Operando Monitoring of Early Ni-mediated Surface. Reconstruction in Layered Lithiated Ni-Co-Mn

Transcription

1 Supporting Information Operando Monitoring of Early Ni-mediated Surface Reconstruction in Layered Lithiated Ni-Co-Mn Oxides Daniel Streich, Christoph Erk, Aurelie Guéguen, Philipp Müller, Frederick-Francois Chesneau, and Erik J. Berg * Paul Scherrer Institute, Electrochemistry Laboratory, 5232 Villigen PSI, Switzerland BASF SE, Ludwigshafen, Germany * Corresponding author: erik.jaemstorp-berg@psi.ch, S1

2 1. OEMS experiment with CAM-free electrode Figure S1.Comparison of CO 2 and O 2 gas evolution from LiCoO 2 / Li and Carbon / Li half-cells. The dotted red line marks the onset of O 2 evolution in the LiCoO 2 /Li half-cell. Among the investigated layered oxide cathode active materials, LiCoO 2 has the highest average potential during charge. Therefore, its galvanostatic cycling profile was selected as a mimick for a control experiment with a half-cell comprising a carbon black (38 wt% Super C65, 19 wt% SFG6, 43 wt% PVDF) instead of a layered oxide composite electrode. Figure S1 clearly shows that CO 2 and O 2 evolution from the carbon black/li half-cell are smaller by orders of magnitude compared to the LiCoO 2 /Li half-cell and that there is a distinct correlation between CO 2 and O 2 evolution at cell voltages > 4.5 V. These control experiments provide further proof that the active material is involved in CO 2 and O 2 evolution and strongly suggests that both CO 2 and O 2 formation depend on the availability of reactive oxygen species as common reactants. S2

3 2. OEMS experiment with oxygen-free electrolyte Figure S2. Comparison of the potential, as well as CO 2 and O 2 gas evolution profiles in O- containing and O-free electrolyte as observed by OEMS analysis of an NCM811/Li half-cell in 1 M LiPF 6 3:7 (w/w) EC:DEC and an NCM811/FP half-cell in 1 M LiPF 6 acetonitrile. Both halfcells were galvanostatically cycled at a specific current of 15 ma per g of active material. The dashed grey lines indicate the upper cut-off potentials for the NCM811/FP half-cell, which was lower than for the NCM811/Li half-cell due to the decreased stability of LiPF 6 in acetonitrile. S3

4 In order to investigate whether O 2 evolution requires O-containing electrolyte as a source of oxygen we performed a comparative control experiment with (1) a NCM811/Li half-cell in 1 M LiPF 6 in 3:7 (w/w) EC:DEC and (2) a NCM811/delithiated LiFePO 4 (FP) half-cell in 1 M LiPF 6 in acetonitrile. In the half-cell operated with acetonitrile electrolyte the only source of oxygen is the CAM. Figure S2 clearly shows that oxygen from EC:DEC is not necessary for O 2 evolution to take place. In fact, the O 2 evolution curves are very similar in both types of electrolyte, suggesting that O 2 formation predominantly involves oxygen originating from the CAM. Note that due to the limited anodic stability of LiPF 6 in acetonitrile the upper cut-off potentials were lower for the NCM811/FP than for the NCM811/Li cell (dashed light grey line in Figure S2). This control experiment proves that O 2 originates from CAM rather than electrolyte. 3. Definition and derivation of state of Ni oxidation (SNOX) Table S1. Individual transition metal contributions to theoretical CAM specific charges. charge Ni/t charge tot [mah g -1 ] charge Ni [mah g -1 ] ot [%] charge Co [mah g - 1 ] charge Co/tot [%] charge Mn [mah g -1 ] LiCoO NCM NCM NCM NCM charge Mn/tot [%] Table S1 summarizes the contributions of the individual transition metals to the overall theoretical specific charges of the investigated materials. The theoretical specific charge values are derived from the CAM compositions, presuming charge neutrality in the fully lithiated pristine materials with all Mn in formal oxidation state 4+, all Co in formal oxidation state 3+, and Ni compensating for the excess positive charge of Mn 4+ (i.e. [Ni 2+ ] = [Mn 4+ ], [Ni 3+ ] = [Ni] [Ni 2+ ]), due to transition metal reduction potentials increasing in the order Mn 4+/3+ Ni 3+/2+ Ni 4+/3+ Co 4+/3+. Assuming that charge neutrality is maintained and that, in line with the reduction potential series, transition metal oxidation occurs stepwise in the order Mn 3+/4+ Ni 2+/3+ Ni 3+/4+ Co 3+/4+ during charging, the momentary state of Ni oxidation was defined and derived as SNOX [t] = (specific charge [t] - charge Mn ) / ( charge Ni ) Although the assumption that the Ni redox is activated first during galvanostatic charge is generally supported by DFT calculations 1 3 and experimental work 4 6, deviations from this ordered stepwise oxidation of the transition metals (Ni 2+/3+ Ni 3+/4+ Co 3+/4+ ) could partly occur at high electrode potentials, but it should only have a minor effect on the general validity of our assumption concerning the first part of the galvanostatic charge. S4

5 4. Comparison of state of charge and state of Ni oxidation dependence of OEMS data Figure S3 clearly shows that the state of Ni oxidation (SNOX) provides a more consistent picture of the electrochemical and gas evolution behavior across all investigated cathode active materials than the state of charge (SOC) does. All CO 2 and O 2 evolution onsets fall into a very narrow SNOX range between 85 and 100 %. No curve crossing is observed and the rising and sloping characteristics consistently follow an order that correlates with active material Ni content. Figure S3. Dependence of potential and gas evolution on state of charge (SOC, a) and state of Ni oxidation (SNOX, b) during 1 st charge. All materials were electrochemically cycled at a specific current of 15 ma per g of active material between 2.0 and 4.7 V vs. Li + /Li in 3:7 (w/w) EC:DEC electrolyte containing 1 M LiPF 6. The dashed grey lines indicate the respective SOC and SNOX ranges into which the gas evolution onsets fall. The dashed black lines are fits of the sigmoidal curve (given in Eq. 1) to the O 2 evolution rate r(o 2 ) profiles (R > 0.99 for all fits). The insets show the linear correlation between the rise coefficient, χ (= 1/dx), and the Ni/Co ratio. S5

6 5. Estimation of reconstructed surface layer thicknesses According to Kojima et al., 7 a likely reconstruction mechanism for the active materials investigated here is the conversion of layered LiMO 2 (M = Ni, Co, Mn) into rock-salt structured MO as outlined in Figure S4. Based on the assumption that surface reconstruction gives rise to a homogeneous shell around a spherical active material particle, the surface layer thickness, d, can be estimated from the total amount of O released during both electrochemical cycles and the average particle diameter, d particle, as derived from the specific BET surface area, BET active Table S2), and the crystallographic density of the corresponding active material, ρ true (Table S2). 8 The average particle diameter, d particle, was estimated from the BET specific surface area using the well-known relationship This yields the average particle volume d particle = 6/(BET active ρ true ) (1) and the average particle mass V particle = 4/3 π (d particle /2) 3 (2) m particle = V ρ true (3) Assuming that one and two LiMO 2 are converted per detected CO 2 and O 2, respectively, the specific mass element affected by LiMO 2 conversion as well as the corresponding specific volume element dm = released O M r (4) dv = dm / ρ true (5) can be easily calculated from the total amounts of released O per g of active material obtained directly from the OEMS measurements. Multiplication of the specific volume element with the average particle mass yields the converted volume element per particle dv particle = dv m particle (6) If the outer radii of the LiMO 2 sphere and the MO shell are defined as r sphere and r shell = d/2 respectively, these radii need to satisfy the condition This can be rearranged to finally yielding 4/3 π ([d particle /2] 3 r sphere 3 ) = dv particle (7) r sphere = ((d particle /2) 3-3 dv particle / (4 π)) 1/3 (8) S6

7 d = r shell r sphere = d particle /2 r sphere (9) As apparent from Figure S4. Model for estimating reconstructed surface layer thickness from the amount of released CO 2 and O 2. The resulting estimated surface layer thicknesses considering only O 2 release, d(o 2 ), are on the order of a few Å. If CO 2 release is also taken into account (d(o 2 +CO 2 )), these values increase by about an order of magnitude. Due to limited OEMS capture efficiencies and the possibility that structural oxygen loss may yield further, non-monitored species these estimates should be understood as lower limits. Table S2. Surface layer thicknesses estimated from released O 2 and CO 2. * ρ true d ** M r released released O *** [g cm -3 [g mol - O *** d (O (O 2 ) ] [µm] 2) (O 1 ] [µmol g -1 [nm] 2 + CO 2 ) ] [µmol g -1 ] NCM NCM NCM NCM * from XRD (data not shown); ** Equation 1, BET active according to Table S 3; *** total amounts released during 1 st and 2 nd cycle. d (O 2 + CO 2 ) [nm] S7

8 6. X-ray Powder Diffraction (XRD) of pristine and galvanostatically cycled NCM811 (charge cut-off 4.7 V) electrodes Figure S5 shows that structural analysis by XRD reveals no structural difference between pristine and cycled electrodes. No formation of rock-salt or spinel crystalline phases were observed. The only observable differences between a cycled and a reference NCM811 electrode are very minor peak shifts that can be explained by a slight deviation in state of charge between these electrodes. This deviation arises because the common state of charge was calculated as the specific charge difference between charge and discharge of the cycled electrode, neglecting any loss of charge due to parasitic side-reactions. This control experiment shows that structural analysis by XRD is incapable of identifying and quantifying surface species as the size of the formed layers is expected to be below 3 nm, which is below the detection limit of standard XRD devices. Figure S5. Structural analysis of NCM811 electrodes by X-ray diffraction. The cycled electrode was charged up to 4.7 V followed by a discharge to 2 V vs. Li + /Li. The reference electrode was charged to the state of charge determined from the specific charge difference between the charge and discharge of the cycled electrode. XRD spectra were normalized with respect to the (003) diffraction peak maximum of the reference electrode. Lines for LiMn 2 O 4 spinel and NiO rock-salt are included for comparison. S8

9 7. Transmission Electron Microscopy (TEM) of NCM811 secondary particles before and after cycling In order to verify that our materials undergo surface reconstruction from rhombohedral LiMO 2 (R-3m) to defect spinel (Fd-3m) and rock-salt (Fm-3m) phases TEM micrographs of Pristine NCM811 (Figure S6), NCM811 cycled galvanostatically 2 times to 4.7 V at C/10 (Figure S7), and NCM811 galvanostatically cycled 100 times to 4.2 vs. Li + /Li at C/3 (Figure S8) were acquired. Electron transparent lamellae of fully lithiated NCM811 secondary particles from the pristine powder or the cycled electrodes with an approximate thickness of 100 nm were produced by Focused Ion Beam (FIB) milling (Strata 400 DualBeam; FEI Company, Hillsboro, USA) using the in-situ lift-out technique. The samples were imaged by TEM using a Tecnai Osiris machine (FEI Company) operated at 200 kev under bright-field (BF) conditions. The crystallinity of the samples was studied by selected area electron diffraction (SAD) as well as by high-resolution (HR)-TEM. Images, Fast Fourier Transforms (FFT) of the HRTEM images and diffraction patterns were evaluated using the item (Olympus, Tokyo, Japan, version: ), Prodas (Proscope, Gangelt, Germany, version: 1.4), TIA (FEI, version: ), Eva (Bruker, Germany) and Digital Micrograph (Gatan, Pleasanton, USA, version 2.11) software packages. Figure S6. TEM micrographs of pristine NCM811 powder (a) in overview and (b) at high magnification. (c) Electron diffractogram from a large cross-section of the FIB milled secondary particle from both pristine and galvanostatically cycled NCM811 powder. Pristine powder of rhombohedral NCM811 consists of nanometer-sized (> 100 nm) primary particles, which are merged into micrometer-sized (> 10 m) secondary particles (Figure S6 a). Images taken at higher magnification (Figure S6 b) and Electron diffractograms collected over a large area of the secondary particle cross-section (Figure S6 c) show that no oxygen deficient cubic or MO-type phases could be resolved. These results are fully consistent with previous studies by Hwang et al 9 11, Kim et al. 12, and Noh et al. 13 S9

10 Figure S7. TEM micrographs of NCM811 secondary and primary particles cycled to 4.7 V vs Li + /Li (a) in overview and (b) at high magnification. Insets show the fast Fourier transform (FFT) image analysis of the respective areas. Secondary particles from electrodes cycled twice to high potential of 4.7 V (Figure S7 a) generally retain their structure. However, closer inspection at high magnification of primary particles located at the surface provides clear evidence of cation mixed phases (Figure S7 b and c), which is the first step in the formation of surface-reconstructed spinel and cubic rock-salt phases. Delithiating the NCM811 obviously leads to Li/Ni layer mixing (displayed as apparent layer distance ~0.24 nm, Figure S7 c) which starts at the surface while the more ordered rhombohedral phase (apparent layer distance ~0.47 nm, Figure S7 c) is gradually retained towards the center of the primary particles. Most importantly, the formation of the observed cation mixed phases occurs inhomogeneously and to different extents along the NCM811 secondary particle surface. The subsequent transformation into the defect spinel and cubic rocksalt reconstructed phases most likely also occurs inhomogeneously and is probably related to the exposed primary particle facet (as evidenced below in samples retrieved after 100 cycles). One has to keep in mind that in this TEM characterization, we only investigated one secondary NCM811 particle at very few locations. In any case, these results are fully consistent with previous studies by Hwang et al 9 11, Kim et al. 12, and Noh et al. 13 Figure S8. TEM micrographs of NCM811 (a) secondary and (b-c) primary particles galvanostatically C/3 cycled 100 times to 4.2 V vs Li + /Li. Insets show (b) the selected area electron diffractograms of single primary particles and (c) the fast Fourier transform (FFT) image analysis of a primary particle at the surface. S10

11 After 100 cycles to 4.2 V, the secondary particles still retain their shape (Figure S8 a). Electron diffractograms, collected over a large area of the secondary particle cross-section (Figure S6 c), mainly evidence the rhombohedral layered phase LiMO 2, although the presence of a second phase may be anticipated. However, closer inspection of the NCM811 particles cycled 100 times to 4.2 V reveals that several primary particles close to the surface almost fully converted into the cubic rock-salt phase (Figure S8 b) while primary particles in the bulk of the secondary particle remain in the initial rhombohedral phase. At higher magnifications (Figure S8 c) it is obvious that the cubic MO phase may persist hundreds of nanometers from the surface and inwards. In conclusion, oxygen-deficient defect spinel and rock-salt phases appear to nucleate at the surface of the secondary NCM particles and gradually grow inwards during cycling. Already moderate electrochemical cycling of NCMs in a commercially relevant potential window of V vs. Li + /Li induces significantly sized (> 100 nm) domains of reconstructed MO phases at the surface of the secondary particles. The results are fully consistent with previous studies by Hwang et al 9 11, Kim et al. 12, and Noh et al Complementary data and visualizations of OEMS results A complete compilation of OEMS data for the first 2 galvanostatic cycles is provided in Figure S6 and complemented by an overview of selected active material and electrode properties summarized in Table S3. LiCoO 2 is included in the compilation and summary to provide a comparison with a state-of-the art Li-ion battery CAM. Figure S9. Electrochemical behavior and gas evolution characteristics with respect to cathode potential. All materials were galvanostatically cycled at a specific current of 15 ma per g of active material between 2.0 and 4.7 V vs. Li + /Li in 3:7 (w/w) EC:DEC electrolyte containing 1 M LiPF 6. Momentary specific charges, CO 2 and O 2 evolution rates, r(co 2 ) and r(o 2 ), are plotted against potential vs. Li + /Li for the 1 st charge (a), 1 st discharge (b), 2 nd charge (c), and 2 nd discharge (d). S11

12 Table S3. Active material and electrode properties. charge tot [mah g -1 ] BET active [m 2 g -1 ] BET electrode * [m 2 g -1 ] avg. potential 1 st (2 nd ) charge [V] 4.3 V V LiCoO (4.07) 4.54 (4.53) NCM (3.93) 4.52 (4.51) NCM (3.93) 4.50 (4.50) NCM (3.94) 4.51 (4.49) NCM (3.90) 4.47 (4.44) *calculated from individual component BET specific surface areas: BET electrode = 93% BET active % BET SuperC % BET SFG6 + 3% BET PVDF. In accordance with the suppliers specifications: BET SuperC65 = 62 m 2 g -1, BET SFG6 = 17 m 2 g -1, BET PVDF = 0 m 2 g -1. References (1) Dixit, M.; Kosa, M.; Lavi, O. S.; Markovsky, B.; Aurbach, D.; Major, D. T. Thermodynamic and Kinetic Studies of LiNi0.5Co0.2Mn0.3O2 as a Positive Electrode Material for Li-Ion Batteries Using First Principles. Phys. Chem. Chem. Phys. 2016, 18, (2) Hwang, B. J.; Tsai, Y. W.; Carlier, D.; Ceder, G. A Combined Computational/experimental Study on LiNi1/3Co 1/3Mn1/3O2. Chem. Mater. 2003, 15, (3) Hoang, K.; Johannes, M. D. Defect Physics and Chemistry in Layered Mixed Transition Metal Oxide Cathode Materials: (Ni,Co,Mn) vs (Ni,Co,Al). Chem. Mater. 2016, 28, (4) Mauger, A.; Gendron, F.; Julien, C. M. Magnetic Properties of LixNiyMnyCo1 2yO2 ( y 0.5, 0 x 1). J. Alloys Compd. 2012, 520, (5) Tsai, Y. W.; Lee, J. F.; Liu, D. G.; Hwang, B. J. In-Situ X-Ray Absorption Spectroscopy Investigations of a Layered Lithium Batteries (6) Balasubramanian, M.; Sun, X.; Yang, X. Q.; Mcbreen, J. In Situ X-Ray Absorption Studies of a High-Rate LiNi Co O 2 Cathode Material. J. Electrochem. Soc. 2000, 147, (7) Kojima, Y.; Muto, S.; Tatsumi, K.; Kondo, H.; Oka, H.; Horibuchi, K.; Ukyo, Y. Degradation Analysis of a Ni-Based Layered Positive-Electrode Active Material Cycled at Elevated Temperatures Studied by Scanning Transmission Electron Microscopy and Electron Energy-Loss Spectroscopy. J. Power Sources 2011, 196, (8) Strehle, B.; Kleiner, K.; Jung, R.; Chesneau, F.; Mendez, M.; Gasteiger, H. A. The Role of Oxygen Release from Li- and Mn-Rich Layered Oxides During the First Cycles Investigated by On-Line Electrochemical Mass Spectrometry. J. Electrochem. Soc. 2017, 164, (9) Hwang, S.; Kim, S. Y.; Chung, K. Y.; Stach, E. A.; Kim, S. M.; Chang, W. Determination of the Mechanism and Extent of Surface Degradation in Ni-Based Cathode Materials after Repeated Electrochemical Cycling. APL Mater. 2016, 4, S12

13 (10) Hwang, S.; Kim, S. M.; Bak, S. M.; Chung, K. Y.; Chang, W. Investigating the Reversibility of Structural Modifications of LixNiyMnzCo1-Y-zO2 Cathode Materials during Initial Charge/Discharge, at Multiple Length Scales. Chem. Mater. 2015, 27, (11) Hwang, S.; Kim, S. Y. M.; Bak, S. M.; Cho, B. W.; Chung, K. Y.; Lee, J. Y.; Stach, E. A.; Chang, W. Using Real-Time Electron Microscopy To Explore the Effects of Transition- Metal Composition on the Local Thermal Stability in Charged LixNiyMnzCo1-Y-zO2 Cathode Materials. Chem. Mater. 2015, 27, (12) Kim, N. Y.; Yim, T.; Song, J. H.; Yu, J.-S.; Lee, Z. Microstructural Study on Degradation Mechanism of Layered LiNi0.6Co0.2Mn0.2O2 Cathode Materials by Analytical Transmission Electron Microscopy. J. Power Sources 2016, 307, (13) Noh, H. J.; Youn, S.; Yoon, C. S.; Sun, Y. K. Comparison of the Structural and Electrochemical Properties of Layered Li[NixCoyMnz]O2 (X = 1/3, 0.5, 0.6, 0.7, 0.8 and 0.85) Cathode Material for Lithium-Ion Batteries. J. Power Sources 2013, 233, S13