Structural Transformations in High-Capacity Li2Cu0.5Ni0.5O2

Mar 9, 2017 - Cathode materials that can cycle >1 Li+ per transition metal are of substantial interest for increasing the overall energy density of li...
8 downloads 0 Views 6MB Size
Article pubs.acs.org/cm

Structural Transformations in High-Capacity Li2Cu0.5Ni0.5O2 Cathodes Rose E. Ruther,*,† Amaresh Samuthira Pandian,† Pengfei Yan,‡ Johanna Nelson Weker,§ Chongmin Wang,‡ and Jagjit Nanda*,† †

Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, Washington 99354, United States § SLAC National Accelerator Laboratory, Stanford Synchrotron Radiation Lightsource, Menlo Park, California 94025, United States ‡

S Supporting Information *

ABSTRACT: Cathode materials that can cycle >1 Li+ per transition metal are of substantial interest for increasing the overall energy density of lithium-ion batteries. Li2Cu0.5Ni0.5O2 has a very high theoretical capacity of ∼500 mAh/g assuming both Li+ ions are cycled reversibly. The Cu2+/3+ and Ni2+/3+/4+ redox couples are also at high voltage, which could further boost the energy density of this system. Despite such promise, Li2Cu0.5Ni0.5O2 undergoes irreversible phase changes during charge (delithiation) that result in large first-cycle irreversible loss and poor long-term cycling stability. Oxygen evolves before the Cu2+/3+ or Ni3+/4+ transitions are accessed. In this contribution, X-ray diffraction, transmission electron microscopy (TEM), and transmission X-ray microscopy combined with X-ray absorption near edge structure (TXM− XANES) are used to follow the chemical and structural changes that occur in Li2Cu0.5Ni0.5O2 during electrochemical cycling. Li2Cu0.5Ni0.5O2 is a solid solution of orthorhombic Li2CuO2 and Li2NiO2, but the structural changes more closely mimic the changes that the Li2NiO2 endmember undergoes. Li2Cu0.5Ni0.5O2 loses long-range order during charge, but TEM analysis provides clear evidence of particle exfoliation and the transformation from orthorhombic to a partially layered structure. Linear combination fitting and principal component analysis of TXM−XANES are used to map the different phases that emerge during cycling ex situ and in situ. Significant changes in the XANES at the Cu and Ni K-edges correlate with the onset of oxygen evolution.



capacity of 250 mAh/g per Li+.15 Nearly 400 mAh/g of capacity is obtained during the first charge, but only ∼125 mAh/g is reversible. The large first-cycle irreversible loss was attributed to oxygen evolution and structural transformations. Gas evolution measurements, in situ Raman spectroscopy, and in situ X-ray absorption spectroscopy were used to track the chemical and structural changes. In this contribution, we present a more in-depth analysis of the structural transformations that occur in Li2Cu0.5Ni0.5O2 cathodes during electrochemical cycling. Ex situ X-ray diffraction (XRD) and transmission electron microscopy (TEM) are used to characterize new phases that form during the initial charge. Full field transmission X-ray microscopy combined with X-ray absorption near edge structure (TXM− XANES) is used to follow chemical changes spatially at the Cu and Ni K-edges. TXM−XANES combines a high spatial resolution (∼30 nm) and a large field of view (∼30 μm), allowing the spatial homogeneity of reactions to be mapped across one or more cathode particles.16,17 The results of XRD, TEM, and TXM−XANES are discussed in the context of our earlier work.15 A complete picture of the chemical and structural changes that occur in Li2Cu0.5Ni0.5O2 during lithiation and delithiation is important for improving the electrochemical reversibility. The ultimate goal is to develop

INTRODUCTION One approach to increase the specific energy of lithium-ion batteries is to develop materials that support multielectron redox reactions.1 For example, sulfur cathodes,2 conversion electrodes,3 and many organic materials4 react with more than one lithium per formula unit. Among the transition metal oxides, a few chemistries with more than one lithium per transition metal have been investigated for high-capacity lithium-ion cathodes. Examples include lithium-rich NMCs [xLi2MnO3·(1−x)LiMO2, where M = Mn, Ni, or Co],5 antifluorite-type oxides,6 and several polyanionic compounds.7 While many of these cathodes have very high theoretical capacities, none have proven to be practical for commercial cells. Lithium-rich NMCs lose both lithium and oxygen during the initial charge,8 and they undergo slow phase transformations during subsequent cycling.9 These structural changes lead to capacity loss, voltage fade, and impedance rise.10 The anti-fluorite-type oxides such as Li5FeO4 also lose oxygen (as Li2O) during charge.11 Generally, only one or fewer Li+ ions per formula unit can be cycled reversibly.6 Successful cycling of more than one Li+ per formula unit has been demonstrated with polyanionic compounds, although typically only one redox couple of the transition metal is within a useful voltage range.12 Li2MnSiO4 is a notable exception, but this material suffers from poor transport and structural instability.13,14 Recently, we characterized the electrochemical performance of Li2Cu0.5Ni0.5O2 cathodes, which have a very high theoretical © 2017 American Chemical Society

Received: December 23, 2016 Revised: March 9, 2017 Published: March 9, 2017 2997

DOI: 10.1021/acs.chemmater.6b05442 Chem. Mater. 2017, 29, 2997−3005

Article

Chemistry of Materials

points across the Ni and Cu K-edges from 8280 to 9115 eV. The energy spacing was not uniform; rather, finer spacing was used near the absorption edge to detect small shifts. Two-dimensional (2D) FF TXM−XANES data were also collected in situ as cathodes were cycled. For in situ measurements, pouch cells were constructed using heatsealable laminated pouches (Shield Pack, Bemis), which have a 9 μm thick layer of aluminum. The aluminum is an effective barrier to oxygen and water but allows for reasonable transmission of the X-ray beam. Pouch cells were assembled on site at SLAC in an argon-filled glovebox using lithium counter electrodes, a 1 M LiPF6 EC/DMC [1:1 (v/v)] electrolyte, and a Celgard 2325 separator. Cells were cycled at 20 mA/g of Li2Cu0.5Ni0.5O2. To minimize beam damage, in situ XANES maps were collected at a lower energy resolution with images taken at 21 points across the Ni K-edge and 20 points across the Cu Kedge. To further minimize beam damage and to increase the size of the sampling area, data collection alternated across six different regions of the electrode (three each for Cu and Ni). TXM-Wizard was used for all data analysis, including principal component analysis and linear combination fitting.

practical cathodes that take advantage of multielectron processes.



EXPERIMENTAL SECTION

Preparation of Li2Cu0.5Ni0.5O2 Electrodes. Li2Cu0.5Ni0.5O2 was synthesized according to our previously published procedure.15 Electrodes were prepared by tapecasting conventional slurries onto carbon-coated aluminum foil (12 μm, z-flo 2651, Coveris). Each slurry contained Li2Cu0.5Ni0.5O2 powder, carbon black (Super P Li, TIMCAL), graphite (C-NERGY KS 6L, TIMCAL), and polyvinylidene difluoride (Solvay) in N-methyl-2-pyrrolidone (Sigma-Aldrich). Electrodes with higher loadings (3−7 mg of Li2Cu0.5Ni0.5O2/cm2) were prepared for TEM and XRD, and electrodes with lower loadings (0.5 mg of Li2Cu0.5Ni0.5O2/cm2) were prepared for TXM−XANES to allow microscopy of individual particles. Electrodes for TEM and XRD consisted of 84 wt % Li2Cu0.5Ni0.5O2, 5 wt % carbon black, 3 wt % graphite, and 8 wt % PVDF. TXM−XANES electrodes consisted of 60 wt % Li2Cu0.5Ni0.5O2, 20 wt % carbon black, 10 wt % graphite, and 10 wt % PVDF. Electrodes were dried at 80 °C under vacuum and stored in an argon-filled glovebox. Cell Cycling for ex Situ Measurements. Samples for ex situ analysis were cycled in coin cells (316L stainless steel, size CR2032, Hohsen Corp.) with lithium counter electrodes (99.9%, Alfa Aesar) and polymer separators (Celgard 2325). The electrolyte was 1.2 M LiPF6 in a 3:7 (v/v) ethylene carbonate (EC)/dimethyl carbonate (DMC) mixture (Selectilyte, BASF). Cells were cycled at a constant current of 50 mA/g of Li2Cu0.5Ni0.5O2. After being cycled, the cells were disassembled inside an argon-filled glovebox. The cathodes were rinsed with dimethyl carbonate and dried under vacuum. Samples for TEM and TXM−XANES were sealed under argon in aluminized pouches (Shield Pack, Bemis) for shipment. Samples for TXM− XANES were sealed under argon in laminated polyester/polyethylene pouches (Kapak SealPAK 4.5 mil) for data collection. Samples for XRD were sealed under a polyimide film for data collection. XRD. Crystallographic phase characterization was conducted on a PANalytical X’pert Pro Powder Diffractometer with Cu Kα radiation (λ ≈ 1.5418 Å) in the 2θ range of 10−70° at a scan rate of 0.25°/min. The operating voltage and current were 45 kV and 40 mA, respectively. TEM. Electrode powders were scraped off of the aluminum current collectors and directly dusted on carbon grids for TEM observations. An FEI Titan 80-300 transmission electron microscope with a probe corrector was used for STEM−HAADF imaging. An FEI Titan 80-300 transmission electron microscope with an image corrector was used for HRTEM and SAED characterizations. Both microscopes operate at 300 kV. EDS measurements were performed on a JEOL ARM200 microscope equipped with a high-efficiency SDD detector. Overall, we consider any effects from electron beam damage to be negligible for the results reported here.18 The pristine sample was not beam sensitive; thus, a series of lattice images from the same region were acquired using STEM−HAADF with no appreciable change. Beam damage was observed only in cycled samples when the beam probe was parked on the thin region of the sample at high magnifications (>1000000×). Therefore, only low-magnification STEM−HAADF images were taken for cycled samples. Lattice images of cycled samples were taken by conventional TEM, which reduced the dose rate significantly (by several orders of magnitude); thus, no appreciable beam effect was found while the HRTEM images were being recorded. TXM−XANES. Hard X-ray full field (FF) transmission X-ray microscopy (TXM) was performed at beamline 6-2c of the Stanford Synchrotron Radiation Lightsource (SSRL) at the SLAC National Accelerator Laboratory. Measurements were taken over both the Ni and Cu K-edges at approximately 8330 and 8980 keV, respectively. The spatial resolution of the microscope was ∼30 nm. A detailed description of the hardware configuration can be found elsewhere.19 Image preprocessing, including magnification correction and image registration, was performed using TXM-Wizard, a software package developed at SLAC.20 XANES maps with high energy resolution were collected on cathodes prepared ex situ with images taken at 137 energy



RESULTS AND DISCUSSION The XRD pattern of Li2Cu0.5Ni0.5O2 powder shows that the material is nearly phase pure (Figure 1). All of the most intense

Figure 1. (a) XRD pattern of Li2Cu0.5Ni0.5O2 powder. (b) XRD pattern of Li2Cu0.5Ni0.5O2 powder with the y-axis expanded to show Li2CO3 impurity.

reflections are attributed to the expected Immm orthorhombic phase (Figure 1a).21−23 Lithium carbonate is present as a minor impurity (Figure 1b). The crystallographic phase of the pristine material was also confirmed using TEM (Figure 2). Selective area electron diffraction (SAED) patterns and scanning TEM− high-angle annular dark field (STEM−HAADF) lattice images confirm that the crystal symmetry is Immm orthorhombic. The voltage profiles of Li2Cu0.5Ni0.5O2 from a sample cycled for ex situ XANES measurements (Figure 3) show a large (∼50%) first-cycle irreversible loss. After the first cycle, the cathode continues to lose capacity gradually. The reversibility does not improve if the upper cutoff voltage is limited to 3.6 V (the first voltage plateau) or 3.9 V (the second voltage plateau).15 In our earlier work, we attributed the poor electrochemical reversibility to oxygen evolution accompanied by significant changes in the crystal structure. 15 Drawing from the literature22,23 and our own Raman spectroscopy results,15 we 2998

DOI: 10.1021/acs.chemmater.6b05442 Chem. Mater. 2017, 29, 2997−3005

Article

Chemistry of Materials

Figure 2. TEM characterization of the pristine Li2Cu0.5Ni0.5O2 sample. (a) A [010] zone axis SAED pattern from a pristine particle (inset image). (b) STEM−HAADF lattice image from the [010] zone axis. (c) STEM−HAADF lattice image from the [001] zone axis.

emerge during charge. The peak near a 2θ value of 19° corresponds to the (003) reflection of LiNiO2 with a layered R3̅m structure.24,25 The transformation from orthorhombic to layered has been predicted computationally for Immm Li2NiO2.22 Broad peaks centered near 2θ values of 38, 44, and 63 are a good match for LixNi1−xO or NiO with the cubic, rock-salt-type lattice (x ≤ 0.3).26 The peaks correspond to the (111), (200), and (220) reflections, respectively. In addition to the nickel oxide phases, LiCuO, Li2CO3, and Li2O are also formed, as indicated in Figure 4. Li2Cu0.5Ni0.5O2 is a solid solution of orthorhombic Li2NiO2 and Li2CuO2. Cu2+ (0.73 Å) and Ni2+ (0.69 Å) have similar ionic sizes, and Li2Cu1−xNixO2 forms a solid solution over the entire composition range from x = 0 to x = 1.23 The structural changes in Li2Cu0.5Ni0.5O2 closely resemble those of the parent Li2NiO2 composition during charge.22,23,27 Immm Li2NiO2 also showed weak diffraction peaks for layered LiNiO2 and rock-salt NiO after charging.27 Another XRD study of Immm Li2NiO2 found no diffraction lines after electrochemical cycling.22 The structure collapsed, forming amorphous material, or very small crystallites, which were not detected by XRD. The morphological and structural evolution of Li2Cu0.5Ni0.5O2 was further investigated by TEM (Figures 5

Figure 3. Voltage profiles of Li2Cu0.5Ni0.5O2 from an electrode prepared for ex situ XANES. The cathode was cycled between 2.25 and 4.3 V vs Li/Li+.

proposed the lattice transformed from the Immm orthorhombic phase to a layered structure similar to 1T-Li2NiO2 (P3̅m1 trigonal) or LiNiO2 (R3̅m trigonal). For a more complete analysis of the phases that form, we performed XRD on cathodes harvested from cells charged (and discharged) to different cutoff voltages (Figure 4).

Figure 5. STEM−HAADF images showing the change in particle morphology after charge−discharge cycling. (a) Pristine Li2Cu0.5Ni0.5O2 particles. (b) Li2Cu0.5Ni0.5O2 particle after one charge−discharge cycle. (c) Li2Cu0.5Ni0.5O2 particle after 100 charge−discharge cycles.

Figure 4. XRD patterns of cathodes taken from cells charged (and discharged) to different end points during the first full cycle (charge to 4.3 V and discharge to 2.25 V). The cathodes discharged to 3.0 and 2.25 V were first fully charged to 4.3 V. The top trace is from a cathode that underwent one full charge−discharge cycle and was recharged to 4.3 V.

and 6). The pristine particles are uniformly dense (Figure 5a). After one charge−discharge cycle, the particle morphology converts to a form with exfoliated layers (Figure 5b). More examples can be found in Figure S2. Exfoliation of the pristine particles into layers accommodates the significant change in density that occurs when the particles transform from orthorhombic (3.7 g/cm3) to trigonal (4.8 g/cm3 for LiNiO2) and cubic (6.7 g/cm3 for NiO). The XRD patterns (Figure 4) suggested layered LiNiO2 and rock-salt NiO phases were formed during the first cycle. TEM investigations further

The diffraction lines for the pristine material disappear during the first charge. In fact, the structure appears to transform completely during the first voltage plateau from OCV to 3.4 V. Charging to 3.4 V requires ∼120 mAh/g of capacity or 0.5 Li+ per formula unit. On the basis of earlier in situ Raman spectroscopy results,15 the structure begins to change with removal of 1 Li+ is removed.21,36,37 Results of density functional theory calculations indicate that the instability with respect to oxygen evolution is due to strong overlap between the O 2p and Cu 3d states in the valence band.36 It is possible that Li2Cu0.5Ni0.5O2 follows a two-step reaction pathway similar to 3002

DOI: 10.1021/acs.chemmater.6b05442 Chem. Mater. 2017, 29, 2997−3005

Article

Chemistry of Materials

Figure 11. (a) Average (bulk) XANES at the Ni K-edge from a cell charged in situ. (b) Magnified view of the absorption edge in panel a. (c) Average (bulk) Ni K-edge XANES from an uncycled cell and cell charged ex situ. (d) Phase maps from TXM−XANES data collected in situ during charge. Phase maps were generated by linear combination fitting using the bulk XANES shown in panel c. An absorption image at 8465 eV is also shown.

that of Li2NiO2, where TXM−XANES is more sensitive to chemical changes that occur with oxygen evolution above 3.9 V. Further work, such as in situ XRD and pair distribution function analysis, will be needed to fully interpret the phases that form during the initial charge.36 TXM−XANES maps collected in situ at the Ni K-edge during charge closely follow the results from the Cu K-edge (Figure 11). Panels a and b of Figure 11 show the average (bulk) XANES at the Ni K-edge for a cell charged in situ. No systematic shift in the edge position is observed. This disagrees with our earlier result, where the Ni K-edge shifted by approximately 2 eV toward a higher binding energy during charge.15 For our earlier work, we collected only the bulk XANES spectra at a higher energy resolution. To minimize Xray beam damage, the energy resolution of TXM−XANES is more limited and small shifts may not be apparent. Nonetheless, significant changes in the shape of the absorption edge are evident after charging above 4.0 V. Figure 11c compares XANES from a pristine, uncycled electrode with XANES from the cell fully charged ex situ. The in situ and ex situ XANES spectra show the same change in shape after charge. Figure 11d shows phase maps generated by linear combination fitting using the XANES spectra in Figure 11c. A mixture of phases is present at 4.0 V, but at 4.3 V, the particle is almost fully transformed. Only one of the electrode positions that was analyzed at the Ni K-edge is shown in Figure 11. The quality of the data collected at another position was compromised by gas bubbles in the electrolyte within the field of view. The third region that was investigated had large amounts of the impurity phase shown in Figure 7. Interestingly, the impurity phase appears to be electrochemically active and undergoes a transformation during charge like the main phase does (Figure S5). The phase change shown in Figures 10 and 11 is not reversible upon discharge (relithiation). Figure 12 shows the average (bulk) XANES spectra during the first discharge. The cell was charged ex situ and discharged in situ to reduce the total X-ray dose. No systematic change was observed at the Cu Kedge during discharge (Figure 12a,b). This is consistent with our earlier work, where we proposed Cu was not redox active between 2.0 and 4.3 V versus Li/Li+.15 However, our earlier report on bulk XANES showed a shift in the Ni K-edge during discharge consistent with the reduction of Ni3+ to Ni2+. With TXM−XANES, the Ni K-edge appears to shift to a lower binding energy between 3.9 and 3.4 V. No additional shift is

Figure 12. (a) Average (bulk) Cu K-edge XANES from a cell charged ex situ and discharged in situ. (b) Magnified view of the absorption edge in panel a. (c) Average (bulk) Ni K-edge XANES from a cell charged ex situ and discharged in situ. (d) Magnified view of the absorption edge in panel c.

apparent, but this may be due to the energy resolution and signal-to-noise ratio of TXM−XANES being lower than those of bulk XANES. Nonetheless, it is clear that the structure permanently changes during the first charge and does not reform.



CONCLUSIONS The Immm orthorhombic structure of Li2Cu0.5Ni0.5O2 collapses with the extraction of