of 55 kHz. A rotor-synchronized spin-echo pulse sequence was used with a π/2 excitation pulse length of 1.5 µs; if required to suppress the paramagnetic background observed in some samples, the interpulse delays were each set to 10 rotor periods (~180 µs), i.e., a T2 filter was applied. Typical recycle delays were 10 s (for diamagnetic samples) or 250 ms (for paramagnetic, i.e., cycled samples). For cycled samples, saturation-recovery measurements were performed to measure the T1 (spin-lattice) relaxation time, with a maximal delay of ~1.9 s. Additionally, measurements of the T2 (spin-spin) relaxation time were performed, by progressively incrementing the interpulse delays in a series of Hahn echo measurements. 27Al chemical shifts were externally referenced to AlF3 at -15.5 ppm.51 NMR data were processed with TopSpin 3.2, and spectral fits were carried out using the dmfit software.52 d. Thermodynamic Calculations: Density functional theory (DFT) was implemented in the Vienna Ab Initio Simulation Package (VASP),53,54 where the PAW method55,56 and the optB86b-vdW functional57–60 were used. Calculations were performed in supercells consisting of 16 transition metal cations and 32 oxygen anions. Converged energy cut-off and k-point densities were used of at least 530 eV and 38 Å, respectively. Structure, input, and result files are publicly accessible online (link provided after initial review or upon request). Results: i.

Overview of Materials

LiM1-yAlyO2 (M = Co, Ni) layered oxide materials where y = 0, 0.05, 0.1, or 0.2 were synthesized in order to investigate the role that Al substitution has on the phase evolution and structural stability of LiCoO2 and LiNiO2 at high degrees of delithiation. Two batches of the LiNi0.8Al0.2O2 chemistry were fabricated. The first batch, denoted as LiNi0.8Al0.2O2 (DO), was chemically disordered as a result of poor transition metal mixing and increased phase segregation while the second was well-ordered with good Ni/Al mixing, and named LiNi0.8Al0.2O2 (O). Results from both LiNi0.8Al0.2O2 materials are presented as their contrast highlights the impact of chemical homogeneity on the structural development and electrochemical performance of these materials. In addition to the Al-substituted LiCoO2 and LiNiO2 materials, LiNi0.8Co0.2O2 and LiNi0.8Co0.15Al0.05O2 were also fabricated. Commercially produced LiCoO2 (cLCO) and LiNi0.8Co0.15Al0.05O2 (cNCA) were also tested to benchmark the quality of the materials synthesized in-house and to evaluate the impact of a contrasting particle size. Figure S1 compares representative FESEM images of cNCA and LiNi0.85Co0.15Al0.05O2 and shows that while both materials have similar primary particle shapes and form aggregates, the synthesized LiNi0.85Co0.15Al0.05O2 has a much smaller average particle size relative to cNCA.



Group 1, which includes cLCO, LiCoO2, LiCo0.95Al0.05O2, LiCo0.9Al0.1O2, and LiCo0.8Al0.2O2, addresses the impact of Al substitution in LiCoO2. Similarly, group 2 addresses the impact of Al substitution in LiNiO2 and thereby consists of LiNiO2, LiNi0.95Al0.05O2, LiNi0.9Al0.1O2, LiNi0.8Al0.2O2 (O), and LiNi0.8Al0.2O2 (DO). Finally, the third group focuses on the impact of Al and/or Co substitution for 20% Ni in a LiNiO2 host. This includes LiNiO2, LiNi0.8Co0.2O2, LiNi0.8Al0.2O2 (O), LiNi0.8Co0.15Al0.05O2, and cNCA. This latter group enables an investigation into the effect of the mixed impact of Ni and Co along with Al substitution. All synthesized and commercial layered oxide materials from groups 1 to 3 were analyzed by XRD (Figure 1) and were observed to be well-crystallized, mainly phase pure (R3m), and with a good (003)/(104) peak ratio (> ~1). The small, broad reflections observed at ~22o and ~26o on all scans are attributed to the grease and Kapton film used to hold the sample. Most materials were > 98% phase pure (Ni-rich) and all of them > 96% as demonstrated in Table SI, which presents the Rietveld calculated phase impurity percentage and the transition metal occupancy in the Li 3a site. Representative measured and Rietveld calculated diffraction profiles are shown in Figure S2. Quantification of the amount of impurity phases were estimated from the Rietveld refinement calculations of the impurity and layered materials. The synthesized Co-rich materials (group 1) were found to only contain Co3O4 as a secondary phase (~3.0-3.9%) with no other impurity phases, signifying that the materials were partially Lideficient.61 All of the Co-rich materials had very low amounts of Co in the Li 3a site, which is typical of layered oxide materials with high Co contents. Of the Ni-rich materials, only LiNi0.8Al0.2O2 (O and DO) had any impurity phases. These two samples had ~1.3-2.0% γ-LiAlO2, which begins to form from αLiAlO2 at temperatures around 750oC.62 The γ-LiAlO2 impurity phase is a byproduct of the phase segregation that is commonly observed when high Al dopant concentrations are used in Ni-rich materials.42


Page 8 of 55

Page 9 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 1: Powder XRD patterns of (a) group 1, (b) group 2, and (c) group 3 materials. 12 h-XRD patterns were recorded at a scan rate of 0.08 o/min over a 2Θ range of 15-70o under ambient conditions. Indexed peaks are associated with the layered (R𝟑m) structure, while the peaks denoted by a * correspond to the Co3O4 (group 1) or γ-LiAlO2 (group 2-3) impurity phases.

The a and c lattice parameters and unit cell volume (Table SI) were derived by Rietveld refinements of the XRD patterns shown in Figure 1 in order to evaluate the impact of all substitutions on the lattice parameters of layered oxide materials. For LiCo1-yAlyO2 materials, the in-plane a lattice parameter



systematically decreased with increasing Al substitution consistent with Vegard’s rule for the substitution of the smaller Al3+ for Co3+, while the interlayer c parameter increased (Figure S3 (a)). Similarly, these trends also held true for the LiNi1-yAlyO2 samples (Figure S3 (b)), excluding LiNi0.8Al0.2O2 (DO), which had poor transition metal mixing (discussed later). The observed impact of Al substitution on the lattice parameters of the cobaltates and nickelates are well supported by the literature.21,34,35,42,45,63 A more detailed explanation of how Al substitution impacts the lattice parameters of LiCoO2 and LiNiO2 can be found in the supplementary information. As previously mentioned, all synthesized materials exhibited good (003)/(104) peak ratios, indicative of little transition metal occupancy in the Li layer determined by Rietveld refinement. While Ni-rich layered oxide materials are usually prone to having divalent Ni ions on the 3a site, the materials fabricated inhouse had relatively low Ni concentrations on the 3a site. The Ni occupancy on the 3a site in LiNi0.8Co0.15Al0.05O2 (0.02502(5)) was similar to cNCA (0.0130(3)). In addition, as opposed to previous reports which typically demonstrated more Ni/Li mixing, the LiNiO2 fabricated herein only had a Ni occupancy of 0.0336(10) in the 3a site. However, both highly substituted LiNi0.8Al0.2O2 materials (Table SI) had significantly more Ni in the Li layer (0.06661(7) and 0.08900(5) in LiNi0.8Al0.2O2 (O) and (DO), respectively), and a clear trend was observed where Al-substitution above 5% in LiNiO2 increased the amount of Ni in the 3a site. The slightly smaller c axis and reduced Ni occupation in the 3a site for LiNi0.95Al0.05O2 relative to LiNiO2 is supported by a steric-driven process discussed in previous findings by Zhong and Sacken.35 In summary, all materials used herein were crystalline, mostly phase pure, and with little transition metal mixing, thus enabling relevant comparisons between materials to be made. ii.

Group 1; LiCo1-yAlyO2 a. Electrochemical Characterization: LixCo1-yAlyO2

An understanding of the phase transitions that occur during delithiation was obtained through multiple electrochemical characterization techniques. Cells comprising LiCo1-yAlyO2 positive electrodes and Li metal negative electrodes were cycled slowly: i) in galvanostatic mode between 4.75 V and 2.75 V at 5 mA/g (g of active material) at 24oC and their resulting voltage and associated differential charge (dQ/dV) profiles are shown in Figure 2 (a), and ii) using a PITT protocol with 10 mV steps and a 1 mA/g current cutoff in the same potential window resulting in the first cycle voltage and current plots depicted in Figure 2 (b) along with their calculated Li+ diffusivities in Figure 2 (c). The voltage and differential charge plots for the unsubstituted cLCO and LiCoO2, shown in the top two panels of Figure 2 (a), initially have a very flat voltage profile at ~3.92 V between lithiation values of 0.99 and 0.74 consistent with a two-phase reaction process. This two-phase reaction is confirmed by the non-Cottrellian current response in the PITT experiments and consistent with the known semiconductor-to-metal transition.4,64 During this transition both materials have slow apparent Li+ diffusivities, in the range of 10-14 to 10-15 cm2/s.


Page 10 of 55

Page 11 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 2: Group 1 LiCoO2-based materials tested vs. Li metal in (a) galvanostatic and (b-c) potentiostatic mode. (a) First charge voltage (black) and differential charge (blue) profiles obtained at 5 mA/g between 4.75 and 2.75 V. First cycle (b) voltage (black) and current (blue) profiles, and (c) the voltage (black) and diffusivity (blue) profiles obtained with 10 mV steps and 1 mA/g current cutoffs. The red boxes in the graphs denote the theoretically inaccessible lithiation regions based on the amount of transition metals that can be electrochemically oxidized in each material.

As these materials (cLCO and LiCoO2) continue to delithiate, the two-phase reaction becomes singlephase which is concurrent with an increase of the voltage and Li+ diffusivity. At 0.55 ≤ x ≤ 0.45, two peaks and a trough were observed in the differential charge profiles which originate from the hexagonal to monoclinic to hexagonal phase Li+ ordering transitions.24,65 Similar to the previous two-phase reaction, longer potentiostatic steps were observed during these phase transitions, and an order of magnitude increase in the Li+ diffusivity was measured. The differential voltage profile exhibits two features, a small peak centered at ~x = 0.3 (~4.47 V) and a much larger peak centered at ~x = 0.2 (~4.54 V) that arises from the monoclinic/hexagonal distortions that are known to occur at 0.21 ≤ x ≤ 0.18.2 Finally, H1-3 staging transitions associated to features at x < 0.2 were observed in both the commercial and in-house LiCoO2, as a sharp peak x = 0.11 (4.62 V) in the former and as a broader feature at x = 0.15 (4.61 V) in the latter. The substitution of even 5% Al for Co in LiCoO2 has enormous ramifications on the material’s structural changes during charging and discharging. While the initial two-phase transition is still present, albeit with lower intensity, the addition of either 5% or 10% Al has nearly eliminated the differential charge peaks associated with the hexagonal to monoclinic to hexagonal ordering transitions. Furthermore, at higher states of charge, LiCo0.95Al0.05O2 and LiCo0.9Al0.1O2 had much broader and less intense differential charge peaks than LiCoO2, with LiCo0.9Al0.1O2 being almost featureless. When 20% of the Co atoms are replaced with Al, all structural transitions become single-phase reactions over the entire delithiation range. This is supported by the lack of clear peaks in the dQ/dV profiles and a Cottrellian response at each potential step for LiCo0.8Al0.2O2. The two-phase reaction observed in LixCoO2 at approximately 0.95 > x > 0.74 was concurrent with the semiconductor-to-metal transformation which was induced by the systematic shrinkage of the MO2 a lattice parameter with Co4+ formation that eventually causes overlap of the transition metal orbitals. The smaller a lattice parameter from the substitution of Al3+ for Co3+ may induce



transition metal orbital overlap and metallic behavior much sooner thereby resulting in the elimination of the two-phase behavior during delithiation. Indeed the a lattice parameter of 2.809 Å for the LiCo0.8Al0.2O2 material was consistent with the approximately 2.81 Å lattice parameter previously observed for the metallic phase for LiCoO2.66 Ultimately, a high concentration of Al3+ may result in semiconducting rather than metallic behavior, but because of other factors such as the differences in the electronic configurations of Al3+ and Co3+ or the impact of Al ions on electronic delocalization, further study would be required. The addition of Al also has a profound impact on the Li+ diffusivity. As more Al was substituted into the structure, the Li+ diffusivity becomes less dependent on the material’s state of lithiation, and was more consistently in the ~10-14 to 10-15 cm2/s range. This is in sharp contrast to LiCoO2, for which the Li+ diffusivity systematically rose from ~10-15 to ~10-13 cm2/s during charging to 4.15 V, and then decreased back to around 10-16 cm2/s after further delithiation. The voltage, current, and diffusivity profiles from the PITT experiments show that the structural transitions that occurred during delithiation were highly reversible upon subsequent discharge despite the near full delithiation state at 4.75 V. Little hysteresis was observed in the PITT voltage profiles, and the difference in diffusivity between charge and discharge at each lithiation value is usually less than an order of magnitude. While some irreversible capacity is expected during the first cycle, especially when charging to near full delithiation at 4.75 V, the amount of irreversible capacity was small. The irreversible capacity was observed to increase significantly and systematically with Al content, especially when 20% Al was substituted for Co. Since there was little change in the average Li+ diffusivity between materials, the increase in irreversible capacity with Al substitution is likely dependent on other factors which will be discussed later. b. Structural Characterization: LixCo1-yAlyO2 In-situ XRD was utilized to provide structural data to complement the high-resolution electrochemical results presented in the previous section. Figure 3 shows the in-situ XRD development of the (003) Bragg reflection (in R3m) as a function of the lithium content for each group 1 material. The (003) reflection is influenced entirely by the interlayer spacing and is indicative of the evolution of the c lattice parameter. To the left of each in-situ XRD contour plot are the voltage and lithiation profiles. The red dashed line superimposed on each sub-figure corresponds to the onset of the 4.75 V constant voltage segment.


Page 12 of 55

Page 13 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 3: In-situ XRD contour plots of the (003) peak over the 2Θ range of 18-20.5° and corresponding electrochemical profiles for (a) cLCO, (b) LiCoO2, (c) LiCo0.95Al0.05O2, (d) LiCo0.9Al0.1O2, and (e) LiCo0.8Al0.2O2. Cells (vs. Li metal) were charged at a constant current of C/50 to 4.75 V, and then subjected to a 4.75 V constant voltage step, while 1 h-XRD scans were recorded concurrently with Δx ≤ 0.02 per scan. The red dashed line superimposed on each sub-figure corresponds to the onset of the 4.75 V constant voltage segment.

Comparison of commercial (Figure 3 (a)) and fabricated in-house (Figure 3 (b)) LiCoO2 revealed similar results with nearly identical changes to the (003) peak during charging, consistent with previously published data and the high-resolution electrochemistry discussed in the previous section. A two-phase region clearly started at ~x = 0.95 and continued until ~x = 0.74, followed by a single-phase reaction which persisted until the phase transformation to and from the monoclinic structure. The calculated c and a lattice parameters from the in-situ XRD experiment for each group 1 material, are shown in Figure 4 (a) and (b), respectively. In order to facilitate the visual analysis of the results, all lattice parameters were determined using a R3m structure, even though the presence of the monoclinic phase was confirmed. The calculated lattice parameters confirmed the existence of the initial two-phase region, and quantified the changes to the unit cell as a result of the hexagonal to monoclinic to hexagonal transformations. During charging to ~x = 0.45 the (003) reflection systematically moved toward lower angles, indicative of an expanding c axis. When delithiated beyond ~x = 0.45 the c axis contracted, and at x ≤ 0.3, the rate of contraction in the c axis accelerated rapidly. At very high states of charge (x ≤ 0.2) a sudden shift of the (003) peak towards higher angles was observed. The stepwise shift in the (003) reflection was attributed



to staging reactions, where some of the Li layers of the R3m structure were completely emptied. Based on the peak positions at the end of charging, both types of LiCoO2 were fully delithiated and converted to the O1 phase (AB stacking). In agreement with previously reported results, during delithiation the a lattice parameter (Figure 4 (b)) of cLCO and LixCoO2 contracts until ~x = 0.5 and then systematically expands.2,4,67 When highly delithiated, but prior to the staging phase transitions, the a parameter for cLCO and LiCoO2 was roughly the same as in the initial fully lithiated state. The structural changes observed by in-situ XRD results on cLCO and LiCoO2 are in excellent agreement with our differential charge and PITT cells shown in Figure 2, as well as with previously published studies.2,4,67

Figure 4: (a) c and (b) a lattice parameters plotted as a function of delithiation for cLCO (black), LiCoO2 (red), LiCo0.95Al0.05O2 (blue), LiCo0.9Al0.1O2 (green), and LiCo0.8Al0.2O2 (orange) positive electrodes derived from in-situ XRD experiments. Cells (vs. Li metal) were charged at a constant current of C/50 to 4.75 V, and then subjected to a 4.75 V constant voltage step, while 1 hXRD scans were recorded concurrently with Δx ≤ 0.02 per scan. Lattice parameters were calculated from a rhombohedral unit cell using the location of the (003) and (101) peaks. Note that a break was included in the y-axis of plot (b).

Similar to the differential charge and PITT results, in-situ XRD shows that when Al was substituted into LiCoO2, the structural transitions throughout delithiation were significantly reduced. While the initial two-phase transition was still present when 5% or 10% Al was substituted into LiCoO2, the two phases were less distinguishable in LiCo0.95Al0.05O2 than in LiCoO2. When 20% Al was substituted into LiCoO2 the initial delithiation reaction appeared to be single-phase. The percentage change of the c and a lattice parameters for the group 1 materials, shown in Figure S4 (a, d), clearly demonstrates that as Al was added to the positive electrode material the lattice parameters changed less during delithiation. This trend was especially apparent as the materials approached higher degrees of delithiation. Even 5% Al was sufficient to prevent the stepwise shift in the (003) reflection that was observed for LiCoO2 near full delithiation. Thus, substituting 5% Al into LiCoO2 was effective at preventing the formation of the O1 structure and in minimizing the staging reactions. The absence of these structural transitions will reduce the lattice strain when accessing such high states of charge. LiCo0.9Al0.1O2 is a particularly important sample, since it showed that Al substitution can prevent the staging transitions even when the material was sufficiently delithiated. Comparison of the c lattice parameters (Figure 4 (a)) for LiCo0.9Al0.1O2 and LiCoO2 shows that 10% Al substitution completely averted the staging reactions even though the materials were at similar states of delithiation. LiCo0.8Al0.2O2 presents the clearest example of how significant Al substitution can be in suppressing the structural changes during delithiation. Due to the presence of 20%


Page 14 of 55

Page 15 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Al and ≥ 20% Li in LiCo0.8Al0.2O2, even when fully charged, all phase transitions during charging were gradual and single-phase, and there was relatively little change in the lattice parameters compared to LiCoO2. Additional PITT cells, identical to those previously reported, were charged to 4.75 V and then disassembled in an Ar-filled glovebox. Ex-situ XRD samples prepared from the positive electrodes retrieved from these cells present the advantage of much better signal-to-noise ratios than the in-situ experiments due of the absence of the Be window. Figure S5 depicts the XRD results of the (003) and (104) peaks associated with the 18-21.5o and 42-47o 2 regions respectively, for all group 1 materials charged to 4.75 V under PITT conditions. The ex-situ findings confirm the conclusions derived from the in-situ experiments, demonstrating the Al-substitution limits the degree of structural transformation at high states of charge. The corresponding voltage profiles are presented in Figure S6 (a), and a brief discussion of the results is given in the supplemental information. The electrochemical results for the 4.75 V ex-situ XRD cells including the final open circuit voltage (OCV) after a 1 h-rest, the final lithiation, and the percentage of the Ni and Co (% Ni/Co) that were oxidized during charging are presented in Table SII for each positive electrode material. The OCV obtained 1 h after the end of charge gives insight into how the transition metal chemistry affects the quasiequilibrium voltage of each material. Increasing Al substitution is expected to systematically lower the maximum degree of delithiation due to the decrease of the amount of Co available for oxidation; however, it is important to note that Co is still fully oxidized. The percentage of Co that was oxidized during charging to 4.75 V is a useful measure of the state of charge of the material. The percentage of Co oxidation for the Al-substituted group 1 materials was similar to but slightly lower than LiCoO2. Despite this, all of the Al-substituted samples showed a very significant increase in the OCV after relaxation, relative to LiCoO2. This can be attributed to the reduced potential associated with re-intercalation of Li into O1 (vs. O3)-type electrodes, and provides further evidence that the O1 structure did not form in any of the Al-substituted materials.2 Operando isothermal microcalorimetry experiments enabled the measurement of the heat flux as a function of charge. Cells placed in the microcalorimeter were cycled between 4.75 V and 2.75 V at a constant current of 25 mA/g at 30oC. The resulting voltage and heat flux profiles for all group 1 materials are presented in Figure 5. As charging began, both LiCoO2 materials had a small endothermic heat flux. Two exothermic peaks found at ~4.10 V (~x = 0.55) and 4.28 V (~x = 0.45) arose from the ordering transition from and to the hexagonal structure. With further delithiation, the topotactic reaction became gradually more exothermic. The heat flux profiles for cLCO and LiCoO2 during delithiation are consistent with previously reported results.68–71 Interestingly, as Li+ was initially reinserted back into the fully charged cLCO and LiCoO2 materials, the heat flux remained exothermic, which could be a result of the heat generated by the sluggish Li+ diffusivity and/or heat from parasitic reactions at such high states of charge. When the lithiation of cLCO and LixCoO2 was between 0.25 and ~0.5, the heat flux became endothermic because of the entropic changes to the host material. Based on the heat flux peaks in the range of ~x = 0.45-0.6, it is evident that the hexagonal to monoclinic to hexagonal phase transitions were reversible. As expected, at lower states of charge the Li+ insertion reaction was exothermic and of approximately the same magnitude as the endothermic deintercalation reactions at equivalent states of charge. At the end of discharge, the heat flux became very exothermic as an abundance of irreversible heat was generated by the low Li+ diffusivity of the material (Figure 2 (c)) and high internal resistance.



Figure 5: First cycle operando isothermal microcalorimetry voltage (black) and heat flux (blue) measurements for each group 1 material. Cells (vs. Li metal) were cycled at a constant current of 25 mA/g between 4.75 V and 2.75 V, while the chamber was isothermal at 30oC.

The bottom three panels of Figure 5 display the voltage and heat flux profiles for LiCo0.95Al0.05O2, LiCo0.9Al0.1O2, and LiCo0.8Al0.2O2, respectively. Instead of showing distinct peaks related to the monoclinic phase, the heat fluxes for all Al-substituted LiCoO2 materials gradually increased from endothermic to exothermic upon delithiation. Since the monoclinic phase changes observed in LiCoO2 are eliminated in the Al-substituted materials, their heat flux profiles did not exhibit the monoclinic transition heat signature but instead simply became steadily more exothermic during charge, analogous to LiCoO2. Operando microcalorimetry confirmed the results previously shown by the differential charge, PITT, and in-situ XRD experiments that the substitution of even 5% Al into LiCoO2 prevented the hexagonal to monoclinic transitions. c. Cycling Performance: LixCo1-yAlyO2 Of critical importance to the analysis of the structural stability of these positive electrode materials near full delithiation is the impact of the structural transitions on the electrochemical performance. Cells based on each positive electrode material were cycled versus Li metal by charging at 20 mA/g to Vmax (4.5 or 4.75 V), maintaining a constant potential until the current dropped below 10 mA/g, and then discharging


Page 16 of 55

Page 17 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


at 10 mA/g to 2.75 V. Cycle life performance for all group 1 positive electrode materials is depicted in Figure 6 for both cycling cut-off voltages of 4.5 V (a) and 4.75 V (b). Upon charging up to either 4.5 V or 4.75 V, the discharge capacity of cLCO was slightly higher than that of LiCoO2. This result was likely associated to the higher electrode-electrolyte reactivity of LiCoO2 stemming from its higher surface area. The increased magnitude of the electrode-electrolyte reactions on LiCoO2 would also explain the larger difference in discharge capacities between cLCO and LiCoO2 at 4.75 V, when electrolyte decomposition accelerates. Table SIII provides the first cycle charging and discharging capacity as well as the irreversible loss for each material in Figure 6.

Figure 6: Cycling performance plotted as discharge capacity (mAh/g) vs. cycle number for all group 1 LiCoO2-based materials charged to (a) 4.5 V or (b) 4.75 V (vs. Li metal). Duplicate cells were charged at 20 mA/g to Vmax (4.5 or 4.75 V), held at constant potential until the current dropped below 10 mA/g, and then discharged at 10 mA/g to 2.75 V, all at room temperature.

The addition of Al to LiCoO2 was found to have a significant deleterious effect on both the charge and discharge capacity regardless of the cutoff potential. The capacities obtained are well below the expected lower theoretical capacities associated with substitution of electrochemically inactive Al. Furthermore, Table SIII shows a clear relationship between the irreversible capacity loss in the first cycle and the amount of Al in the positive electrode material. These results are in good agreement with the literature.72 Since all the synthesized materials have similar surface areas, the decrease in discharge capacity and increase in irreversible loss cannot be primarily attributed to surface area effects. Figure S7 (a-e) shows the 1st, 5th, 10th and 15th cycle voltage profiles for group 1 materials. From the voltage profiles of cLCO (Figure S7 (a)) and LiCoO2 (Figure S7 (b)) it is clear that while all of the phase transitions were completely reversible on the first cycle, they rapidly become less distinct or non-existent with cycling. By the 5th cycle, the signature hexagonal to monoclinic to hexagonal transformations were already absent from the voltage profile. The voltage profiles were particularly useful for understanding the impedance development during cycling, which was determined by the polarization and hysteresis in the potential curves. Both cLCO and LiCoO2 had only moderate impedance growth in the first 15 cycles, and the rate of impedance growth appeared to be slowing down as cycling progressed. However, as the Al content in the electrode material increased, so did the total impedance after 15 cycles. Moreover, materials with higher Al substitution such as LiCo0.9Al0.1O2 and LiCo0.8Al0.2O2 continued to have rapid impedance growth even after 10 cycles. The amplified impedance growth that scales with the magnitude of the Al



substitution into LiCoO2 was largely responsible for the poor electrochemical cycling performance of the Al-substituted electrodes. Figure S8 (a, d) shows the percentage discharge capacity retention when charged to 4.5 and 4.75 V, respectively. The capacity retention results make evident that Al substitution into LiCoO2 had a profound negative impact on the material’s capacity retention, especially when charged to 4.75 V. d. Dissolution: LixCo1-yAlyO2 As previously reported, at high states of charge and elevated temperature, HF etching of the positive electrode material accelerates, and transition metal dissolution from the positive electrode becomes prevalent leading to a porous surface.23 To investigate the impact of Al substitution on the surface chemical stability, additional cells were fabricated and charged to 4.75 V at 25 mA/g and held at 4.75 V for 10 h at 60oC. All positive materials were tested in duplicate by ICP-OES and the resulting average percentage Ni and Co dissolution amounts are presented in Table I. It is important to note that the ICPOES measurements only determine the amount of Ni and Co dissolved from the positive electrode and then reduced on the negative electrode. It does not quantify any of the dissolved transition metals still in the electrolyte when the cell was disassembled.


Page 18 of 55

Page 19 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Table I: Ni and Co metal dissolution measurements by ICP-OES based on a Li metal negative electrode and separator retrieved from LixCo1-yAlyO2, LixNi1-yAlyO2, LixNi0.8Co0.2O2, LixNi0.8Co0.15Al0.05O2, Co3O4, NiO and baseline cells (1 ≥ x ≥ 0 and 0.2 ≥ y ≥ 0). Cells (vs. Li metal) were charged to 4.75 V at 25 mA/g, held at 4.75 V for 10 h, at 60oC. The percentage Ni and Co dissolution is based on the quantity of each metal in the positive electrode material. The Ni and Co dissolution from the baseline samples was below the detection limit of the instrument.

Despite being held at 4.75 V and 60oC for only 10 h, significant Co dissolution was measured for all group 1 materials. LiCoO2 (5.24%) lost substantially more of its initial Co than cLCO (3.83%), because of its higher electrode-electrolyte interfacial area. 5% Al substitution for Co in LiCo0.95Al0.05O2 (2.80%) enabled a 47% reduction in Co dissolution. Additional Al substitution in LiCo0.9Al0.1O2 (2.48%) and LiCo0.8Al0.2O2 (1.77%) was effective in further reducing the Co dissolution significantly. A Co3O4 reference material was found to have exceedingly low (0.05%) Co dissolution, indicating the impressive stability of Co in the cobalt spinel and inferior stability of the layered oxides in such an aggressive environment. This result is consistent with that reported for cobalt spinel coated LiCoO2.73 The Co dissolution was directly related to the corrosion current, which resulted from parasitic electrodeelectrolyte reactions during the 4.75 V potentiostatic segment. Figure S8 shows the current profiles for the ICP-OES cells for all materials. The increase in current during the potentiostatic step, as clearly evidenced in the LiCoO2 cells (Figure S8 (a)), was rooted in the exothermic electrode-electrolyte reactions and transition metal dissolution that were extensively described in a previous report.11 As a



result of its lower surface area, less parasitic current was measured for cLCO relative to LiCoO2, which also supports its lower Co dissolution. The corrosion currents were observed to decrease substantially as the concentration of the Al in LiCo1-yAlyO2 increased, which is also consistent with the dissolution results. Even though Al substitution into LiCoO2 had many beneficial effects such as suppressed phase and structural changes, improved average Li+ diffusivities, and reduced Co dissolution at high states of charge, the Al-substituted LiCoO2 materials had substantially increased first cycle irreversible losses and worse cycling performance. This poor cycling stability occurs despite the Al-substituted materials being sufficiently oxidized, and the irreversible loss not being attributed to the reaction kinetics. As a result, the source of the degradation in electrochemical performance was likely of electronic or internal structural origin, and will be elaborated upon in the discussion section. iii.

Group 2; LiNi1-yAlyO2 a. Electrochemical Characterization: LixNi1-yAlyO2

Group 2 materials comprising LiNiO2, LiNi0.95Al0.05O2, LiNi0.9Al0.1O2, LiNi0.8Al0.2O2 (O), and LiNi0.8Al0.2O2 (DO), were investigated in a systematic way similar to the group 1 LiCoO2-based materials in order to achieve an in-depth understanding of the role of Al on the structural transformations and subsequent cycling performance on LiNiO2. Figure 7 presents the dQ/dV and PITT results obtained with group 2 LiNiO2-based materials while the group 1 results were shown in Figure 2.

Figure 7: Group 2 LiNiO2-based materials tested vs. Li metal in both (a) galvanostatic and (b-c) potentiostatic mode. (a) First charge voltage (black) and differential charge (blue) profiles obtained at 5 mA/g between 4.75 and 2.75 V. First cycle (b) voltage (black) and current (blue) profiles, and (c) the voltage (black) and diffusivity (blue) profiles were obtained with 10 mV steps and 1 mA/g current cutoffs. The red boxes in the graphs denote the theoretically inaccessible lithiation regions based on the amount of transition metals that can be electrochemically oxidized in each material.

Consistent with earlier reports, the delithiation of LiNiO2 proceeded within a single-phase as evidenced by the broad dQ/dV peak centered at x = 0.9 in the top panel of Figure 7 (a).3,74,75 Similar to LiCoO2, during this highly lithiated region, the Li+ diffusivity (Figure 7 (c)) of LiNiO2 was relatively low (~10-15 to 10-16 cm2/s). The LixNiO2 structural transition to the monoclinic phase began at ~x = 0.75 (3.76 V) and is


Page 20 of 55

Page 21 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


shown by the wide differential charge peak at x = 0.65. In contrast to LiCoO2, the monoclinic phase in LiNiO2 is stable for a much larger lithiation range. The transformation back to a rhombohedral structure is represented by the dQ/dV peak at ~x = 0.41 (4.01 V). This single rhombohedral phase was stable until x = 0.24 (4.21 V), where a two-phase reaction is obtained. Throughout the monoclinic and single rhombohedral phase structures the Li+ diffusivity of LiNiO2 was high (~10-13 to 10-14 cm2/s), which was a result of the ample lithium vacancies and expanded c axis. The transition to this phase was enormously sluggish with Li+ diffusion decreasing by almost two orders of magnitude. At the very end of the charging process, a final peak in the differential charge plot was observed at ~x = 0.06. This peak could be indicative of the ongoing transition to the fully delithiated structure. All structural transitions in LiNiO2 were highly reversible as demonstrated by the very small hysteresis in the PITT voltage profile, and plateaus at similar lithiation contents and potentials. The LiNiO2 phase transitions documented here are in good agreement with previous observations.3,7,25,74 Similar to the LiCo1-yAlyO2 (y > 0) materials, the substitution of Al for the Ni in LiNiO2 decreased the intensity of the phase transitions and smoothed the voltage profile. However, in the 5% Al-substituted material, LiNi0.95Al0.05O2, there were still very distinct differential charge peaks associated with each phase transition, and the voltage profile of the PITT cell remained highly faceted. As with LiNiO2, the final phase transition in LiNi0.95Al0.05O2 was accompanied by a very significant decrease in diffusion, but in contrast to LiNiO2, this transition seemed to occur with greater hysteresis. As more Al was substituted into the LixNiO2 structure, the lower state of charge (x > 0.2) phase transitions became less apparent even though they were within the smaller delithiation range induced by the Al substitution. Interestingly, as 5 and 10% Al was substituted into LiNiO2 the intensity of the final dQ/dV peak grew substantially, but this peak was not visible for LiNi0.8Al0.2O2 (O). As previously indicated this peak was correlated to the final delithiation of the material and not only did the magnitude of this peak change with the amount of Al substitution, but so did its position: x = 0.06 (4.61 V) for LixNiO2, x = 0.10 (4.64 V) for LixNi0.95Al0.05O2, x = 0.18 (4.63 V) for LixNi0.9Al0.1O2, x = 0.19 (4.71 V) for LixNi0.8Al0.2O2 (DO), but absent in LiNi0.8Al0.2O2 (O). The shift in the lithiation at which this final dQ/dV peak occurs was directly related to the amount of Al substitution. When there was less Ni in the structure as a result of Al substitution, full oxidation of the positive electrode occurred at lower lithiation values since Al is electrochemically inactive. In agreement with previous findings, the amount of Al shifts the voltage profile higher so that the potential at which this transition occurs is correlated to the amount of Al in the structure.22,35,45 We specifically find this increase in slope and average voltage with Al substitution readily apparent at x < 0.3. The absence of this final delithiation peak in LiNi0.8Al0.2O2 (O) but its presence in LiNi0.8Al0.2O2 (DO) may be related to the homogeneity of the Ni and Al mixing. At the end of charging, the homogeneous LiNi0.8Al0.2O2 (O) material still retained ~20.8% of its initial Li in the structure. As a result, LiNi0.8Al0.2O2 (O) was not sufficiently delithiated for the final transition to be observed. Conversely, in LiNi0.8Al0.2O2 (DO) the transition metal mixing was poor so the material was comprised of “Ni-rich” and “Al-rich” domains. Thus while the “Ni-rich” domains were capable of some phase transition, the “Al-rich” domains were not, and the dQ/dV plot, which is an averaged measurement of the material, only exhibits a small peak. TEM nano-EDS mapping confirmed poor Ni and Al mixing in LiNi0.8Al0.2O2 (DO) (Figure S10 (bd)) as opposed to the homogeneous distribution of LiNi0.8Al0.2O2 (O) (Figure S10 (f-h)). Like LiNiO2, the LiNi0.95Al0.05O2 and LiNi0.8Al0.2O2 (O) materials exhibit very little hysteresis under PITT cycling conditions. While the poorly mixed LiNi0.8Al0.2O2 (DO) sample was expected to show a substantial voltage hysteresis, it was surprising to find a larger hysteresis in LiNi0.9Al0.1O2 than in



LiNi0.8Al0.2O2 (O). Such a result suggests that LiNi0.9Al0.1O2 may have also formed some “Ni-rich” and “Al-rich” domains from an imperfect synthesis, although not nearly to the same degree as LiNi0.8Al0.2O2 (DO). When homogeneous transition metal mixing occurred, the addition of Al to LiNiO2 did not negatively impact the Li+ diffusivity. As such, LiNi0.95Al0.05O2 and LiNi0.8Al0.2O2 (O) showed similar average Li+ diffusivities as LiNiO2. In fact, the LiNi0.8Al0.2O2 (O) sample exhibited diffusivities almost an order of magnitude larger than LiNiO2 towards the end of charge. b. Structural Characterization: LixNi1-yAlyO2 To confirm the difference in structural transitions suggested by the electrochemical characterization, group 2 materials were characterized by in-situ XRD with the resulting evolution of the (003) peak upon charging presented in Figure 8 and the derived c and a lattice parameters in Figure 9. In agreement with the differential charge and PITT results just presented, the in-situ XRD analysis of LiNiO2 (Figure 8 (a)) did not show the development of a two-phase structure in the initial stages of the delithiation process. Instead, the delithiation reaction appeared to be topotactic until the monoclinic phase started forming at approximately x = 0.75. While the structural change to and from the monoclinic phase was not easily apparent from the (003) peak contour plot over the 2Θ range of 18-20.5°, an increase in the (003) peak width was observed when 0.75 ≥ x ≥ 0.36. This supports the lithiation range over which the monoclinic region was observed in the dQ/dV profile. Near the end of charge (x = 0.16), a two-phase region appeared as the delithiation reaction started to create empty lithium planes. As LiNiO2 proceeded to higher states of charge, the intensity of the new peak centered at ~19.8o increased substantially while the original (003) peak centered at ~18.6o gradually disappeared. When LiNiO2 approached full delithiation, the material became single-phase and exhibited a contracted c axis stabilized at 13.39 Å (Figure 9). The relative change to the c axis was much smaller for LiNiO2 (~5.64%, Figure S4 (b)) than for LiCoO2 (~9.25%, Figure S4 (a)). Regardless, the in-situ XRD results for LiNiO2 at the end of charge were best fit to a refined NiO2 material with either a R3m or C2/m structure, but could not be definitively distinguished because of the insufficient scan resolution to observe peak splitting. The C2/m structure had been previously associated to NiO2.67,76 Unlike the CdI2 (P3m1) structure that was achieved by CoO2, the octahedral environment of the Li site in the C2/m structure does not share faces and as a result more closely resembled the R3m (O3) structure. As such, this octahedral environment can stably host Li+ or even Ni ions, in sharp contrast to the P3m1 structure. The ability of this octahedral environment to host some cations eliminates the harsh two-phase transition that occurs when intercalating Li back into the fully delithiated P3m1 (e.g. CoO2) structure, and would reduce the capacity degradation associated with that transition. An elaboration on the differences between the C2/m, R3m, and P3m1 structures is provided in the supplementary information section and shown in Figure S11.


Page 22 of 55

Page 23 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 8: In-situ XRD contour plots of the (003) peak over the 2Θ range of 18-20.5° and corresponding electrochemical profiles for (a) LiNiO2, (b) LiNi0.95Al0.05O2, (c) LiNi0.9Al0.1O2, (d) LiNi0.8Al0.2O2 (O), and (e) LiNi0.8Al0.2O2 (DO). Cells (vs. Li metal) were charged at a constant current of C/50 to 4.75 V, and then subjected to a 4.75 V constant voltage step, while 1 h-XRD scans were recorded concurrently with Δx ≤ 0.02 per scan. The red dashed line superimposed on each sub-figure corresponds to the onset of the 4.75 V constant voltage segment.

Compared to LiNiO2, the Al-substituted LiNiO2 materials had broader peaks, smoother phase transformations and smaller changes to the lattice parameters. The in-situ XRD results obtained with LixNi0.95Al0.05O2 (Figure 8 (b)) appeared similar to LixNiO2 when x > 0.3, but at higher states of charge the (003) reflection in LixNi0.95Al0.05O2 had a gradual and continuous shift to higher angles. Indeed, as more Al was substituted into LiNiO2, the peak shift representing the structural transition towards the highly delithiated phase decreased in magnitude and occurred more progressively. The 5% Al substitution was sufficient to prevent any ambiguity over the formation of a C2/m phase as obtained in delithiated LiNiO2, and instead a highly delithiated phase retaining the O3 R3m structure was clearly observed. Figure 9 (a) clearly shows that as Al was added to LiNiO2 the host materials began to contract in the c-direction at higher states of charge as opposed to going through a distinct two-phase reaction. The relative change of the c axis depicted in Figure S4 (b) decreased steadily as more Al was substituted into the structure. As such, the final single-phase c spacing of the Al-substituted materials was markedly larger than that of pure LiNiO2, likely due to the presence of more residual Li remaining in LiNi1-yAlyO2 (y > 0) at the end of charge.



The a lattice parameter of the fully lithiated materials was observed to be highly dependent on the Al concentration, and as more Ni was replaced with Al, the a parameter systematically decreased. However, Figure S4 (e) shows that during delithiation the relative change to the a parameter was not strongly related to the concentration of Al in the host material. Since LiNi0.8Al0.2O2 (O) was homogeneously mixed the a lattice parameter was significantly smaller than for LiNi0.9Al0.1O2, but for LiNi0.8Al0.2O2 (DO), the a parameter was larger than for LiNi0.9Al0.1O2 because of the presence of “Ni-rich” and “Al-rich” domains. In distinct contrast to the LiCoO2-based materials, none of the group 2 LiNiO2-based materials had an increase in the a lattice parameter during charging. Instead, the a parameter for all of the Al-substituted LixNiO2 electrodes decreased until ~x = 0.2 and then remained constant.

Figure 9: (a) c and (b) a lattice parameters plotted as a function of delithiation for LiNiO2 (black), LiNi0.95Al0.05O2 (red), LiNi0.9Al0.1O2 (blue), LiNi0.8Al0.2O2 (O) (green), and LiNi0.8Al0.2O2 (DO) (orange) positive electrodes derived from in-situ XRD experiments. Cells (vs. Li metal) were charged at a constant current of C/50 to 4.75 V, and then subjected to a 4.75 V constant voltage step, while 1 h-XRD scans were recorded concurrently with Δx ≤ 0.02 per scan. Lattice parameters were calculated from a rhombohedral unit cell using the location of the (003) and (101) peaks.

As previously performed with the group 1 materials, group 2 samples were PITT charged to 4.75 V and disassembled for ex-situ XRD analysis. The charging voltage profiles for these materials are shown in Figure S6 (b), while the corresponding ex-situ scans obtained over the (003) and (104) peak regions are presented in Figure 10. While the addition of 5% Al into LiNiO2 did not impact charge capacity significantly (Table SII), it affected sensitive structural features such as the (003) and (104) peaks that shifted to lower angles compared to LiNiO2. As observed with the in-situ XRD samples, the shifts of the LiNiO2 reflections to higher angles were the result of the transition to a C2/m structure that occurred at the highest states of charge. At higher Al concentrations such as LiNi0.9Al0.1O2 and LiNi0.8Al0.2O2, the (003) and (104) peaks remained at lower angles compared to the non-substituted LiNiO2 material at the end of charge. These results support the relationship determined by the in-situ experiments. As more Al was added into the LiNiO2 structure the degree of structural transformation in the positive electrode material diminished, even at equivalent states of delithiation. The OCV 1 h after being PITT charged to 4.75 V, presented in Table SII, demonstrates that a strong relationship exists between the quantity of the Al ions and the OCV. In agreement with the group 1 materials, the OCV voltage after a 1 h-rest increased with increasing Al-content in LiNi1-yAlyO2. As with the group 1 materials, no trend relating the amount of Al in the positive electrode structure and the Ni oxidation percentage could be determined. This could be


Page 24 of 55

Page 25 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


attributed to the non-negligible amount of parasitic reactions that were occurring on these materials at such high states of charge, and explain why the Ni oxidation percentage of some of these materials were observed to exceed the theoretical limit.

Figure 10: Ex-situ XRD scans of the positive electrodes materials charged to 4.75 V (vs. Li metal) under PITT conditions (10 mV step, 1 mA/g cutoff) in the regions of the (003) (18-21.5o 2Θ) and (104) (42-47o 2Θ) peaks for LiNiO2 (black), LiNi0.95Al0.05O2 (red), LiNi0.9Al0.1O2 (blue), LiNi0.8Al0.2O2 (O) (green), and LiNi0.8Al0.2O2 (DO) (orange). All cells were disassembled in an Ar-filled glovebox where the XRD samples were sealed with Kapton film to prevent contamination from the ambient atmosphere.

The potential and heat flux profiles of group 2 materials determined by operando microcalorimetry are displayed in Figure 11. In contrast to group 1, the LiNiO2-based group 2 materials exhibited an exothermic heat flux as charging began. However, LixNiO2 did become slightly endothermic at ~x = 0.75, which was related to the beginning of the transformation to the monoclinic phase. Although the heat flux peaks associated with the hexagonal to monoclinic to hexagonal phase transformations were of much weaker intensity in the case of LiNiO2 than LiCoO2, they were still visible. The less distinct transitions to and from the monoclinic phase for LiNiO2 were in agreement with the previous dQ/dV, PITT (Figure 7), and in-situ XRD results (Figure 8). For all Al-substituted LiNiO2 materials, the heat flux remained exothermic throughout the charge process to 4.75 V with no features related to the structural transformations to the monoclinic phase. The region during charging where LiNi0.9Al0.1O2 was temporarily endothermic was due to a small temperature deviation (< 0.1oC) in the microcalorimeter chamber and was not associated with the electrode material. At higher potentials (> 4.3 V), the heat flux of all LiNiO2-based materials rapidly became more exothermic, and continued to increase until the end of charge. The only exception to this rule was the LiNi0.8Al0.2O2 (O) sample which maintained a very low exothermic reaction for the majority of the delithiation and finally resulted in a slight (approx. 50% less than other samples) exothermic rise at the end of charge towards 4.75 V. Knowing that the ionic transport was improved within the relevant region of lithiation in LiNi0.8Al0.2O2 (O), we can state that a majority of the exotherm was related to kinetics. Throughout the entirety of the discharge to 2.75 V, all group 2 materials remained exothermic, although the LiNi0.8Al0.2O2 (O) sample much less so. Since the entropic changes of the positive electrode should be the dominant influence on the heat flux, the heat flux during



discharge is expected to be equal and opposite to the heat flux during charge, so some of the heat generated may be related to parasitic reactions or impedance.

Figure 11: First cycle operando isothermal microcalorimetry voltage (black) and heat flux (blue) measurements are presented during the first cycle for each group 2 material. Cells (vs. Li metal) were cycled at a constant current of 25 mA/g between 4.75 V and 2.75 V, while the chamber was isothermal at 30oC.

c. Cycling Performance: LixNi1-yAlyO2 The electrochemical cycling performance was also evaluated for the group 2 positive electrode materials. The discharge capacities plotted as a function of cycle number are shown in Figure 12, and the associated first cycle charge/discharge capacities as well as the irreversible capacity loss are provided in Table SIV. For cells charged to 4.5 V and 4.75 V, the first cycle discharge capacity decreased systematically with increasing Al content. In contrast to the LiCo1-yAlyO2 materials, the first cycle irreversible capacity loss (Table SIV) did not uniformly increase with higher Al substitution. While LiNi0.95Al0.05O2 did exhibit a higher first cycle irreversible capacity loss than LiNiO2 with a 4.5 V cutoff potential, a steady decrease in the irreversible capacity loss was observed as the Al-content was increased beyond 5%. In contrast, with a 4.75 V cutoff potential the first cycle irreversible capacity losses for LiNi0.95Al0.05O2 and LiNi0.8Al0.2O2 (O) were lower than for LiNiO2, but LiNi0.9Al0.1O2 and LiNi0.8Al0.2O2 (DO), which are both somewhat inhomogeneous, had increased capacity losses.


Page 26 of 55

Page 27 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 12: Cycling performance plotted as discharge capacity (mAh/g) vs. cycle number for all group 2 LiNiO2-based materials charged to either (a) 4.5 V or (b) 4.75 V (vs. Li metal). Duplicate cells were charged at 20 mA/g to Vmax (4.5 or 4.75 V), held at constant potential until the current dropped below 10 mA/g, and then discharged at 10 mA/g to 2.75 V, all at room temperature.

As expected, LiNiO2 demonstrated very poor cycling stability with either 4.5 or 4.75 V cutoff potentials, as shown in Figure 12. The electrochemical instability of LiNiO2 when cycling to high potentials was likely caused by the intense two-phase reaction that occurs at high states of charge, which results in a significant incoherence of strain along the phase boundaries. The substantial improvement in capacity retention even at lower Al-substitutions is consistent with the elimination of this two-phase region. Beyond the two phase reaction, loss of oxygen could occur at the surface resulting in: 1) a more resistive cathode-electrolyte interface layer, and 2) the subsequent propensity for divalent Ni atoms to migrate into the 3a site. Such developments would further impede Li+ diffusion. It is evident from the percentage capacity retention results in Figure S8 (b, e) that all Al-substituted LiNiO2 materials were markedly more stable and had substantially better capacity retention than LiNiO2. Interestingly, little difference was observed in the capacity retention at either cutoff potential between LiNi0.95Al0.05O2, LiNi0.9Al0.1O2, and LiNi0.8Al0.2O2 (O), which indicates that only enough Al to eliminate the two-phase reaction towards C2/m was necessary. In contrast to LiCoO2, the voltage profile of LiNiO2 remained highly faceted even after 15 cycles to 4.75 V (Figure S12). Furthermore, LiNiO2 had negligible impedance development after the first cycle, and unlike Al-substituted LiCoO2, substituting Al for Ni in LiNiO2 actually decreased the amount of impedance development. The 15th cycle voltage profile of LiNi0.8Al0.2O2 (O) to 4.75 V (Figure S12 (d)) looks nearly identical to the 5th cycle, indicating the stability of the electrode’s impedance with extended cycling. The minimal impedance development on LiNi0.95Al0.05O2, LiNi0.9Al0.1O2 and LiNi0.8Al0.2O2 (O) electrodes explains why those materials had much better cycling stability. While LiNi0.8Al0.2O2 (DO) experiences large changes to its voltage profile after the first cycle, the general shape of the voltage profile remains, and there was little impedance growth after the 5th cycle. This was in sharp contrast to LiCo0.8Al0.2O2 (Figure S7 (e)), which continued to have substantial impedance development after its 5th cycle. d. Dissolution: LixNi1-yAlyO2



Similar to LiCoO2, Al substitution into LiNiO2 was found to stabilize the structure against HF attack and impede Ni dissolution. Table I lists the average percentage Ni dissolution from each group 2 material after charging to 4.75 V followed by a constant voltage step of 10 hours at 60oC. Remarkably, LiNiO2 (0.97%) had a substantially lower transition metal dissolution than LiCoO2 (5.24%). This was particularly noteworthy since NiO loses 10.21% of its Ni under identical conditions and highly delithiated LiNiO2based materials are known to transform into a rocksalt-like structure that is Ni2+ rich at the surface.68,77,78 Similar to the group 1 LiCoO2-based materials, as Al was substituted into LiNiO2 the magnitude of Ni dissolution generally decreased. The average percentage of Ni dissolution in LiNi0.95Al0.05O2, LiNi0.9Al0.1O2, and LiNi0.8Al0.2O2 (O) positive electrodes amounted to 0.82%, 1.01%, and an amazingly low 0.28%, respectively. Only LiNi0.9Al0.1O2 and LiNi0.8Al0.2O2 (DO), which had an average Ni dissolution of 1.01% and 2.08%, respectively, did not follow the trend of decreasing Ni dissolution with increasing Al content. Because these two materials have “Ni-rich” domains, with low Al contents, it was expected that they would have more Ni dissolution than some of the other LiNi1-yAlyO2 (y > 0) materials. However it was unexpected that LiNi0.8Al0.2O2 (DO) would have a significantly higher Ni dissolution rate than LiNiO2, suggesting that somehow the inhomogeneous Ni/Al mixing actively increased the propensity for Ni dissolution. Unlike the LiCoO2-based materials, none of the group 2 materials had a noteworthy increase in current during the 10 h-potentiostatic segment. However, the corrosion current results in Figure S9 (b) show a subtle trend that, as more Al was substituted into LiNiO2, the parasitic current decreased. Both the dissolution results and the corrosion current data indicate that Co was more catalytic towards the electrolyte and more likely to dissolve than Ni. Similar to the LiCoO2-based materials, Al-substitution into LiNiO2 suppressed phase and structural transformations, reduced the variability of the Li+ diffusivity, and stabilized the structure against HF etching. In stark contrast to Al substitution for Co in LiCoO2, which was found to significantly reduce discharge capacity, adding Al to LiNiO2 had beneficial effects on the cycling stability by minimizing the capacity fade and impedance development in LiNiO2. iv.

Group 3; 20% Substituted LiNiO2

Presently, most state of the art positive electrodes for secondary Li-ion batteries consist of Ni-rich layered oxide materials.6 These materials have a LiNiO2 base structure with varying amounts of Al, Co, or Mn substituted for Ni. Among the most commonly used compounds are LiNi0.8Co0.15Al0.05O2 or LiNixMnyCozO2 (where x + y + z = 1).1,6 For this study, LiNi0.8Co0.15Al0.05O2 was investigated as a model commercial positive electrode material with the objective of understanding the role that both Co and Al have in improving its electrochemical performance. As previously indicated, cNCA (commercially produced LiNi0.8Co0.15Al0.05O2) was used to benchmark the LiNi0.8Co0.15Al0.05O2 sample that was synthesized in-house. LiNi0.8Co0.2O2 and LiNi0.8Al0.2O2 (O), which have 20% of the Ni substituted for either Co or Al, were studied alongside cNCA, LiNi0.8Co0.15Al0.05O2, and LiNiO2, and together make up the group 3 materials. a. Electrochemical Characterization: 20% Substituted LiNiO2


Page 28 of 55

Page 29 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 13: Group 3 LiNiO2-based materials tested vs. Li metal in both (a) galvanostatic and (b-c) potentiostatic mode. (a) First charge voltage (black) and differential charge (blue) profiles obtained at 5 mA/g between 4.75 and 2.75 V. First cycle (b) voltage (black) and current (blue) profiles, and (c) the voltage (black) and diffusivity (blue) profiles obtained with 10 mV steps and 1 mA/g current cutoffs. The red boxes in the graphs denote the theoretically inaccessible lithiation regions based on the amount of transition metals that can be electrochemically oxidized in each material.

As discussed in the previous section and presented again in the top panel of Figure 13 (a), the first charge differential profile for LiNiO2 exhibited several well-defined dQ/dV peaks. In contrast, all group 3 substituted LiNiO2-based materials presented smoother differential profiles. While the LiNi0.8Co0.2O2 and LiNi0.8Al0.2O2 (O) profiles in the second and third panels of Figure 13 (a) seem devoid of any evidence for distinct phase transitions during delithiation, a close evaluation of the results revealed the existence of 5 and 4 small dQ/dV features, respectively. The first two peaks, which are the most intense, in all 20%substituted group 3 materials are associated with the initial single-phase delithiation of the material (1 > x > 0.9) and the transition to the monoclinic structure (0.75 > x > 0.65), similar to LiNiO2. The third dQ/dV peak, related to the transition back to the rhombohedral phase, of the substituted materials located at ~x = 0.43 (3.98 V) for 20% Co and at ~x = 0.50 (4.02 V) for 20% Al is much less intense than in the unsubstituted LiNiO2 at ~x = 0.41 (4.01 V), shown in panels 2, 3 and 1 respectively of Figure 13 (a). In short, the 20% Co and Al substitution into LiNiO2 caused a drastic decrease in the intensity of the peak associated with the monoclinic to rhombohedral phase transition, suggesting that substitution renders such a structural transition less favorable. In addition, substitution was also observed to shift the transition peaks to a more lithiated state relative to LiNiO2. Upon further delithiation, the group 3 materials exhibit another differential peak indicative of the transition to the hexagonal phase. These peaks were observed at ~x = 0.28 (4.19 V) in LixNi0.8Co0.2O2, ~x = 0.34 (4.23 V) in LixNi0.8Al0.2O2 (O), and at ~x = 0.24 (4.21 V) in LixNiO2 (Figure 13 (a)). As with the monoclinic to rhombohedral transition, the dQ/dV peaks representative of the transition to the hexagonal phase were much less intense for the substituted materials and were shifted to higher lithiations compared to LiNiO2, with the 20% Al-substituted material showing the highest Li-content at each transformation. Since Co can be oxidized and theoretically all of the Li in LixNi0.8Co0.2O2 can be extracted, a small feature in the dQ/dV profile related to the final delithiation process was measured at ~x = 0.10 (4.64 V). Such a feature was absent in LixNi0.8Al0.2O2 (O) since Al is electrochemically inactive and therefore the material can not be fully delithiated. As observed for the



monoclinic and hexagonal transitions, this final process occurred at a higher lithium content when 20% Co was substituted for Ni relative to LiNiO2 with x = 0.10 versus x = 0.06. In addition, the final reaction occurred at a higher potential upon Co substitution (4.64 V) compared to LiNiO2 (4.61 V). Excluding the final delithiation dQ/dV peak, which was not observed in LiNi0.8Al0.2O2 (O) since the material could not be completely delithiated, the substitution of either 20% Al or Co for Ni caused less defined structural transitions relative to LiNiO2 and induced these transitions in a more lithiated state. The potential, current and Li+ diffusivity profiles obtained and derived from PITT evaluation of the group 3 materials are shown in Figure 13 (b-c). While LixNiO2 exhibited a relatively flat voltage at low states of charge above x ≥ 0.75 and at high states of charge over 0.20 ≥ x ≥ 0.06, both LiNi0.8Co0.2O2 and LiNi0.8Al0.2O2 (O) had more sloped profiles and a typical Cottrellian response at each potential step indicative of single-phase processes. In addition, LiNi0.8Co0.2O2 and LiNi0.8Al0.2O2 (O), similar to LiNiO2, revealed very low hysteresis in the first cycle, showing that the reactions occurring during charge are highly reversible. Finally, substituting 20% Co or Al for Ni improved the Li+ diffusivity, particularly in the lithiation ranges where two-phase transformations were observed in LiNiO2. In conclusion, PITT results clearly support the dQ/dV conclusions that all substitutions into LiNiO2 dramatically reduced the intensity of the phase transformations, led to low hysteresis, and improved Li+ diffusivity. LiNi0.8Co0.15Al0.05O2, which is isostructural to LiNiO2 but has a mixture of Co and Al substituted for 20% of the Ni, was expected to show properties between that of LiNi0.8Co0.2O2 and LiNi0.8Al0.2O2 (O). The differential charge plots for cNCA and LiNi0.8Co0.15Al0.05O2, presented in Figure 13 (a), are very similar to the materials obtained with 20% Co or 20% Al substitution, and clearly different from LiNiO2. Like LiNi0.8Co0.2O2 and LiNi0.8Al0.2O2 (O), the cNCA and LiNi0.8Co0.15Al0.05O2 materials also revealed smooth, gradual voltage profiles. In addition, both cNCA and LiNi0.8Co0.15Al0.05O2 exhibited 5 weak dQ/dV peaks associated with the same structural transitions as those described for LiNi0.8Co0.2O2, which occurred through single-phase reaction processes as evidenced by the Cottrellian current response demonstrated throughout the entire lithiation range, analogously to LiNi0.8Co0.2O2. While cNCA and LiNi0.8Co0.15Al0.05O2 had similar voltage profiles indicating that both materials are chemically comparable, the intensity of the highly delithiated differential peak was much larger for cNCA than for either LiNi0.8Co0.15Al0.05O2 or LiNi0.8Co0.2O2, which impacted Li+ diffusivity as discussed below. A slightly larger voltage hysteresis was observed for LiNi0.8Co0.15Al0.05O2 relative to cNCA, and was attributed to the higher surface area of LiNi0.8Co0.15Al0.05O2 which caused more polarization from increased electrode-electrolyte reactions. While the Li+ diffusivities for cNCA and LiNi0.8Co0.15Al0.05O2 generally follow the same trends, on average the Li+ diffusivity for LiNi0.8Co0.15Al0.05O2 was incrementally lower than cNCA, which is also associated with the differences in the particle size of the materials. Major differences between the Li+ diffusivities of cNCA and LixNi0.8Co0.15Al0.05O2 arose when 0.3 > x > 0.05. The Li+ diffusivity of LixNi0.8Co0.15Al0.05O2 began to decrease at a lower state of charge (~x = 0.3) than cNCA (~x = 0.1). Additionally, at ~x = 0.15 the Li+ diffusivity of LixNi0.8Co0.15Al0.05O2 suddenly decreased, while it increased for cNCA at the same lithiation. This contrast is clearly correlated to the larger feature in the dQ/dV profile of cNCA at x = ~0.11 relative to the corresponding peaks for LiNi0.8Co0.15Al0.05O2 or LiNi0.8Co0.2O2. As the Ni-Co-Al ternary materials approached full delithiation (x < 0.1), they all showed similar decreases in the Li+ diffusivity, indicating that as the remaining Li ions become scarce the Li+ diffusivity declines irrespective of the specific transition metal chemistry or particle size.


Page 30 of 55

Page 31 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


The LiNi0.8Co0.15Al0.05O2 material, like LiNi0.8Co0.2O2 and LiNi0.8Al0.2O2 (O), had suppressed structural transformations and less variation in the Li+ diffusivity compared to LiNiO2. The differential charge and PITT results for LiNi0.8Co0.15Al0.05O2 were similar to LiNi0.8Co0.2O2, with the main difference being that LiNi0.8Co0.15Al0.05O2 was slightly more lithiated at the end of charge because of the 5% Al substitution. The Li+ diffusivity over 1 > x > 0.8 was much lower for LixNiO2 than for any of the other group 3 materials, indicating the Al and Co substitution into LiNiO2 improved the Li+ diffusivity even at low states of charge. b. Structural Characterization: 20% Substituted LiNiO2 In-situ XRD results for all group 3 LiNiO2-based materials upon charge up to and at 4.75 V are presented in Figure 14. The calculated c and a lattice parameters are shown in Figure 15, while the percentage change in the lattice parameters are depicted in Figure S4 (c, f). Each 20%-substituted LiNiO2 material proceeded via a single-phase reaction throughout the entire lithiation range. While phase evolutions were observed in every material at high states of charge, the structural transformations were always gradual with no evidence of two-phase transitions. Because of its larger particle size, cNCA (Figure 14 (e)) exhibited narrower and more intense diffraction reflections with a better signal-to-background ratio compared to LiNi0.8Co0.15Al0.05O2 (Figure 14 (d)). Despite cNCA having a slightly smaller a axis than the more nanostructured LiNi0.8Co0.15Al0.05O2, the variation of the a lattice parameter during delithiation was nearly identical for both LiNi0.8Co0.15Al0.05O2 materials. Qualitatively, the change in the c axis during charging was consistent between cNCA and LiNi0.8Co0.15Al0.05O2, but the magnitude of the expansion and contraction was much larger for cNCA (Figure S4 (c)). At ~x = 0.4, when the c lattice parameters were the largest, the c parameter for cNCA expanded by 2.30%, while that of LiNi0.8Co0.15Al0.05O2 only increased by 1.11%. At full delithiation, the c parameter for cNCA decreased by 5.56% but only by 4.35% for LiNi0.8Co0.15Al0.05O2. The percentage change of the c lattice parameter was smaller for LiNi0.8Co0.15Al0.05O2 than for cNCA because either i) LiNi0.8Co0.15Al0.05O2 had more Ni in the 3a site, which made the structure more rigid or ii) the true degree of delithiation was slightly less in LiNi0.8Co0.15Al0.05O2 relative to cNCA. The c lattice parameters for each electrode, displayed in Figure 15 (a), emphasize the differences of the Al and Co substitution in LiNiO2. When fully charged, the 20% Al substitution resulted in a c lattice contraction of only 2.09%, whereas the 20% Co led to a 5.48% decrease. Substitution comprising both 5% Al and 15% Co in LiNi0.8Co0.15Al0.05O2 revealed a 4.35% contraction of the c parameter, falling between the results for LiNi0.8Al0.2O2 (O) and LiNi0.8Co0.2O2. As such, a comparison of the LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2 materials revealed that Al ions were more efficient at minimizing the c axis structural change than Co ions, even at the same degree of delithiation. At lower states of charge (x > 0.2) little difference was observed in the percentage change of the a lattice parameter (Figure S4 (f)) between cNCA, LixNi0.8Co0.15Al0.05O2, and LixNi0.8Co0.2O2 during delithiation. While the higher Al content in LiNi0.8Al0.2O2 (O) did cause a smaller a parameter, the relative change compared to LiNi0.8Co0.15Al0.05O2 or LiNi0.8Co0.2O2 was minimal and there was no fundamental change in the trend of the a lattice parameter during delithiation.



Figure 14: In-situ XRD contour plots of the (003) peak over the 2Θ range of 18-20.5° and corresponding electrochemical profiles for (a) LiNiO2, (b) LiNi0.8Co0.2O2, (c) LiNi0.8Al0.2O2 (O), (d) LiNi0.8Co0.15Al0.05O2, and (e) cNCA. Cells (vs. Li metal) were charged at a constant current of C/50 to 4.75 V, and then subjected to a 4.75 V constant voltage step, while 1 h-XRD scans were recorded concurrently with Δx ≤ 0.02 per scan. The red dashed line superimposed on each sub-figure corresponds to the onset of the 4.75 V constant voltage segment.

While the Co-substituted materials comprising cNCA, LixNi0.8Co0.15Al0.05O2, and LixNi0.8Co0.2O2 sustained an a lattice expansion at very high states of charge at x < ~0.15, the 20% Al-substituted LiNi0.8Al0.2O2 (O) did not present any evidence for such an occurrence, as shown in Figure 15 (b). Even though LiNi0.8Al0.2O2 (O) was fully oxidized, it may not have been sufficiently delithiated for the a parameter to increase. However, none of the LiNi1-yAlyO2 materials evaluated above that could reach such high states of charge (Figure 9 (b)) exhibited a similar expansion during delithiation. In contrast, all of the LixCo1-yAlyO2 materials (Figure 4 (b)) displayed an increasing a axis when x < ~0.5. For cNCA, LiNi0.8Co0.15Al0.05O2, and LiNi0.8Co0.2O2, the lithiation point at which the a axis began to expand appears to be correlated to the amount of Co in the electrode. Due to their lower redox potential relative to Co, Ni ions should have been oxidized to 4+ before the oxidation of Co ions to 4+.29 Thus, the expansion of the a lattice parameter seems to only occur when Co was being oxidized, which is in agreement with the a lattice parameter results of the group 1 and 2 materials. The in-situ XRD results of LixNi0.8Co0.15Al0.05O2 and LixCo1/3Ni1/3Mn1/3O2 by Yoon et al. also show an expansion of the a parameter at higher states of


Page 32 of 55

Page 33 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


charge, and that the increase of the a parameter occurs at lower states of charge for LixCo1/3Ni1/3Mn1/3O2, than for LixNi0.8Co0.15Al0.05O2 because of its higher Co content.41

Figure 15: (a) c and (b) a lattice parameters plotted as a function of delithiation for LiNiO2 (black), LiNi0.8Co0.2O2 (red), LiNi0.8Al0.2O2 (O) (blue), LiNi0.8Co0.15Al0.05O2 (green), and cNCA (orange) positive electrodes derived from in-situ XRD experiments. Cells (vs. Li metal) were charged at a constant current of C/50 to 4.75 V, and then subjected to a 4.75 V constant voltage step, while 1 h-XRD scans were recorded concurrently with Δx ≤ 0.02 per scan. Lattice parameters were calculated from a rhombohedral unit cell using the location of the (003) and (101) peaks.

As performed for group 1 and 2 materials, cells comprised of each group 3 material were charged to 4.75 V under PITT conditions and examined by ex-situ XRD in order to corroborate the in-situ XRD results and improve the scan resolution. The (003) and (104) peak regions of the resulting ex-situ XRD scans at the end of charge are presented in Figure 16. LiNiO2 and LiNi0.8Co0.2O2, which have no Al in their structures and therefore could be fully delithiated, had (003) and (104) reflections at the highest angles of all group 3 materials. Similar to LiNiO2, LiNi0.8Co0.2O2 was indexed to a NiO2-type material in either the R3m or C2/m space groups at the end of charge. Although LixNi0.8Co0.2O2 retained less Li in its structure than the LixNiO2 sample at the end of charge (Table SII, x = 0.015 vs. x = 0.047), its (003) and (104) peaks were still shifted to lower angles compared to LixNiO2 (Figure 16). Such a finding confirms that substitution of Co for Ni induced a contraction in the c lattice parameter and helped resist large, deleterious changes to the unit cell, without preventing full delithiation of the material. In agreement with the conclusions drawn from the previous sections, Al substitution limited the structural transformations that occurred at the highest states of charge and minimized the collapse of the c parameter at full delithiation. cNCA and LiNi0.8Co0.15Al0.05O2, which substituted 5% Al for Co in LiNi0.8Co0.2O2, had peak positions at slightly lower angles than LiNi0.8Co0.2O2. In the LiNi0.8Al0.2O2 (O and DO) samples, where all of the Co in LiNi0.8Co0.2O2 was exchanged for Al, the peak positions were shifted even further to the lowest angles of the group 3 materials. In short, the shift of the (003) and (104) peaks towards the lower angles scaled with the Al-content in the group 3 LiNiO2-based materials. Finally, all Al-substituted materials in group 3 formed a partially delithiated R3m structure at full state of charge.



Figure 16: Ex-situ XRD scans of positive electrode materials charged to 4.75 V (vs. Li metal) under PITT conditions (10 mV step, 1 mA/g cutoff) in the regions of the (003) (18-21.5o 2Θ) and (104) (42-47o 2Θ) peaks for LiNiO2 (black), LiNi0.8Co0.2O2 (red), LiNi0.8Al0.2O2 (O) (blue), LiNi0.8Co0.15Al0.05O2 (green), and cNCA (orange). All cells were disassembled in an Ar-filled glovebox where the XRD samples were sealed with Kapton film to prevent contamination from the ambient atmosphere.

Similar to the group 1 and 2 materials, the OCV results after PITT charging to 4.75 V (Table SII) gave insight into the effect of each transition metal on the electrode’s stability at high states of charge and the final equilibrium redox potential. All group 3 materials quickly achieved stable OCV potentials above 4.4 V, indicative that appreciable amounts of the layered structure did not degrade into other structures (e.g. spinel or rocksalt). The OCV of LiNi0.8Co0.2O2 (4.65 V) was substantially higher than LiNiO2 (4.47 V) which was likely due to the presence of Co in the material, thus enabling the material to oxidize to higher potentials than Ni and thereby increasing the equilibrium voltage. As previously discussed, the OCV of LiN1-yAlyO2 rose with increasing Al content. At a given capacity, a higher redox potential is desirable since it increases the energy density of the cell. A comparison of LiNi0.8Co0.2O2, LiNi0.8Co0.15Al0.05O2, and LiNi0.8Al0.2O2 demonstrated that substituting Ni with Co instead of Al could increase the OCV even further. In particular, LiNi0.8Co0.2O2 had an incrementally higher OCV (4.65 V) than LiNi0.8Co0.15Al0.05O2 (4.64 V) despite being at a slightly lower state of charge, and had a substantially higher OCV than LiNi0.8Al0.2O2 (O) (4.61 V). Thus for LiNiO2-based electrodes at similar states of charge, Co ions were more effective in raising the OCV than Al ions, and as a result of the higher potential as a function of lithiation, the Co ions increased the energy density of the LiNiO2-based materials. The operando microcalorimetry results of the group 3 materials are presented in Figure 17. Similar to the Al-substituted LiNiO2 materials, cNCA, LiNi0.8Co0.15Al0.05O2, and LiNi0.8Co0.2O2 were exothermic when charging began and showed no heat flux peaks related to the monoclinic phase. All group 3 materials had very little hysteresis in the heat flux profiles, which is in agreement with the effect of Al substitution observed in the microcalorimetry results of the group 1 and 2 materials. The lack of observable phase changes or hysteresis in the heat flux profiles supports the conclusion that Co and Al substitution into LiNiO2 suppressed phase changes within the electrode material. In general, the heat flux profiles for cNCA, LiNi0.8Co0.15Al0.05O2, and LiNi0.8Co0.2O2 were always exothermic, and in good agreement with the


Page 34 of 55

Page 35 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


LiNiO2-based group 2 materials. The negative heat flux region between 0.85 ≥ x ≥ 0.54 during the discharge of LixNi0.8Co0.15Al0.05O2 was due to a small drift in the heat flux measurement and was not representative of the material showing endothermic behavior. Of particular note was that LiNi0.8Co0.2O2 exhibited a much lower exothermic signal at the end of charge than all the corresponding synthesized (high surface area) materials. The only material which was comparable was the much lower surface area cNCA.

Figure 17: First cycle operando isothermal microcalorimetry voltage (black) and heat flux (blue) measurements for each group 3 material. Cells (vs. Li metal) were cycled at a constant current of 25 mA/g between 4.75 V and 2.75 V, while the chamber was isothermal at 30oC.

c. Cycling Performance: 20% Substituted LiNiO2 To fully understand the impact that Al and Co substitution into LiNiO2 had on the materials’ electrochemical capabilities at high states of charge, all group 3 materials were cycled identically to the previously discussed group 1 and 2 materials. The cycling capacity and percentage capacity retention results of the group 3 materials are shown in Figure 18 and Figure S8 (c, f), respectively. The first cycle charge capacity, discharge capacity, and irreversible loss are presented in Table SV. When 4.5 V was used as the upper cutoff potential, LiNiO2 had the highest first cycle discharge capacity, but realized very poor cycling stability. The first cycle discharge capacities for LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2 (~



210 mAh/g) were higher than for cNCA (~ 201 mAh/g). However, as is evident from Figure 18 (a), the discharge capacities for LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2 decreased during the first few cycles, while the discharge capacity of cNCA actually increased slightly. As a result, for the majority of the cycles evaluated herein the discharge capacity of cNCA is ~14 mAh/g higher than LiNi0.8Co0.15Al0.05O2 and ~22 mAh/g above LiNi0.8Co0.2O2. The difference in cycling behavior between cNCA and LiNi0.8Co0.15Al0.05O2 during the first few cycles could be attributed to the difference in the surface areas of the two materials, and is discussed in more detail later. With a 4.5 V potential limit, cNCA and LiNi0.8Co0.15Al0.05O2 showed similar capacity loss rates after the first 5 cycles, which was associated to the identical chemistries of the two materials. Upon increase of the upper cut-off voltage to 4.75 V (Figure 18 (b)), cNCA initially exhibited a higher discharge capacity but also a much more rapid capacity loss than LiNi0.8Co0.15Al0.05O2. As a result of its ability to undergo full Li deintercalation, LiNi0.8Co0.2O2 had higher first cycle charge and discharge capacities than LiNi0.8Co0.15Al0.05O2, however, the cycling stability of LiNi0.8Co0.15Al0.05O2 was slightly better than LiNi0.8Co0.2O2 in both potential windows due to the structural benefits of Al substitution. In agreement, LiNi0.8Al0.2O2 (O) showed the lowest initial discharge capacity of the group 3 materials which is attributed to the electrochemical inactivity of the 20% Al, but among the best cycling stability, especially at 4.75 V (Figure S8 (f)).

Figure 18: Cycling performance plotted as discharge capacity (mAh/g) vs. cycle number for all group 3 LiNiO2-based materials charged to either (a) 4.5 V or (b) 4.75 V (vs. Li metal). Duplicate cells were charged at 20 mA/g to Vmax (4.5 or 4.75 V), held at constant potential until the current dropped below 10 mA/g, and then discharged at 10 mA/g to 2.75 V, all at room temperature.

LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2, which both had higher surface areas than cNCA, likely experienced significantly more parasitic electrode-electrolyte reactions on the first charge relative to cNCA. These parasitic surface reactions caused the development of a larger cathode-electrolyte interface and increased the charge transfer impedance. The impact of these reactions is evident by the higher irreversible loss (Table SV) and lower initial discharge capacities for LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2. The decreased amount of parasitic reactions on cNCA had a beneficial effect on the material’s discharge capacity and cycling stability during the first few cycles. At the same time, because cNCA had a larger particle size than the synthesized materials it was more susceptible to fracturing,


Page 36 of 55

Page 37 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


especially when subjected to the enormous structural changes upon charging to 4.75 V. cNCA also had a more intense pseudo-phase transition at high potentials (> 4.5 V) than LiNi0.8Co0.15Al0.05O2, which likely contributed to particle fracturing. As the material fractured, new surfaces were created, which in turn, parasitically reacted with the electrolyte. Thus, a larger change in the c axis is expected to degrade the cycling stability of the material since there would repeatedly be increased strain on the lattice. When charged to 4.5 V, cNCA and LiNi0.8Co0.15Al0.05O2 had similar relative contractions in the c lattice parameter (Figure S4), and nearly identical cycling stabilities. In contrast, when the materials were charged to 4.75 V, cNCA had a substantially larger change in the c axis compared to LiNi0.8Co0.15Al0.05O2. The larger structural variations occurring in cNCA upon cycling could explain its poorer cycling stability relative to LiNi0.8Co0.15Al0.05O2. In summary, while the difference in the initial electrochemical performance between cNCA and LiNi0.8Co0.15Al0.05O2 can be attributed to the effect of the parasitic surface reactions, the discrepancy in the extended cycling stability was mainly determined by the structural changes within each material. Finally, the impedance development for each positive electrode material can be determined from the changes to the voltage profiles during the first 15 cycles (Figure S13). cNCA, LiNi0.8Co0.15Al0.05O2, LiNi0.8Co0.2O2, and LiNi0.8Al0.2O2 (O) showed suppressed impedance development upon cycling compared to LiNiO2 with little impedance growth after the first cycle and then nearly no change to the voltage profile after the 5th cycle. The reduction in impedance growth of the substituted group 3 materials confirmed the stabilizing role that Al and Co ions had on the LiNiO2 structure. d. Dissolution: 20% Substituted LiNiO2 The Ni and Co dissolution results obtained by ICP-OES for the group 3 materials, presented in Table I, show the effect of Al and Co substitution on the magnitude of the transition metal dissolution. In contrast to the much lower dissolution observed for 20% Al substitution into LiNiO2 (0.28% Ni), substituting 20% Co for Ni was found to have significantly increased the amount of transition metal dissolution in LiNi0.8Co0.2O2 (1.75% Ni, 1.86% Co) relative to LiNiO2 (0.97% Ni). cNCA (0.82% Ni, 0.68% Co) had much lower Ni and Co dissolution than LiNi0.8Co0.15Al0.05O2 (1.78% Ni, 1.77% Co) most likely because of its significantly lower surface area. Indeed, LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2, which had similar surface areas, exhibited nearly identical magnitudes of Ni and Co dissolution. In agreement with the previously discussed results, it seems that Co catalyzes dissolution regardless of whether it is the sole transition metal (5.24% Co dissolution in LiCoO2), substituted into LiNiO2, or into a mixture with both Ni and Al ions. From the three materials with both Ni and Co ions, it is evident that there was no significant preferential dissolution between Ni and Co. Interestingly, although 5% Al substitution into LiCoO2 or LiNiO2 was very effective in lowering the dissolution rates of Co and Ni, it was not beneficial when substituting for Co in LiNi0.8Co0.2O2. This may be related to differences in surface chemistry which were not explored in this study. The corrosion current results, presented in Figure S8 (c), act as an electrochemical measure of the transition metal dissolution. As previously shown, LiNiO2 and LiNi0.8Al0.2O2 had no significant amounts of corrosion current during a 10 h-constant voltage step at 4.75 V at elevated temperature. In contrast, LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2 revealed a small increase in current during the constant potential segment relative to the expected exponential decay of the current profile. Such increase could be associated to parasitic reactions with the electrolyte that were induced by the Co ions in both materials.



Indeed, the group 1 and 2 materials demonstrated that only the Co-rich materials exhibited a noteworthy increase in corrosion current, while the Ni-rich materials, even without any Al substitution, did not have a significant increase in parasitic current. The substantially higher corrosion current as well as the increased amounts of Ni and Co dissolution in LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2 compared to LiNiO2 or LiNi0.95Al0.05O2, suggested that Co substitution into Ni-rich materials degrades the material’s resistance to HF attack. In Co-substituted Ni-rich materials, not only did the Co ions dissolve into the electrolyte, but the presence of Co seemed to increase the propensity for the Ni ions to dissolve as well. The higher Ni dissolution in LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2 relative to LiNiO2 and LiNi0.95Al0.05O2 may have been from the increased surface porosity or degraded surface structure that was likely to occur with Co dissolution. Finally, the smaller increase in corrosion current under the potentiostatic conditions found in cNCA relative to LiNi0.8Co0.15Al0.05O2 was most likely correlated to the substantially lower surface area of cNCA. Discussion: i.

Influence of Al:

As was clearly shown with the group 1 and 2 materials, Al substitution into LiCoO2 or LiNiO2 had a very significant impact on preventing phase transformations and in reducing the structural changes to the electrode materials during delithiation and subsequent lithiation. Suppression of the structural transitions and decreased variation of the lattice parameters during charging and discharging was directly correlated to an increasing amount of Al in the electrode material. At full states of charge, the 5% Al in LiCo0.95Al0.05O2 was sufficient to prevent the formation of the O1 phase that was observed for cLCO and LiCoO2, and preserve the R3m structure. Similarly, 5% Al substitution in LiNiO2 was sufficient to eliminate the two-phase transition and maintain the R3m structure at a full state of charge. When more Al was substituted into either LiCoO2 or LiNiO2, the phase transformations became less distinct, and the crystal structure showed smaller changes in the lattice parameters and unit cell volume during delithiation. Moreover, LiCoO2, LiNiO2, and the 20% Al-substituted versions of each base material were discharged down to 0.5 V (not shown) and the Al ions caused no additional phase transformations and did not impede the eventual conversion transformations to Co/Ni metal and Li2O. As a result of the beneficial effects of the Al substitution on the phase progression of LiCo1-yAlyO2 and LiNi1-yAlyO2, less variability in the Li+ diffusivity as a function of Li+ content in Al-substituted materials was observed, especially when lower Al concentrations were used. Additionally, Al substitution substantially reduced the magnitude of the deleterious transition metal dissolution that occurs when the electrode material was near full delithiation (Table I) for both the LiCoO2 and LiNiO2-based materials. Enhanced electrochemical performance was expected for the Al-substituted LiCoO2 materials relative to LiCoO2 upon cycling at high state of charge based on the favorable impact of Al substitution on the electrode’s structural and chemical stability. In contrast, Figure 6 clearly shows that replacing Al for Co did not improve the electrochemical performance of the electrode, and instead increased the first cycle irreversible capacity loss, reduced discharge capacity and resulted in poor cycling for all the LiCo1-yAlyO2 compositions investigated. The negative effect was found to scale with the Al content in LiCoO2. As such, LiCo0.8Al0.2O2 demonstrated the lowest discharge capacity with rapid capacity fade upon cycling to 4.5 and 4.75 V. These results were surprising since LiCo0.8Al0.2O2 was found to be structurally very stable compared to LiCoO2, with unit cell dimensions relatively invariant, while LiCo0.95Al0.05O2 and


Page 38 of 55

Page 39 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


LiCo0.9Al0.1O2 still had substantial changes to the a and c lattice parameters during charging (Figure S4 (a, d)). While the cycling stability of LiCo0.8Al0.2O2 was very poor at either cutoff potential it was markedly worse when charged to 4.75 V, indicating a high voltage degradation mechanism. However, no structural decomposition was evidenced by in-situ XRD or PITT analysis when LiCo0.8Al0.2O2 was charged to 4.75 V. Furthermore, the first cycle Li+ diffusivity in LiCo0.8Al0.2O2 was similar to the other Co-rich materials, suggesting that the material was not initially kinetically limited. Thus, the significant irreversible capacity that developed during cycling must be a result of changes to the LiCo0.8Al0.2O2 and other LiCo1-yAlyO2 materials. When charged and held at 4.75 V under aggressive thermal conditions, the LiCo1-yAlyO2 materials also had substantially less transition metal dissolution (Table I) or corrosion current (Figure S8 (a)) than LiCoO2. To determine if the capacity degradation was related to electrode reactions with the electrolyte salt, LiCo0.8Al0.2O2 electrodes were also cycled with a 1M LiBF4 EC:DMC (1:1) electrolyte to be compared with 1M LiPF6 EC:DMC (1:1) as shown in Figure S14. The electrolyte salt had minimal impact on the resulting electrochemical performance of LiCo0.8Al0.2O2. In both electrolytes, LiCo0.8Al0.2O2 electrodes revealed high first cycle irreversible capacity (> 60 mAh/g), and poor cycling stability. Regardless of electrolyte, all LiCo0.8Al0.2O2 cells had a discharge capacity below 100 mAh/g by the 5th cycle. The LiBF4 electrolyte is known to contain substantially less H2O, which subsequently degrades into HF, than its LiPF6 analogue.11,79 Moreover, the BF4- anion is less easily degraded than PF6-.80 Since a similar cycling performance for LiCo0.8Al0.2O2 was observed for both salts, the acid-induced degradation mechanism of the positive electrode was eliminated as the cause of the poor electrochemical capabilities during the initial cycles. In summary, these results indicate that the degradation of the electrochemical performance when Al was added to LiCoO2 is not from transition metal dissolution of the positive electrode and subsequent poisoning of the negative electrode, or from increased electrolyte decomposition. While structural degradation for any of the LiCo1-yAlyO2 materials was not observed on the first cycle, there was evidence of increased impedance evolution (Figure S7) upon additional cycling. Moreover, a systematic increase of the irreversible capacities for LiCo1-yAlyO2 as a function of Al content on the first cycle was observed. These large irreversible capacities were correlated with the subsequent severe impedance growth of these materials during cycling (Figure S7 (e)). However, it is critical to point out that the high irreversible losses across the entire Al-substituted LiCo1-yAlyO2 series could not be recovered even when cycled at extremely slow, near equilibrium PITT conditions. This suggests that the Alsubstituted LiCoO2 materials underwent some permanent transformation during cycling that had profound negative consequences on the materials’ fundamental electrochemical characteristics. In sharp contrast, when Al was added into LiNiO2 the cycling stability improved markedly (Figure 12) with no increase in irreversible loss on the first cycle, consistent with the decreased transition metal dissolution and the suppressed phase transformations observed. After only 3 cycles, LiNi0.95Al0.05O2 had a higher discharge capacity and lower capacity degradation rate than LiNiO2, when cycling with an upper cutoff potential of 4.75 V. Even increased Al concentrations (10 and 20%), which substantially decreased the electrode’s theoretically obtainable capacity, were observed to have better electrochemical performance than LiNiO2 when charged to 4.75 V. While LiNi0.8Al0.2O2 (O) initially had a lower capacity than LiNiO2, after only 17 cycles LiNi0.8Al0.2O2 (O) maintained a higher discharge capacity and greater stability with additional cycling. Furthermore, Al substitution into LiNiO2 was very effective in reducing the impedance development (Figure S12) and the Ni dissolution from the fully delithiated electrodes (Table I). ii.

Influence of Co:



As observed with Al, Co substitution for Ni increased the average voltage, suppressed phase transitions, and reduced the degree of structural transformations during charging, which in turn, improved the electrochemical performance of Co-substituted LiNiO2 at high states of charge. At 100% Co substitution for Ni, which is equivalent to LiCoO2, the final oxidation plateau was ~0.3 V above that for LiNiO2, and even 20% Co substitution increased the plateau voltage noticeably relative to pure LiNiO2. In addition, Co-substitution enabled full delithiation, in contrast to substituting LiNiO2 with electrochemically inactive Al. As a result, much higher discharge capacities were obtained with LiNi0.8Co0.2O2 electrodes relative to LiNi0.8Al0.2O2 (O). Furthermore, the cycling stability was significantly better for LiNi0.8Co0.2O2 than LiNiO2, and after 20 cycles of charging up to 4.75 V, the discharge capacity of LiNi0.8Co0.2O2 (199 mAh/g) was much higher than that of LiNiO2 (176 mAh/g). The improved electrochemical performance of LiNi0.8Co0.2O2 at high states of charge may be attributed to the decreased structural transformations caused by the Co ions. Co substitution was effective in both suppressing phase transformations and in minimizing the percentage change to the lattice parameters (i.e. lattice strain), which in turn lowered the impedance development during cycling. The cumulative effect of these two phenomena was that the LiNi0.8Co0.2O2 material was more stable at high states of charge and could reversibly cycle in this region substantially better than LiNiO2, despite the fact that the substitution of Co markedly increased the degree of transition metal dissolution. iii.

Influence of Co and Al:

Comparison of LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2 gave direct insight into the role of 5% Al substitution on the electrochemical performance of layered oxides with both Ni and Co ions. Despite being electrochemically inactive and thus preventing the extraction of all of the lithium in the electrode, 5% Al substitution into LiNi0.8Co0.2O2 enabled the increase of the discharge capacity and the improvement of the cycling stability slightly. The in-situ XRD results, displayed in Figure 14 and Figure 15, show a smaller change in unit cell parameters in LiNi0.8Co0.15Al0.05O2 than LiNi0.8Co0.2O2 upon charge to 4.75 V. This enhanced structural stability may be the key reason why the cycling performance of LiNi0.8Co0.15Al0.05O2 incrementally outperformed LiNi0.8Co0.2O2 at high potentials. While the cycling stability of layered oxide electrodes near full state of charge was dictated by multiple factors, some aspects were more influential than others. The structural changes that occurred during cycling, particularly in the c axis, and the electrode impedance development are identified as the primary factors affecting the cycling stability at higher states of charge. Relative to LiNiO2, Co and Al-substituted LiNiO2 materials clearly demonstrated the beneficial impact that reducing structural transformations had on the electrochemical performance. By decreasing the degree of structural change during charging to 4.75 V, the Co and Al substituents considerably reduced the impedance development, and improved the cycling stability of the Ni-rich electrodes. iv.

Theoretical Calculations for Metal Migration:

The significant capacity loss shown during cycling to such high states of charge can arise from one of the many known positive electrode degradation mechanisms, including when cations from the transition metal layer migrate to the Li layer. In this mechanism, isolated cations that migrate to the Li layer leave behind a vacancy in the 3b octahedral sites of the transition metal layer. These vacancies attract pairs of Li ions to the tetrahedral sites immediately above and below the vacant transition metal site.81 The resulting Li-Li dumbbells then block further Li occupancy in six neighboring octahedral sites that share


Page 40 of 55

Page 41 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


faces with the occupied tetrahedral sites. The blocked octahedral sites can be refilled, but at lower voltages, as energy is required to displace the favorable tetrahedral Li to less favorable but more abundant octahedral sites.81 This can result in an overall capacity reduction if the reinsertion voltage falls below the cutoff voltage or if other competing reactions such as conversion occur first. Cation migration to the Li layer becomes more likely at high states of charge when most Li sites are vacant. The first step of the process is a cation hop into the tetrahedral site directly below or directly above the original octahedral site of the transition metal layer. Once in the tetrahedral site, the migrating cation then has the option of hopping back to its original site or of hopping to one of its three nearestneighbor octahedral sites in the Li layer. If the cation picks one of the three sites in the Li layer, it is unlikely to return to its original site in the transition metal layer upon reinsertion of Li, resulting in permanent structural changes. A host is thereby created consisting of vacant sites within the transition metal layer that attract capacity-reducing Li dumbbells. The experimental results previously shown indicate that capacity reduction is highly correlated with transition metal chemistry. The cation migration to the tetrahedral site as a function of the surrounding transition metal chemistry in the fully charged state was also investigated with DFT calculations. To consider the migration of Ni and Co from an octahedral site in the transition metal layer to an adjacent tetrahedral site of the Li layer in a fully delithiated NiO2 or CoO2 host having the O3 crystal structure, four scenarios were investigated: (i) Co migration in CoO2, (ii) Ni migration in NiO2, (iii) the migration of an isolated Co in NiO2 and (iv) the migration of an isolated Ni in CoO2. Calculations were performed in supercells consisting of 16 transition metal cations and 32 oxygen anions. The energy cost of hopping from a transition metal site to a tetrahedral site in an empty Li layer is shown in Figure 19. All four energies are positive, indicating that there is a barrier for Co or Ni migration to the Li layer. The tetrahedral sites are locally stable such that an activation barrier separates the octahedral site from the tetrahedral site. The calculated energies in Figure 19, therefore, correspond to lower bounds on the overall barrier to migrate to the Li layer. The barriers range between 1.29 eV and 2.75 eV. Surprisingly, the barriers are relatively insensitive to the chemistry of the migrating cation (i.e. Co or Ni), but very sensitive to the chemistry of the surrounding host. In the Co-rich hosts, the barrier is approximately 1.3 eV independent of whether Co or Ni hops to the tetrahedral site. In the Ni-rich hosts, in contrast, the barrier is more than 2.6 eV, also independent of whether it is Co or Ni that migrates to the tetrahedral site. This indicates that the chemistry of the host plays a crucial role in determining the susceptibility of cations to migrate to the Li layer. The origin of this remarkable property is rooted in unique differences between the electronic structures of CoO2 and NiO2 and will be described elsewhere.82



Figure 19: Tetrahedral defect formation energies in CoO2 (top) and NiO2 (bottom) hosts, where positive values indicate that the octahedral configuration is more stable. Dark purple and grey bars indicate the energy of Co and Ni ions migrating into adjacent tetrahedral sites in the hosts, respectively. The lighter blue bar indicates the energy of Li-Al dumbbell formation in each host. Exact energies are listed next to each bar.

The barriers for Co and Ni migration to tetrahedral sites in the Li layers of Co-rich and Ni-rich compounds are all positive and quite high. Next, the migration of Al ions to adjacent tetrahedral sites in O3 LiCo15AlO32 and LiNi15AlO32 was considered. Al maintains its formal valence of 3+ during charge and discharge of the compound. One Li per Al ion will therefore remain within the compound in the fully charged state. The most stable configuration when Al migrates to a tetrahedral site is a Li-Al dumbbell as illustrated in Figure S15. The energy changes accompanying the migration of an Al ion from the transition metal layer to form a Li-Al dumbbell are shown in Figure 19.82 The Li-Al dumbbell formation energies indicate that the host chemistry again plays a crucial role in determining the energy barrier. In the Ni-rich host, the barrier is approximately 1.2 eV, while in the Co-rich host the octahedral and tetrahedral sites are essentially degenerate (in fact the energy change is slightly negative). In the fully charged Co-rich host, there is effectively no barrier opposing the first step of Al migration to the Li layer. The effect of Ni or Co alloying, as would be seen in the LixNi1-y-zCoyAlzO2 materials, was explored by replacing some of the Ni surrounding the Al in LiNi15AlO32 with Co. Each Al has six neighboring transition metals and all possible Co-Ni orderings (for Co next to Al) were considered in the aforementioned unit cell. The change in energy as Al migrates from an octahedral site to a tetrahedral dumbbell configuration is shown in Figure 20 as a function of the number of Co in the nearest neighbor shell (multiple points at the same concentration correspond to different configurations). Figure 20 shows that as the Co concentration surrounding the Al ion in the transition metal layer increases, the energy to form a tetrahedral dumbbell configuration decreases. When all transition metals surrounding the Al are Co, even when some Ni remains in the structure beyond the nearest-neighbor shell, the dumbbell formation energy is very slightly negative as was seen in LiCo15AlO32.


Page 42 of 55

Page 43 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 20: Dumbbell formation under varying concentrations of Co to Ni in LiM15AlO32 (M = Co, Ni). The Li-Al dumbbell formation energy decreases as more Co surrounds the Al.

Lastly, a mechanism by which dumbbell formation in Co-rich systems could result in irreversible changes in the material is explored. Configurations using larger unit cells to reduce interactions between periodic images (Li19AlCo23O48) were sampled to determine whether Li-Al dumbbells (Figure S16 (a)) would remain stable upon Li reintercalation. The calculations indicate that the lowest-energy configurations at high Li concentrations are those in which Al returns to the transition metal layer. However, it is possible that in the charged state, Al may migrate into an adjacent octahedral site in the Li layer, with a Li atom taking its place in the dumbbell (Figure S16 (b)). The calculated energy of the Li-Li dumbbell configuration is only 105 meV/Al above the Li-Al dumbbell in LiCo15AlO32 (calculations with the same number of Li atoms were used). If such Li-Li dumbbells form, Al may not be able to migrate back into the transition metal layer even if that would lower the energy of the system. Furthermore, of the calculated configurations that contain Al in the Li layer, the Li-Li dumbbell is lower in energy than configurations with Li in the transition metal layer, suggesting that reinsertion of Li into the structure may be indirectly thwarted by Al in the transition metal layer. These calculations indicate that dumbbell formation and the resulting irreversible changes are unlikely to occur in LiNi15AlO32 but become more prominent with increasing concentration of Co. In the experimental results it is shown that LiCo1-yAlyO2 materials experience extreme irreversible loss after charging to high levels of delithiation, in contrast to LiNi1-yAlyO2 materials (Figure 6 and Figure 12, respectively). Furthermore, when only a small amount of Co is present in LixNi1-y-zCoyAlzO2, capacity loss does not significantly increase (Figure 18) because not enough Co surrounds Al to drive dumbbell formation (Figure 20). In order to avoid possible irreversible dumbbell formation, Co must remain a minority component in LixNi1-y-zCoyAlzO2 materials. v.

Experimental Evidence for Tetrahedral Migration of Al:



To confirm the hypothesized and theoretically expected migration of Al ions to neighboring sites in a LiCoO2–based host, solid-state magic-angle spinning (MAS) NMR spectroscopy was employed to identify changes in the local chemical environment of Al. Solid-state 27Al NMR results of pristine and cycled LiCo0.8Al0.2O2, shown in Figure 21 (a) and Figure S17, support the irreversible migration of Al ions to tetrahedral environments after being charged to high potentials. The spectrum of the pristine sample contains a series of well-resolved resonances between 20 and ~60 ppm that arise from octahedral Al ions with varying numbers of Co3+ neighbors in the first cation coordination shell, where an increased number of Co3+ ions leads to higher chemical shifts; the 62 ppm resonance corresponds to the local environment Al(OCo)6-z(OAl)z where z = 0.83 Samples were then charged to different states of charge (4.2, 4.5 and 4.75 V) and fully discharged to minimize the effects of Co4+. NMR spectra were acquired with and without a T2 (spin-spin) relaxation filter, where the T2 filter suppresses an extremely broad signal centered at around 60 ppm (Figure S18). This signal is tentatively ascribed to Al3+ near paramagnetic Co4+ ions (in distant coordinate shells) or possibly in metallic regions of the sample.84 The series of sharp resonances broaden on increasing the state of charge, and the intensity of the 62 ppm Al(OCo)6 (z = 0) resonance decreases relative to resonances where z > 1 (i.e., less Co3+). At the same time, the development of a high frequency shoulder (centered around 70 ppm) was observed at frequencies higher than the octahedral Al signals after charging LiCo0.8Al0.2O2 to only 4.2 V and subsequently discharging. Since tetrahedral Al sites typically resonate at approximately 60 – 80 ppm in diamagnetic oxides,85 the observed resonance is tentatively assigned to tetrahedral Al3+ in diamagnetic environments, which would result from migrating Al ions. Accessing higher states of charge increased the apparent intensity of the tetrahedral Al signal, especially after charging to 4.75 V (see Figure S17 for deconvolution). Taken together, the results suggest that Al initially nearby more Co3+ are more likely to migrate to tetrahedral sites and/or to be located nearby residual Co4+/metallic regions after cycling to higher voltage. Thus as Al is surrounded by more Co ions, an increased propensity for Al ions to migrate to the tetrahedral site is expected, in agreement with the DFT calculations. Regardless of this possible preference for migration based on the initial local Al environment, the spectra nonetheless provides evidence of a clear migration of Al to tetrahedral environments resonating at 70 ppm. For the LiCo0.8Al0.2O2 sample charged to the lower potential of 4.2 V, the relative proportion of Al ions in tetrahedral environments (with respect to the entire observed Al signal) was estimated at 12% ± 4%. The growth of the signal intensity indicated the concentration of tetrahedral Al in LiCo0.8Al0.2O2 increased substantially as the material accessed higher states of charge. In contrast to LiCo0.8Al0.2O2, no increase in the higher-frequency tetrahedral Al signal was measured for LiNi0.8Al0.2O2 (O) after cycling (Figure 21 (b)). The small tetrahedral Al resonance (75 ppm) in the spectra of the pristine and cycled LiNi0.8Al0.2O2 (O) material arises from a known γ-LiAlO2 impurity phase83; its intensity relative to the octahedral Al features did not increase, even after charging to 4.75 V. Additional analysis and discussion of the 27Al NMR results is provided in the supplemental information.


Page 44 of 55

Page 45 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 21: Solid-state 27Al NMR spectra of (a) LiCo0.8Al0.2O2 and (b) LiNi0.8Al0.2O2 (O) in the pristine state and after charge at 10 mA/g to different cutoff potentials and subsequent discharge at 10 mA/g to 2.7 V (vs. Li metal). The expected region for the tetrahedral Al signal is shown in both plots. All samples were prepared without exposure to ambient atmosphere. Spectra were acquired at a MAS rate of 55 kHz and a field strength of 11.7 T.

The electrochemical experiments clearly demonstrate that Al substitution for Co induces a larger degradation of the electrochemical performance and increased first cycle irreversible loss than Al substitution for Ni. The high irreversible loss after charging to high states of delithiation and subsequent poor cycling capabilities of the Al-substituted LiCoO2 materials were associated to the propensity for Al ions in LiCo1-yAlyO2 to migrate to the tetrahedral or 3a octahedral sites. 27Al NMR results experimentally confirmed the theoretical predictions obtained by the previously discussed DFT calculations that migration of Al ions occurred in a LiCoO2-based host but not in a LiNiO2-based material. This structural degradation mechanism would become progressively more influential when the electrode was repeatedly cycled to very high potentials, and explains the high irreversible loss, and increased impedance development of Al-substituted LiCoO2 relative to the pure LiCoO2 material. vi.

Summary:

Table II provides a succinct overview of the physical and electrochemical characterization results for all layered oxide materials investigated herein at 4.5 and 4.75 V. Key relationships were identified between the structural changes and electrochemical performance upon cycling to different states of charge. Alsubstitution into LiNiO2 improved capacity retention by more than 10% after 15 cycles at both 4.5 and 4.75 V compared to LiNiO2 (Table II). LiNiO2 proceeded via a two-phase reaction process at high states of charge with a large change in the c parameter that likely resulted in structural degradation deleterious to cycling stability. In contrast, Al substitution into LiNiO2 stabilized the layered structure at high potentials and drastically reduced the structural transformations between 4.5 V and 4.75 V, even for 5% Al substitution. Moreover, Al substitution also diminished transition metal dissolution while maintaining Li+ diffusivity. As a result, the cycling stability of the LiNi1-yAlyO2 materials improved slightly as the Al content increased beyond 5%. In comparison to LiNiO2, LiCoO2 had substantially inferior stability when cycled to 4.75 V, and after 15 cycles it only retained 69.8% of its capacity, significantly lower than the 82.6% retained by LiNiO2. This poor cycling performance could be mainly attributed to the two-phase reaction leading to the O1 phase



and its massive c parameter change. This two-phase transition is more drastic and therefore more deleterious than the one to C2/m observed in LiNiO2. The substantial Co dissolution and increased parasitic reactions present on LiCoO2 at high states of charge also contributed to the diminished stability. Despite preventing the two-phase reaction at the end of charge, reducing the Co dissolution, and preserving the reaction kinetics, Al substitution into LiCoO2, in contrast to LiNiO2, severely degraded the materials’ electrochemical stability (Table II). For the LiCo1-yAlyO2 materials, accessing very high states of charge promoted the migration of Al ions into tetrahedral sites, which caused a large increase in the first cycle irreversible loss and degradation of the material’s electrochemical capabilities. In agreement with the theoretical calculations by DFT and experimental results from solid-state NMR, electrochemical analysis suggested that as more Al was substituted into the structure, the more pronounced this failure mechanism became. The cycling stability at high states of charge for the Ni-Co-Al ternary mixtures was also dependent on the phase transitions between 4.5 and 4.75 V. Comparison of LiNi0.95Al0.05O2, LiNi0.8Co0.2O2, and LiNi0.8Co0.15Al0.05O2 is insightful because all three materials have similar theoretical capacities. As with Al, Co substitution into LiNiO2 was effective at eliminating the high state of charge two-phase reactions and in reducing the c lattice parameter variation, resulting in improved capacity retention with cycling. In contrast to Al substitution, Co increased the amount of transition metal dissolution, which is due to the catalytic nature of Co. This is an important result as it seems to show that dissolution is not the primary cause of degradation when cycling at high states of delithiation. While Al substitution into LiNi0.8Co0.2O2 did not reduce the magnitude of Ni/Co dissolution, it definitively minimized the lattice changes and improved the cycling stability of the material, further bolstering this conclusion. Confirmation of the importance that the structural changes have on the cycling stability is apparent in the comparison of cNCA and LiNi0.8Co0.15Al0.05O2. cNCA showed a feature in the dQ/dV and PITT profiles at ~4.6 V that was substantially more prominent than the similar feature in LiNi0.8Co0.15Al0.05O2 or LiNi0.8Co0.2O2. Since the data presented in Table II rules out all other possibilities, the poorer cycling stability of cNCA relative to LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2 when charged to 4.75 V can be attributed to this high state of charge structural transition and larger percentage change to the unit cell parameters. The results presented herein help answer critical questions regarding why Ni-Co-Al ternary mixtures such as LiNi0.8Co0.15Al0.05O2 are among the best positive electrode materials, and establish the primary pathways towards the development of improved materials cycling in this potential range. As previously mentioned, the substitution of Co into LiNiO2 had mixed results. It may be that at higher states of charge, when Co becomes active, one would want to reduce the amount of Co to be consistent with the desired delithiation such that it does not become redox active. However, this work was not an exhaustive evaluation of the impact that Co may have on the performance of the electrode material. Other aspects, such as the electronic conductivity or the impact on the transition metal occupation of the 3a site, were not fully investigated and may prove to be critically important properties. Nevertheless, of all the materials studied, LiNi0.8Co0.15Al0.05O2 and LiNi0.8Co0.2O2 had the best cycling stability near full delithiation. Table II: Important structural and electrochemical results for all materials at 4.5 V and 4.75 V (vs. Li metal). Lithiation (x)/TM oxidation (%) and OCV after charge under PITT protocol up to 4.5 and 4.75 V (Figure S6 and Table SII) as well as the percent change in Li-content/TM oxidation between 4.5 and 4.75 V. First cycle irreversible loss (%) and capacity retention (%) after 15 cycles under constant current cycling up to 4.5 and 4.75 V (Figures 6, 12, 18 and Tables SIII-SV), as well as the difference (%) between those two potential cut-offs. The c parameter and unit cell volume changes (%) at 4.5 V and 4.75 V were calculated


Page 46 of 55

Page 47 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


from the in-situ XRD analysis recorded under constant current charging (Figures 4, 9 and 15). The * denotes a two-phase transition between 4.5 and 4.75 V, and ** signifies that two phases were present simultaneously. The difference in Li+ diffusivity derived from PITT experiments (Figures 2, 7, and 13) between 4.5 and 4.75 V displayed as the absolute and relative (%) value. Transition metal dissolution (%Ni/%Co) was obtained by ICP-OES (Table I).

While it is well documented that carbonate electrolytes are unstable at potentials above 4.5 V and that transition metal dissolution from the positive electrode can degrade cell functionality, neither of these failure mechanisms were found to be the primary reason for poor electrochemical performance of layered oxide materials operating near full delithiation.86,87 The PITT measurements showed that differences in electrochemical performance between positive electrode materials at such high states of charge could not be attributed to differences in the Li+ diffusivity or to kinetic limitations. Table II also shows that the percentage change in the c parameter or in the unit cell volume between 4.5 and 4.75 V was largely independent of the materials’ cycling stability. Furthermore, the change in lithiation and amount of transition metal oxidation between 4.5 and 4.75 V was also unrelated to the electrochemical performance. Instead, the cycling stability of these layered oxide materials at or near full delithiation was mainly dependent on the intrinsic structural stability of the positive electrode material. Specifically, the elimination of a two-phase reaction at the end of charge and the prevention of metal migration into undesirable sites were identified as the primary factors dictating the cycling stability. Thus, for optimal performance at such high states of charge it is critically important that the positive electrode material is devoid of two-phase transitions with significant lattice mismatch, robust enough to withstand the large single-phase structural evolution and the correlated stresses that occur during cycling, as well as preventing internal migration of metal ions to unfavorable sites. After this foundation is established, longterm impedance evolution from failure modes originating at the surface needs to be addressed.

Conclusions: The cycling capacity and stability of layered oxide positive electrodes near full delithiation was determined to be primarily attributed to the intrinsic metal site stability and strains induced within the electrode material. Using an extensive array of electrochemical techniques along with in-situ XRD and operando microcalorimetry, the impact of Al substitution on the structural and electrochemical properties at very high states of charge for LiCoO2 and LiNiO2 was shown. While Al substitution into both LiCoO2-



and LiNiO2-based positive electrodes was observed to stabilize the structural evolution during delithiation, reduce parasitic reactions, and restrict transition metal dissolution, only in the case of adding Al to LiNiO2 did the electrochemical performance of the electrode improve at high degrees of delithiation. Al substitution into LiCoO2 led to a systematic increase in the first cycle irreversible loss and was ineffective in stabilizing the capacity degradation when cycling to high potentials. DFT calculations predicted Al ion migration to tetrahedral sites in LiCoO2-based hosts, but not in LiNiO2-based materials when such structures were near full delithiation. In support of these theoretical predictions at high states of delithiation, solid-state NMR spectra showed evidence of Al migration into neighboring tetrahedral sites for Al-substituted LiCoO2, but not Al-substituted LiNiO2. Both 20% Al or Co substitution into LiNiO2 was found to significantly reduce the strain inducing structural transitions and increase the stability during cycling. LiNi0.8Co0.15Al0.05O2 was shown to have superior electrochemical performance than LiNi0.8Al0.2O2 and LiNi0.8Co0.2O2 at very high states of charge, which can be attributed to the increased structural stability towards phase transitions gained by the mixture of both Al and Co ions. Without exception, addition of Co always increased the total transition metal dissolution (Ni + Co) significantly. Capacity retention with cycling at extremely high degrees of delithiation was primarily but not exclusively tied to mechanical strains induced by phase transitions as opposed to dissolution reactions, parasitic reactions with electrolytes, changes in Li+ diffusivity, or the average delithiation voltage. The results presented herein enhance the collective understanding of how the transition metal chemistry affects the stability and electrochemical performance of layered oxide electrodes at or near full delithiation. Utilization of the lessons derived from this work will aid in the identification of new positive electrode materials that will minimize the gap between the theoretical and practically obtained electrode energy densities. Acknowledgements: The authors would like to thank Gerbrand Ceder, Philip Reeves, and Louis Piper for helpful discussions and experimental assistance during the course of this work. This work was supported in full as part of NECCES, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Award No. DE-SC0012583.

References: (1)

Blomgren, G. E. The Development and Future of Lithium Ion Batteries. J. Electrochem. Soc. 2017, 164, A5019–A5025.

(2)

Amatucci, G. G.; Tarascon, J. M.; Klein, L. C. CoO2, The End Member of the LiCoO2 Solid Solution. Electrochem. Soc. 1996, 143, 1114–1123.

(3)

Ohzuku, T.; Ueda, A.; Nagayama, M. Electrochemistry and Structural Chemistry of LiNiO2 (R3m) for 4 Volt Secondary Lithium Cells. J. Electrochem. Soc. 1993, 140, 1862-1870.

(4)

Ueda, A.; Ohzuku, T. Solid-State Redox Reactions of LiCoO2 (R3m) for 4 Volt Secondary Lithium Cells. J. Electrochem. Soc. 1994, 141, 2972-2977.

(5)

Chen, Z.; Lu, Z.; Dahn, J. R. Staging Phase Transitions in LixCoO2. J. Electrochem. Soc. 2002, 149, A1604-A1609.


Page 48 of 55

Page 49 of 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(6)

Pillot, C. The Rechargeable Battery Market and Main Trends 2016-2025. Adv. Automot. Batter. Conf. 2016.

(7)

Croguennec, L.; Pouillerie, C.; Delmas, C. NiO2 Obtained by Electrochemical Lithium Deintercalation from Lithium Nickelate: Structural Modifications. J. Electrochem. Soc. 2000, 147, 1314-1321.

(8)

Xia, H.; Lu, L.; Meng, Y. S.; Ceder, G. Phase Transitions and High-Voltage Electrochemical Behavior of LiCoO2 Thin Films Grown by Pulsed Laser Deposition. J. Electrochem. Soc. 2007, 154, A337-A342.

(9)

MacNeil, D. D.; Dahn, J. R. The Reaction of Charged Cathodes with Nonaqueous Solvents and Electrolytes: I. Li0.5CoO2. J. Electrochem. Soc. 2001, 148, A1205-A1210.

(10)

Aurbach, D.; Markovsky, B.; Salitra, G.; Markevich, E.; Talyossef, Y.; Koltypin, M.; Nazar, L.; Ellis, B.; Kovacheva, D. Review on Electrode-Electrolyte Solution Interactions, Related to Cathode Materials for Li-Ion Batteries. J. Power Sources 2007, 165, 491–499.

(11)

Faenza, N. V.; Lebens-Higgins, Z. W.; Mukherjee, P.; Sallis, S.; Pereira, N.; Badway, F.; Halajko, A.; Ceder, G.; Cosandey, F.; Piper, L. F. J.; et al. Electrolyte-Induced Surface Transformation and Transition-Metal Dissolution of Fully Delithiated LiNi0.8Co0.15Al0.05O2. Langmuir 2017, 33, 93339353.

(12)

Wang, H.; Rus, E.; Sakuraba, T.; Kikuchi, J.; Yasuyuki, K.; Abruna, H. D. CO2 and O2 Evolution at High Voltage Cathode Materials of Li-Ion Batteries: A Differential Electrochemical Mass Spectrometry Study. Am. Chem. Soc. 2014, 86, 6197–6201.

(13)

Armstrong, a. R.; Holzapfel, M.; Novák, P.; Johnson, C. S.; Kang, S. H.; Thackeray, M. M.; Bruce, P. G. Demonstrating Oxygen Loss and Associated Structural Reorganization in the Lithium Battery Cathode Li[Ni0.2Li0.2Mn0.6]O2. J. Am. Chem. Soc. 2006, 128, 8694–8698.

(14)

Bak, S. M.; Nam, K. W.; Chang, W.; Yu, X.; Hu, E.; Hwang, S.; Stach, E. A.; Kim, K. B.; Chung, K. Y.; Yang, X. Q. Correlating Structural Changes and Gas Evolution during the Thermal Decomposition of Charged LixNi0.8Co0.15Al0.05O2 Cathode Materials. Chem. Mater. 2013, 25, 337– 351.

(15)

Amatucci, G. G.; Tarascon, J. M.; Klein, L. C. Cobalt Dissolution in LiCoO2-Based Non-Aqueous Rechargeable Batteries. Solid State Ionics 1996, 83, 167–173.

(16)

A. Blyr; Sigala, C.; Amatucci, G.; Guyomard, D.; Chabre, Y.; Tarascon, J. M. Self-Discharge of LiMn2O4/C Li-Ion Cells in Their Discharged State. Electrochem. Soc. 1998, 145, 194–209.

(17)

Arora, P.; White, R. E. Capacity Fade Mechanisms and Side Reactions in Lithium-Ion Batteries. J. Electrochem. Soc. 1998, 145, 3647–3667.

(18)

Aoshima, T.; Okahara, K.; Kiyohara, C.; Shizuka, K. Mechanisms of Manganese Spinels Dissolution and Capacity Fade at High Temperature. J. Power Sources 2001, 97–98, 377–380.

(19)

Guilmard, M.; Croguennec, L.; Delmas, C. Thermal Stability of Lithium Nickel Oxide Derivatives. Part II: LixNi0.70Co0.15Al0.15O2 and LixNi0.90Mn0.10O2 (x = 0.50 and 0.30). Comparison with LixNi1.02O2 and LixNi0.89Al0.16O2. Chem. Mater. 2003, 15, 4484–4493.

(20)

Lin, F.; Markus, I. M.; Nordlund, D.; Weng, T.-C.; Asta, M. D.; Xin, H. L.; Doeff, M. M. Surface Reconstruction and Chemical Evolution of Stoichiometric Layered Cathode Materials for Lithium-



Ion Batteries. Nat. Commun. 2014, 5, 1-9. (21)

Jang, Y.; Huang, B.; Wang, H.; Sadoway, D. R.; Ceder, G. LiAlyCo1-yO2 (R3m) Intercalation Cathode for Rechargeable Lithium Batteries. J. Electrochem. Soc. 1999, 146, 862–868.

(22)

Ohzuku, T.; Yanagawa, T.; Kouguchi, M.; Ueda, A. Innovative Insertion Material of LiAl1/4Ni3/4O2 (R3m) for Lithium-Ion (Shuttlecock) Batteries. J. Power Sources 1997, 68, 131– 134.

(23)

Faenza, N. V; Bruce, L.; Lebens-higgins, Z. W.; Plitz, I.; Piper, L. F. J.; Amatucci, G. G. Growth of Ambient Induced Surface Impurity Species on Layered Positive Electrode Materials and Impact on Electrochemical Performance. J. Electrochem. Soc. 2017, 164, A3727–A3741.

(24)

Van der Ven, A. First-Principles Evidence for Stage Ordering in LixCoO2. J. Electrochem. Soc. 1998, 145, 2149-2155.

(25)

Arroyo y de Dompablo, M. E.; Van der Ven, A.; Ceder, G. First-Principles Calculations of Lithium Ordering and Phase Stability on LiNiO2. Phys. Rev. B 2002, 66, 1-9.

(26)

Reimers, J. N.; Dahn, J. R. Short-Range Cation Ordering in LixNi2-XO2. Phys. Rev. B 1993, 47, 8486–8493.

(27)

Van Der Ven, A.; Aydinol, M.; Ceder, G.; Kresse, G.; Hafner, J. First-Principles Investigation of Phase Stability in LixCoO2. Phys. Rev. B 1998, 58, 2975–2987.

(28)

Jang, Y.-I.; Neudecker, B. J.; Dudney, N. J. Lithium Diffusion in LixCoO2 (0.45

Ni - ACS Publications

Recommend Documents