Cunico

Mar 17, 2023

Abstract High entropy oxides (HEOs) with chemically disordered multi-cation structure attract intensive interest as negative electrode materials for battery applications. The outstanding electrochemical performance has been attributed to the high-entropy stabilization and the so-called ‘cocktail effect’. However, the configurational entropy of the HEO, which is thermodynamically only metastable at room-temperature, is insufficient to drive the structural reversibility during conversion-type battery reaction, and the ‘cocktail effect’ has not been explained thus far. This work unveils the multi-cations synergy of the HEO Mg0.2Co0.2Ni0.2Cu0.2Zn0.2O at atomic and nanoscale during electrochemical reaction and explains the ‘cocktail effect’. The more electronegative elements form an electrochemically inert 3-dimensional metallic nano-network enabling electron transport. The electrochemical inactive cation stabilizes an oxide nanophase, which is semi-coherent with the metallic phase and accommodates Li+ ions. This self-assembled nanostructure enables stable cycling of micron-sized particles, which bypasses the need for nanoscale pre-modification required for conventional metal oxides in battery applications. This demonstrates elemental diversity is the key for optimizing multi-cation electrode materials. Introduction Metal oxides anode materials enable conversion reactions and provide high theoretical capacity in lithium ion batteries (LIBs) 1 , 2 , 3 . However, the poor electrical conductivity 4 , 5 , 6 and severe structural disintegration during the reaction 7 , 8 , which result in the notorious ‘size effect’ 9 , 10 , 11 , hinders the materials from practical application except after costly nano-structuring. Design of multi-cation oxides has been demonstrated to be an promising strategy for overcoming the drawbacks in the conversion reaction 12 , 13 , 14 . Adding an additional metal element has been proposed to enhance the electron conductivity of the pristine materials 12 , 15 and the reaction products 16 , 17 , and also to reduce the volume change while reacting with Li ions 12 , 16 . Recently, high entropy oxides (HEOs) as a novel class of multi-cation metal oxides have attracted intensive interest for battery applications 18 , 19 , 20 , 21 , 22 , 23 , 24 . In particular, it has been reported that micrometer-sized HEO particles show extraordinary long-term cycling stability at high capacity without a necessity for nanostructuring when used as anode material in LIBs 20 . It overcomes the ‘size effect’ of conventional oxides during the conversion reaction, resulting in a material with promising chances for practical application. The unexpected performance has been attributed to the randomly mixed five metal elements in a single-phase solid solution without chemical short-range order. This has been supposed to result in a high configurational entropy, stabilizing the crystal structure during lithium storage 18 , 20 , 23 , 24 . However, solely considering the entropy is insufficient to explain the reversibility of the conversion reaction and the high electron conductivity that is required for the long reaction path due to the large particle size. Ghigna et al. recently used operando XAS and realized that the HEO Mg0.2Co0.2Ni0.2Cu0.2Zn0.2O is not reacting fully reversible during electrochemical cycling 25 . This questions the simple high entropy interpretation. The term cocktail effect has been raised by many authors attempting to interpret the observed electrochemical improvements 18 , 19 , 26 , 27 , 28 . However, using term cocktail effect does not really bring any understanding for the synergistic effects of the cations during the electrochemical reaction. Why this HEO performs differently from well-studied binary multi-cation oxides is not clear. In this work, we unveiled the synergistic effects of the cations in the HEO Mg0.2Co0.2Ni0.2Cu0.2Zn0.2O during electrochemical reaction with lithium by a detailed analysis of the valance state of the metal elements and a comprehensive characterization of the micro structure of the material at different cycling states of the HEO-based anode using X-ray absorption spectroscopy (XAS) and analytical (scanning) transmission electron microscopy (S/TEM). This work further contributes to the idea of designing multi-cation materials for high performance ion batteries. Results Valence states and atomic coordination The HEO micro particles as the active material are prepared as composite electrodes and assembled in coin-type half-cells with a carbonate electrolyte using lithium foil as the counter and reference electrode. The half-cells are cycled in the voltage window of 0.01 V (fully discharged state) to 3.0 V (fully charged state) versus Li/Li+. XAS results including X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure Fourier transforms (EXAFS-FT) of the K edge of Co, Ni, Cu and Zn at different electrochemical reaction states are shown in Fig. 1 . The XANES and EXAFS-FT results show the expected 2+ valence state for Co, Ni, Cu and Zn in the as prepared HEO. The valance state of Mg could not be analyzed, because the energy of the Mg-K edge is below the minimum energy of the hard X-ray spectrum available at the beamline. Fig. 1: Valence state and atomic coordination analysis. a–d XANES of Co, Ni, Cu and Zn; e–h corresponding EXAFS–FT, solid lines represent the experimental data and dashed lines are references measured from standard samples provided by the beamline. ‘First discharged’ and ‘First charged’ sample refers to the samples discharged to 0.01 V and charged to 3.0 V (versus Li/Li+). After the first discharging, the XANES spectra of Co, Ni, Cu and Zn (solid violet lines in Fig. 1a–d ) match well to the corresponding metallic references. This proves that Co, Ni, Cu and Zn have been reduced to the metallic state, fitting to the expected conversion reaction. Consistently, the EXAFS-FTs (solid violet lines in Fig. 1e–h ) show that the metal-oxide (M-O) bond distances present in the as-prepared HEO disappeared and metal-metal (M-M) bond distances are present in the discharged sample, suggesting that oxygen was removed from the coordination shell of the Co, Ni, Cu and Zn atoms resulting in a metallic structure. The M-M distances of Co, Ni and Cu agree with metallic fcc Co, Ni and Cu. The M-M distances of Zn do not match metallic Zn (hcp, space group P63/mmc), but fit to Zn-Li distances in fcc LiZn. Alloying of Zn and Li during electrochemical cycling is well known and has been reported by various studies of ZnO batteries 29 , 30 , 31 . Interestingly, after recharging to 3.0 V, Ni and Cu stay in the metallic state as determined from the unchanged XANES pre-edges (Fig. 1b, c , red lines) and the M-M distances present in the EXAFS-FTs (Fig. 1f, g , red lines). Consistently, no M-O distances are observed for Ni and Cu in the EXAFS-FTs. The Co XANES reveals a significant reduction of the pre-edge peak (solid red line in Fig. 1a ) and the presence of both M-M and M-O distances in the Co EXAFS-FT (Fig. 1e , red line). This indicates that a large fraction of Co participates in the redox reaction while some Co remains in a metallic state. Distinct from the other metals, Zn is almost fully reoxidized to the 2+ state (Fig. 1d , red line). The Zn EXAFS-FT (Fig. 1h , red line) shows that the characteristic Zn-O distances reappear, while the Zn-Zn distances disappear. The atomic structure of the samples was further analyzed by electron pair distribution function (ePDF) (Supplementary Fig. 5 ) derived from select area electron diffraction (SAED) patterns. The evolution of the M-O-M and M-O distance and the creation of M-M bonds confirmed the conclusions deduced from the XAS analysis. Details can be found in the Supplementary Fig. 5 . Micro structure, elemental and phase distribution S/TEM analysis was applied to understand the micro structure at the atomic level for the different states. The intensity in high-angle annular dark field STEM (HAADF-STEM) images is roughly proportional to the atomic number squared, the atomic density and the sample thickness. For the as-prepared sample (Fig. 2a ), HAADF-STEM shows a homogeneous intensity distribution with only a few pores present, whereas the images of the cycled samples (Fig. 2b, c) exhibit a heterogeneous nanostructure. In particular, dendritic features can be clearly observed. This observation indicates a nanoscale phase/elemental separation in the electrode after dis/recharging. Fig. 2: Elemental distribution analysis of the as-prepared and cycled samples. a–c HAADF-STEM images of the as-prepared, 1st discharged (discharged to 0.01 V versus Li/Li+) and 1st charged (charged to 3.0 V versus Li/Li+) samples; d, e HAADF-STEM image and STEM-EELS elemental maps from an exemplary location of the 1st discharged sample; g, h HAADF image and EELS elemental maps from the 1st charged sample; f, i combined O and Cu maps. Elemental maps obtained by STEM based electron energy-loss spectroscopy (EELS) mapping explain the HAADF contrast. In the discharged sample (Fig. 2e ), Co, Ni and Cu are spatially well-correlated with each other and correspond to the bright areas (dendritic and bright granular features) in the HAADF image (Fig. 2d ). O and Mg and to some extent Zn are anti-correlated with Cu, Ni and Co and correspond to the dark regions in HAADF-STEM. For better visual guidance, we overlaid the Cu and O maps in Fig. 2f and highlighted two exemplary regions demonstrating the correlation. The result indicates that the discharged sample consists of a phase containing Cu, Ni and Co in the metallic state and a phase containing O, Mg, and presumably Li (which cannot be clearly distinguished in the EELS signal due to an overlap with the M-edges of the transition metals). The Zn distribution fits neither to the Cu-Ni-Co phase nor the MgO(LixO) phase and presumably corresponds to the LiZn distribution, present as indicated by the EXAFS-FT analysis. The phase separation is mostly maintained in the 1st charged sample (Fig. 2 g–i). Different from the discharged case, now Zn exhibits a strong spatial correlation with O and Mg. They form an oxide phase visible as the dark regions in the HAADF image. This is consistent with the 2+ oxidation state observed for Zn in XAS and also suggests that MgO and ZnO are mixed at the sub-nm or atomic level. Cu and Ni are still highly correlated and major features for Co are the same as Cu and Ni. They are anti-correlated with the oxide phase and form a metallic phase as bright granular regions and dendritic features in the HAADF image (Fig. 2c, g). This agrees with Cu, Ni and a large fraction of Co not being oxidized after recharging as determined from the XAS results. However, the granular distribution of Co is less sharp compared to the discharged state, implying some fraction of Co contributing to the oxide. In the high-resolution TEM (HRTEM) images (Fig. 3 ) of the cycled sample, the metal phase appears darker compared to the metal oxide phase. For the charged sample, the reflections in the fast Fourier transform (FFT) (Fig. 3c ) from a dendritic area (Fig. 3b , blue box) can be indexed as [101] zone axis of a fcc metallic structure with space group Fm-3m. The FFT (Fig. 3d ) of an area inside a grain (Fig. 3b , orange box) consists of two sets of reflections. The one with larger reciprocal space distances can be indexed as [101] zone axis of a metal fcc structure with an orientation rotated 59° anti-clockwise to the adjacent dendritic feature at the grain boundary. An enlarged image of the gain boundary is shown in Supplementary Fig. 15 . The unit cell parameter measured from the FFTs of the metallic phase (both the dendritic region and in the grain) is 3.6 Å close to fcc Cu (3.6 Å), Ni (3.5 Å) and Co (3.6 Å) single elemental metals. It confirms this metallic phase to be a CoNiCu alloy. The other set of reflections in the FFT from withni the grain can be indexed as [101] zone axis of a rock-salt metal oxide. Notably, the metal and the oxide phases in the grain exhibit a semi-epitaxial relationship. The same phenomenon was also observed in the discharged sample (Fig. 3 e–g). Fig. 3: Structural characterization of the cycled samples. a HRTEM image of the 1st charged sample, and inset showing a HAADF-STEM image from the same area; b enlarged image of the area marked in a; c, d FFTs of the areas marked in b; e HRTEM image of the discharged sample; f HAADF-STEM image of the same area as e; g FFT of the area marked in e. Scanning nanobeam electron diffraction (4D-STEM) was carried out to map the structure and phase distribution with a larger field of view compared to HRTEM; details are provided in the experimental section. As an example, Fig. 4 shows a 4D-STEM result of the discharged sample. The crystal orientation and phase map can be obtained by indexing the local diffraction patterns (e.g. Fig. 4d–f ). Correlating the HAADF image (Fig. 4a ) and the crystal orientation map (Fig. 4b ), one can clearly see that the dendritic features, which formed during lithiation, are dominantly located at the grain boundaries (e.g. the ones indicated by the arrows in Fig. 4a ). Fig. 4: Orientation and phase distribution analysis of the 1st discharged sample. a HAADF-STEM image; b orientation map obtained by indexing the diffraction patterns of the 4D-STEM data; c typical phase map corresponding to the area marked by the white rectangular box in b; d, e two exemplary diffraction patterns averaged from the area marked by the yellow and blue boxes in c, the circles mark the M and MO reflections; f diffraction pattern averaged from the area marked by the red dotted box in a; g, h line profiles along the arrow in the green and blue dotted rectangular boxes in f. The nanoscale phase map (Fig. 4c , taken from the location marked by the white dotted box in Fig. 4b ) shows the nanoscale separation of the metallic (green) and oxide (red) phase inside the grain and the CoNiCu dendritic-featured phase located at the grain boundaries. The local diffraction patterns (e.g. Fig. 4d , e, two typical ones taken from the regions marked by the yellow and blue boxes in Fig. 4c ) can be indexed to either a simple fcc structure with lattice parameters of crystalline Cu or a rock-salt fcc structure with lattice parameters of crystalline MgO. Due to the overlap of the phases in projection, diffraction spots from both phases are observed in every diffraction pattern. The colors in the phase map represent the phase dominating the signal. The result is consistent with the STEM-EELS elemental maps and also the XAS, ePDF and HRTEM results. All these indicate the formation of the alloy and oxide phase with lattice parameters very close to Cu and MgO. In addition of the CuNiCo phase and the oxide phase, a fcc LiZn phase is observed in the 4D-STEM data. For example, the diffraction pattern (Fig. 4f ) taken from the red box in Fig. 4a with the corresponding intensity profiles (Fig. 4g , h) along the arrows show broad diffraction peaks, which can only be explained by overlapping reflections of the fcc LiZn phase with the CoNiCu alloy and the oxide phase. While the similarity of the d-values between LiZn and the metal oxide together with the limited resolution in reciprocal space make it difficult to obtain a reliable phase map of LiZn, this manual analysis of the diffraction patterns (e.g. Fig. 4f ) unambiguously reveals the presence of LiZn crystals. Note that, the LiZn exhibits a clearly defined orientation relationship with the CuNiCo and the oxide, giving rise to a single crystalline appearance of the grains in Fig. 4b despite 3 different phases being present. Due to a large amount of metal atoms reduced during the 1st discharging and only partial reoxidation during the following charging, an equal amount of lithium ions corresponding to the reduction of these metal atoms are expected to exist both in the discharged and the charged material. Although it is difficult to directly detect nano phase Li2O due to its electron beam sensitivity and overlap of the Li-K edge with the M-edge of Cu, Co, Ni and the tail of the volume plasmon, Li2O with a size above a few ten nanometer can be directly observed in TEM, especially using electron diffraction with controlled dose. This has been widely used in many recent studies, even during in-situ TEM experiments 32 , 33 , 34 . Our 4D-STEM and SAED data confirms that no bulk Li2O is present in the cycled material. However, we cannot rule out the presence of Li2O nanophases. Therefore, it is reasonable to deduce that Li+ is incorporated in the oxide phase either as nano/sub-nanometer scale lithium oxide aggregates or atomically dispersed in the rock-salt structure giving rise to a Mg-Li-O configuration in the discharged state and a Mg-Co-Zn-Li-O configuration in the charged state. In a previous study the possibility of atomically dispersed lithium has been demonstrated, where a series of rock salt (Co,Cu,Mg,Ni,Zn)1-xLixO oxides (x < 0.30) was successfully synthesized and the ionic conductivity of the materials increases with increasing Li amount 35 . Figure  5d, e illustrates the epitaxial relationship observed for all phases inside the grains using the discharged state as an example. The metallic phase and the oxide phase exhibit the same crystal orientation. The large mismatch of the unit cell parameters (0.36 nm for the CoNiCu alloy and 0.42 nm for the oxide) results in a significant concentration of crystal defects (approximately one dislocation necessary every 7 (200) planes), which can be directly observed in the HRTEM images. The associated structural distortion can explain the disappearance of the crystal-field splitting (at 9670 eV and 9674 eV) corresponding to an octahedral environment in the Zn XANES spectrum of the charged sample (Fig. 1d ) 36 , 37 . The large lattice mismatch finally results in the observed nanoscale phase separation and small crystallite size. The {220} lattices of LiZn (0.22 nm) matche the {020} of the oxide phase (0.21 nm), and the {400} lattices of LiZn (0.16 nm) matche the {220} of the oxide phase (0.15 nm) as illustrated in Fig. 5d . Zn and partial Co are oxidized to Zn2+ and Co2+ and merge into the oxide phase in the charged state. Fig. 5: Epitaxial phase relationships in the discharged state. a–c Sketches of the atomic structure of the metal (M) phase, the metal oxide (MO) phase and LiZn phase; d orientation relationship of the oxide phase and LiZn; e a schematic overview of the structure. 3D electron conductive network Electron tomography was used to visualize the 3-dimensional (3D) distribution of the nanophases. The tomographic reconstruction from the 1st charged HEO is shown in Supplementary Movie  1 . Figure  6a shows a visualization from a cropped tomographic reconstruction (the corresponding volume rendering video is shown in Supplementary Movie  2 ). A series of volume rendered images of the area marked by the white box is shown in Supplementary Fig. 17 . The generated video after surface rendering of the grain boundary is shown in supplementary movie  3 . Supplementary Movie  4 shows the animation after combining the volume rendering of grain boundary adjacent area and the surface rendering of the grain boundary. From the Supplementary Movie  3 and 4 , it can be clearly observed that the cold color part (represents the alloy) has a two-dimension shape. Figure  6 b–f is reconstructed slices over a depth range of 51.2 nm normal to the viewing direction in Fig. 6a . The bright parts correspond to the metallic phase due to its significantly higher density than the oxide phase. The dendritic features (e.g. the one indicated by the yellow arrow) are present throughout the depth range revealing their platelet-like shape at the grain boundary instead of a 1-D dendrite structure. The platelets at the boundary of adjacent grains intersect and form a 3D network. Together with the metal phase inside the grains, a fine nanoscale 3D conductive network is formed, which penetrates throughout the micrometer-sized primary particle. The oxide phase fills the region between the metal network forming a 3D framework as counterpart. Fig. 6: Analysis of the 3D conductive network. a Volume rendering based on a tomographic reconstruction; b–f slices through the reconstructed volume normal to the viewing direction in a, the depth is denoted at the bottom right corner in the images. The metallic network provides a highly efficient electron-transfer path, while the metal oxide containing Li ions provides good ionic conductivity. To confirm this, we measured the conductivity of the as-prepared and the charged microparticles using an in-situ TEM-STM holder following the configuration shown in Fig. 7a . The as-prepared sample exhibits an insulating behavior in the voltage range of −15 to 15 V (Fig. 7b ). In contrast, significantly higher currents (0.1 µA) can pass through the charged particle at 15 V, indicating a significantly enhanced electrical conductivity (the high resistance in the voltage range of −5 to 5 V is presumably due to the contact resistance between the tip and particle). We conducted electrochemical impedance spectroscopy (EIS) measurements in a half-cell configuration to separate the contributions of ionic and electrical conductivity and provide more statistics than the microscopic in-situ investigation. One can see a smaller diameter of the semicircle in the EIS spectra of the charged sample (Fig. 7c , red) compared to that in the as-prepared sample (Fig. 7c , black). It confirms the enhanced electrical conductivity in the charged sample. Fig. 7: Conductivity test. a Schematic setup of the in-situ conductivity measurement (a full sketch of the TEM-STM holder is drawn in Supplementary Fig. 19 ); b the current-voltage curves of the as-prepared and 1st charged sample, the insets at the top-left and bottom-right show STEM images corresponding to the measurement of the as-prepared and the charged sample; c electrochemical impedance spectroscopy (EIS) results of the as-prepared and charged sample, the equivalent circuit is shown in the inset. Contributions C.K. and K.W. initialized the project. X.M. designed and supervised the experiments and data analysis. D.S., J.W. and B.B. provided the materials. K.W. prepared the TEM samples and performed the TEM measurements, X.H. performed the 3D tomography and Z.D. did the in-situ electron conductivity tests. W.H. and H.E. contributed the XAS analysis. K.W., Y.C. and Q.W. performed the EIS electrochemical performance testing and contributed to the data analysis. X.M., K.W. and C.K. carried out the main interpretation of the results. K.W. and X.M. wrote the first draft of the manuscript. All authors have contributed to the discussion and revision of the manuscript. Corresponding authors