Degradation mechanisms of lithium-rich nickel manganese cobalt oxide cathode thin films

Abstract

This paper reports a method to prepare Li-rich NMC (Li_1.2Mn_0.55Ni_0.15Co_0.1O₂) thin film cathodes for Li-ion batteries using RF magnetron sputtering and post-annealing in O₂. Thin film cathodes with high reversible capacities (260 mA h g⁻¹) and potential profiles similar to those of the powder material have been obtained. Structural and electrochemical studies show that the grown materials consist of a layered structure with trigonal symmetry in which Li/TM ordering is partially achieved. Using XPS we determine that the surface is comprised of Mn⁴⁺, Co³⁺ and Ni²⁺ cations inside an O²⁻ framework. The loss mechanisms of these electrodes have been studied after 184 cycles. The data after cycling shows the absence of Li/TM ordering, confirming that Li₂MnO₃ activation is irreversible, while electron diffraction data indicates extensive structural modifications upon cycling. In addition, we identified that the surface chemistry is dominated by inorganic species (LiF, Li_x′PO_y′F_z′, Li_xPF_y), along with small amounts of organic species with C–O and O–C=O groups such as PEO, LiOR and RCO₂Li. Moreover, XPS results indicate that Ni and Co migrate into the bulk while the reduction of Mn⁴⁺ into Mn³⁺ is clearly evidenced, as expected from the activation of Li₂MnO₃ domains and discharging to 2.5 V.

---

1. Introduction

The lithium-ion battery (LIB) is the current leading technology for powering all-electric vehicles (EVs).¹ A typical cathode layered material in a LIB has composition LiMO₂ (M = Ni, Mn, Co) and delivers about 150 mA h g⁻¹ when cycled between 3–4.2 V and close to 200 mA h g⁻¹ when charged up to 4.6 V (LiNi_1/3Mn_1/3Co_1/3O₂).¹ An exciting strategy was developed to obtain layered materials with higher capacities through the use of excess amounts of Li and Mn.^2–4 These structurally integrated Li₂MnO₃-stabilized composite structures, i.e. ‘layered–layered’ xLi_4/3Mn_2/3O₂ + (1 − x) LiMO₂ (M = Ni, Mn, Co), hereafter abbreviated as Li-rich NMC,^4–6 opened a route to achieve larger discharge capacities (250–300 mA h g⁻¹) with an operating voltage window of 2.0–4.8 V.^2–8

Unfortunately, Li-rich NMC materials suffer from several issues such as voltage fade and impedance rise.^9,10 The voltage fade is generally attributed to cation reordering upon cycling^2–8,11 possibly leading to the formation of spinel-like domains. The rise in electrode impedance concomitant to a decline in capacity can either be related to these internal structural changes or to the passivation of the surface due to electrolyte decomposition.¹² Indeed, operating at higher voltages (>4.2 V) is known to induce the partial oxidation of the electrolyte salt and solvents.^12–17 The electrolyte decomposition leads to the formation of a passive film on the active particles commonly referred to as cathode electrolyte interface (CEI) or solid electrolyte interphase (SEI).^12–17

One of the difficulties in understanding the electrodes interface is the complex surface chemistry of composite electrodes, which generally consist of the active material particles coated by conductive carbon particles, embedded in a polymer binder matrix. Thin film electrodes are an ideal platform to understand how the SEI layer influences capacity losses.^13,14 Not only do they have a well-defined geometric surface area but they lack binders and conductive additives, which facilitate to study the intrinsic properties of the cathode material using surface sensitive techniques such as X-ray photoelectron spectroscopy (XPS).^13,14 The first report on Li-rich NMC films yielded thin film electrodes with a reversible capacity of about 180 mA h g⁻¹,¹⁸ which is lower than the expected 250–300 mA h g⁻¹. Unfortunately, no voltage profiles were reported, which prevents a careful evaluation of the film structural-electrochemical properties.¹⁸ Another work discussed the properties of Li-rich NMC films prepared by pulsed laser deposition.¹⁹ These films showed a very large irreversible capacity during the first cycle, and exhibited potential profiles fairly different from those expected for the powder. In addition, both works did not investigate the degradation mechanisms of the electrodes after cycling.

Herein, we present the synthesis of Li-rich NMC thin film cathode with the full electrochemical storage capacity (260 mA h g⁻¹). The thin film cathodes have been characterized by galvanostatic cycling, X-ray diffraction (XRD), XPS and transmission electron microscopy (TEM) coupled with selected area electron diffraction (SAED) measurements. Moreover, to study the cathode material loss mechanisms we performed TEM/SAED analysis after prolonged cycling (184 cycles) and detailed XPS measurements at various cycling stages.

2. Experimental

2.1. Thin films preparation

Al₂O₃ substrates of 1 cm diameter and 380 μm thickness (99.6%, Valley Design) were coated on both sides with 0.5 μm Pt using DC magnetron sputtering.^13,14 Li-rich NMC thin films were deposited onto Pt-coated Al₂O₃ substrates by means of RF magnetron sputtering using a homemade target. HE5050 powder was provided by Argonne National Laboratory and purchased from TODA America, Inc. The composition of the powder was characterized as Li_1.2Mn_0.55Ni_0.15Co_0.1O₂, as has already been reported elsewhere.^5,8,11 The as-received powder was thoroughly ball-milled for 1 hour with yttria-stabilized zirconia balls, recovered, pressed as a 2′′pellet, and fired in air at 950 °C for 10 hours. The resulting pellet was bound to a Cu plate and used as sputter target using an anode-shielded sputter gun geometry. RF sputtering was carried inside a chamber with a base pressure of 10⁻⁶ Torr, using an Ar plasma at 80 W power and 20 or 25 mTorr pressure. A quartz microbalance was used before and after the deposition to measure the amount of mass deposited by unit of time. The thickness was back-calculated based on the expected density. The weight of the samples was also checked by weighting the samples multiple times with a Mettler microbalance (±10 μg), and was found in good agreement with the values expected from the QCM. A film of 0.5 μm deposited over 1 cm² has a weight of about 220 μg while thicker films of 1.5 μm weight about 660 μg. Based on the weighting accuracy, the error on the reported storage capacities is estimated to be lower than 5%. For XRD characterization, thick layers of ∼1.5 μm were prepared whereas thinner films of circa 0.5 μm were employed for other characterizations. The grown thin films were annealed in pure O₂ at 700 °C for 1 hour. After preparation, the samples were stored inside an Ar-filled glovebox.

2.2. Electrochemical characterization

Electrochemical characterization was conducted at 25 °C inside a thermostatic incubator using 2-electrode coin cells (2032 hardware, 316L, Hohsen) prepared inside an Ar (99.99%) glovebox. The cells consisted of a pure Li (Alfa Aesar) counter electrode, 1.2 M LiPF₆ in dimethyl carbonate (DMC) and ethylene carbonate (EC) (Novolyte) as the electrolyte, Celgard 2500 separators, and the thin film working electrode. Galvanostatic cycling was performed on a Maccor 4000 series. For ex situ characterizations, specimens were extracted from coin cells inside an Ar-filled glovebox and rinsed with anhydrous DMC (Sigma-Aldrich).

2.3. Structure and surface chemistry characterizations

A Scintag diffractometer with a Cu Kα source (45 kV, 40 mA) and fixed divergence slits or a PANalytical X'Pert Pro MPD with Mo Kα source (60 kV, 40 mA) and programmable variable slits were used to characterize the structure of the films. Surface chemistry was probed using a PHI 3056 XPS spectrometer equipped with Al Kα radiation (1486.6 eV) at a measurement pressure below 10⁻⁸ Torr. The energy scale calibration of the instrument is checked regularly using Ag and Au standards. Samples rinsed with anhydrous DMC were transferred from the glovebox to the XPS chamber under vacuum using an air-tight transfer device. High resolution scans were acquired at 350 W with 23.5 eV pass energy and 0.05 eV energy step. Survey scans were measured at 350 W with 93.9 eV pass energy and 0.3 eV energy step. The binding energies were shifted by setting the strongest carbon signal to 284.8 eV to account for charging. Li₂SiF₆ (Alfa Aesar) powder was measured as received for reference.

High resolution transmission electron microscopy (HRTEM) and SAED were collected using a Hitachi HF3300 TEM at 300 kV. TEM specimens were prepared by scratching off the powder from the thin film electrodes directly on the TEM Cu grid and transferred inside the microscope in a few minutes. To assist in the assignment, the SAED patterns were simulated by Desktop Microscopist software by using the unit cell of LiMO₂ structure with R3̄m space group and Li₂MnO₃ with C2/m space group, as reported elsewhere.²⁰

Inductively coupled plasma optical emission spectroscopy (ICP-OES) was measured on post-annealed films made at 10 and 20 mTorr for determining the Li:Mn:Ni:Co molar ratios using a Thermo Jarrell Ash IRIS unit. Three mL of freshly prepared aqua-regia (3 : 1 mixture of hydrochloric acid and nitric acid) were used to dissolve the films for analysis followed by dilution in 18.3 MΩ deionized water. A series of ICP standards were prepared by the serial dilution of standards purchased from Alfa Aesar.

3. Results and discussion

3.1. Structural, electrochemical and surface characterizations of the starting thin films

To control the stoichiometry and structure of Li-ion thin film materials, one can tune the starting target stoichiometry,^13,14 adjust the deposition pressure or vary the gas mixtures.^21–23 Sputtering the stoichiometric target at 10 and 20 mTorr resulted in Li/TM ratios of 1.63 and 1.43, respectively, as determined by ICP-OES. The ratio obtained at 20 mTorr is very close to that of the starting powder (1.41), whereas it is clear that deposition at lower pressures induce Li enrichment of the deposited material. Suitable growth conditions to prepare Li-rich NMC films with expected electrochemical properties were thereby employed by sputtering a stoichiometric Li-rich NMC target in pure Ar at deposition pressures of 20 or 25 mTorr.

The XRD patterns of the produced films post-annealed in pure O₂ are presented in Fig. 1. Apart from the reflections of the Al₂O₃ substrate and Pt current collector, blue and red vertical bars respectively, the patterns all show reflections attributed to a layered α-NaFeO₂-type structure with lattice parameters obtained by Rietveld refinement of a = b = 2.843 Å (±0.001) and c = 14.38 Å (±0.01), with a goodness-of-fit (χ²) = 6.21 (Fig. S1†). The c axis is more elongated than expected, yielding a c/a ratio of 5.06 larger than the typical value of 4.99.^2,8 The exact origins of this larger value are not known at this moment and could result from defects in the structure such as Li vacancies in the Li layers or Li excess in the transition metal (TM) layers, a work of future studies using grazing incidence or pole figure diffraction experiments. However, it should also be noted that these values have been obtained with the constraint that the Pt and Li rich NMC layers have the same height offset. If the NMC layer height is freely refined the lattice parameters shift to a = 2.836 Å (±0.001) and c = 14.23 Å (±0.02) and c/a = 5.02, with a goodness-of-fit (χ²) = 6.06 due to the additional height variable; this would seem reasonable, but the refined heights are unphysical due to an incorrect ordering of the layers. The results must be taken with care for thin film data, with few reflections to define the lattice parameters of the Li-rich NMC phase, and with complications from height offset, spherical harmonic preferred orientation correction (applied to all phases), and sample transparency; indeed, only three relatively strong reflections – (003), (101) and (104) – are present in the XRD data. However, for our analysis the Rietveld refinement is suggestive of a high c/a ratio for this thin film sample.

Fig. 1 X-ray diffraction patterns of Li-rich NMC thin films deposited at 20 and 25 mTorr sputter pressure and post-annealed in O₂. Reference patterns for Al₂O₃ and Pt are included as vertical blue and red bars, respectively. The strongest diffraction lines for the thin film material are indicated with Miller indices.

The microstructure of layered cathodes is normally related to the α-NaFeO₂ structure with trigonal symmetry (O3 phase). This structure consists of a cubic close-packed oxygen array with alternating slabs of Li and TM ions occupying the octahedral sites forming layers. For Li-rich Mn-rich NMC materials, the excess Li atoms occupy octahedral sites of the TM layers and may induce cation ordering resulting in Mn⁴⁺ locally confined to the Li₂MnO₃ phase. The LiMO₂ and Li₂MnO₃ structures have similar cubic closed-packed layers characterized by an interlayer spacing of ∼4.7 Å ((001) for layered monoclinic and (003) for layered trigonal), which facilitates the formation of a composite.⁸ While charging to 4.5 V, Li ions are extracted from the Li layers without disturbing the Li⁺/Mn⁴⁺ ordering in the TM layers. Beyond 4.5 V, the Li₂MnO₃ domains become electrochemically active and Li ions can be extracted from the TM layers on a plateau around 4.55 V. Previous studies suggest that the reaction proceeds, overall, according to Li₂MnO₃ → ‘Li₂O’ + MnO₂.^2–4 The net removal of ‘Li₂O’ corresponds to the extraction of the excess Li⁺ (+e⁻), and possibly O₂ gas evolution.^24,25 During discharge, the layered phase remains electrochemically active and the formed MnO₂ becomes active at lower potentials, typically below 3.5 V, providing the additional storage capacity up to 250–300 mA h g⁻¹.

The XRD patterns collected for these samples do not exhibit the superlattice reflections (Fig. 1) normally observed around 2θ = 20–25° originating from long range cation ordering. This is an interesting observation considering, as discussed latter, the electrochemical and electron microscopy data suggest local ordering of the Li and TM atoms. Furthermore, in the XRD data there is a lower (003)/(104) peak intensity ratio measured (0.7) than expected, compared to the typical (003)/(104) ratio of 1.2 or more.⁸ However, it is well-known that the crystal growth direction can be influenced by the substrate and current collector textures as well as the volume strain energy, and that there may be an insufficient number of grains present for a proper statistical average. The substrate shows a strong 〈111〉 preferred orientation of the Pt film. The (111) planes of face centered cubic Pt form a hexagonal array with closest Pt–Pt distances near to 2.78 Å (PDF 01-087-0642). This matches closely the a = b lattice parameter of LiCoO₂ and Li-rich NMC oxides with values around 2.82 (ref. 26) and 2.85 Å,⁸ respectively. It could therefore be expected that it is energetically more favorable to form crystallites with the R3̄m (003) planes parallel to the Pt (111) planes. However, it should be noted that the thin film crystal growth is achieved during post-annealing at high temperatures. Thus, the crystal growth does not purely result from the substrate texture but is strongly influenced by the film thickness due to volume strain energy considerations.²⁶ It has been shown for LiCoO₂ sputtered films that their (104) and (101) planes are preferentially aligned parallel to the substrate for large layer thicknesses (>1 μm) whereas extremely thin films (50 nm) show a strong preferential growth of the (003) planes parallel to the Pt (111) current collector.²⁶ In summary, the absence of random orientation of the crystallites due to a preferred orientation effect may significantly contribute to the lower than expected (003)/(104) intensity ratio.

The electrochemical potential profiles obtained during the first two cycles are presented in Fig. 2 for films grown at 20 or 25 mTorr. The potential profiles during charge exhibit a sloping profile until 4.35 V where a small plateau is present. The slope likely results from Li extraction from the Li slabs of the layered structure^2–4,8 whereas the latter feature is characteristic of the films and may result from a slight phase impurity. This process is reversible, as evidenced by the small plateau at 4.05 V during discharge. After these features, the potential reaches a wide plateau centered on 4.5 V, which is generally attributed to the conversion of the monoclinic domains,^2–4,8 and a final slope up to 4.9 V is measured. During discharge, a slope is measured till the small plateau around 4.05 V is reached, followed by a slope till 3.25 V which corresponds to Li insertion in the Li slabs.^2–4,8 At potentials below 3.3 V, the MnO₂ material formed by the conversion of Li₂MnO₃ participates in the reaction, as suggested by the change in slope. During the second cycle, the charge profile starts with a slope centered on 3.35 V corresponding to Li removal from Li_xMnO₂ domains, and is followed by a slope till 4.35 V corresponding to Li removal from the layered domains. The plateau at 4.35 V is present during the second cycle, and is followed by a slope till the potential reaches 4.9 V. Finally, the second discharge is the same as during the first cycle with a reversible capacity of 260 mA h g⁻¹, which indicates the good reversibility of the electrode thin film material after the activation of the first charge. These profiles match fairly well the profiles obtained for the parent powder.⁸ Despite the absence of superlattice reflections in the XRD pattern, the electrochemical results suggest that the starting films possess some Li/Mn ordering.

Fig. 2 Electrochemical potential profiles of Li-rich NMC thin film electrodes deposited at 20 (black) and 25 (gray) mTorr sputter pressure and post-annealed in O₂. Cycling was performed at 10 μA cm⁻² (C/20).

The presence of Li₂MnO₃ domains, suggested by the plateau measured around 4.5 V during the first charges, was further investigated by studying the microstructure of the films deposited at 20 and 25 mTorr using TEM/SAED. A representative HRTEM image from the Li-rich NMC thin film prepared at 20 mTorr is presented in Fig. 3a. The SAED pattern from the particle is shown in Fig. 3b. The distinct lattice fringes in the HRTEM image confirm the crystallinity of the sample. The SAED pattern also matches the simulated pattern from [01̄1] zone axis in a trigonal unit cell (not presented). Moreover, superlattice reflections can be observed at 1/3 interval between the fundamental trigonal spots, which indicate the presence of Li and TM ordering in the TM layer. The HRTEM photograph collected for a Li-rich NMC thin film prepared at 25 mTorr is presented in Fig. 3c. The SAED pattern collected from the whole crystal (Fig. 3d) matches with the [1̄12] zone axis in a trigonal symmetry. Moreover, it is clear that some superlattice spots, indicated by arrows, also appear at 1/3 spacing between the fundamental trigonal spots. Similar to the material made at 20 mTorr, these reflections are attributed to the cation ordering between Li and TM ions in the TM layers, and thereby supports the presence of a monoclinic Li₂MnO₃ phase discussed earlier and suggested by the electrochemical data.

Fig. 3 (a) High resolution TEM image and (b) SAED pattern for the Li-rich NMC thin films prepared at 20 mTorr. The SAED pattern shows the trigonal fundamental reflections from [01̄1] zone axis. (c) High resolution TEM image and (d) SAED pattern for the Li-rich NMC thin films prepared at 25 mTorr. The SAED pattern shows the trigonal fundamental reflections from [1̄12] zone axis. The arrows show the superlattice reflections indicating the cation-ordering in the transition metal layers possibly arising from the monoclinic C2/m phase.

The surface chemistry of the pristine thin film material prepared at 20 mTorr is presented in Fig. 4. An extensive study of various core levels was conducted to unambiguously determine the oxidation state of the TM in the starting structure. As seen in the upper left plot with Ni3p, Co3p, Li1s and Mn3p, assignments to Ni²⁺, Co³⁺, Li⁺ and Mn⁴⁺ are in good agreement with former reports with binding energies of 67.6, 61.4, 54.5 and 50.1 eV for Ni3p, Co3p, Li1s and Mn3p, respectively.^13,14,17 3s core levels have also been studied with obtained values of 103.0, 112.5, 89.1 and 84.8 for Co3s, Ni3s, Mn3s ⁵S and Mn3s ⁷S, respectively; Mn3s doublet is due to correlation interactions between unpaired 3d electrons and the 3s electron which can have parallel (⁷S) or antiparallel (⁵S) spins.^13,27,28 The energy separation of 4.30 eV for the Mn3s split levels is consistent with the presence Mn⁴⁺ and excludes the presence of Mn³⁺.^13,27,28 The 2p core levels, which are most often considered although they are more difficult to interpret due to multiplet splitting effects and ligand to metal charge transfer,^{13,27,29–32} confirm the assignments to Mn⁴⁺, Ni²⁺, and dominantly Co³⁺.^13,16,17 Binding energies of 642.7, 654.3, 780.6, 795.5, 855.2 and 872.2 eV are measured for Mn2p3/2, Mn2p1/2, Co2p3/2, Co2p1/2, Ni2p3/2 and Ni2p1/2, respectively, in good agreement with the presence of Mn⁴⁺, Co³⁺ and Ni²⁺. For Co2p core level, a very small proportion of Co²⁺ is evidenced by a weak Co2p3/2 shake-up line around 785.5 eV,¹⁶ which is possibly related to the presence of hydroxides evidenced in O1s at 531.6 eV along with O²⁻ from the lattice (529.6 eV) (see Fig. 7). Nonetheless, the majority of Co is Co³⁺, as evidenced by the shake-up line around 790 eV.¹⁶

Fig. 4 X-ray photoelectron spectra of a pristine film deposited at 20 mTorr pressure. C1s and O1s spectra are presented with the data for cycled electrodes in Fig. 7. Assignments are indicated on the figure.

3.2. Structural and surface characterizations of cycled electrodes

For the investigation of the degradation mechanisms of the cathode material, the films deposited at 20 mTorr were used. The evolution of the potential profiles upon cycling is presented in Fig. 5. As measured for powder electrode samples,¹⁰ the capacities decrease upon cycling concomitant to a decrease of the discharge voltage profiles (Fig. S2†). The capacity retention of the films is not as high as for the powder samples, which may result from the larger electrolyte/electrode weight ratio in the case of the films. Protecting the surface with a thin coating could significantly promote the capacity retention, as already demonstrated for high voltage spinel cathode thin films.^13,34 The response at higher voltages (3.0–4.0 V) gradually gets suppressed whereas a marked inflection is now visible around 2.7–2.9 V. The former may be related to the gradual loss of activity of the layered component whereas the appearance of the latter has been attributed to the formation of a spinel-like phase upon cycling due to TM cations rearrangement.^2,8,11,35

Fig. 5 Evolution during cycling of the electrochemical potential profiles of 20 mTorr deposited Li-rich NMC thin films for discharge and charge. Selected cycle numbers are 2, 20, 55, 80, 100, 125, 150 and 180, from high to lower total capacities, as indicated by the arrows and corresponding to the cycling experiment presented in Fig. S1.†

The structure evolution of the electrode material after 184 cycles studied by TEM/SAED is shown in Fig. 6. Representative HRTEM images (Fig. 6a and b) show that in some particles, different crystals coexist in a single grain. In addition, it was observed that after such extensive cycling period, the crystallinity of the particles was lost in some regions. Moreover, the SAED patterns collected from different particles (∼15 particles) did not reveal cation ordering spots of the C2/m structure (dashed arrows, Fig. 6c), as normally measured for the pristine material (Fig. 3b and d). However, the SAED patterns from several particles, of which a particle oriented along the [001] zone axis is shown in Fig. 6c, exhibit reflections which can be referred to as “forbidden” reflections (see the arrow pointing at the center of the triangle formed by fundamental O3 reflections) in the trigonal symmetry.^8,33 These reflections generally indicate the presence of TM ions in the Li layer and/or stacking faults in the lattice,^8,11,33 which might occur during extensive cycling and may be contributing to the voltage and capacity decay (Fig. 5 and S2†). The absence of monoclinic superlattice reflections in the examined SAED patterns (absence of spots at the positions of the dashed arrows) indicate that Li/TM ordering has been lost, likely during the initial high-voltage delithiation process correlated with the conversion of Li₂MnO₃ domains.³⁵

Fig. 6 (a and b) High resolution TEM images and (c) SAED pattern for the Li-rich NMC thin film electrode after 184 cycles. The HRTEM in (b) shows the presence of different crystallites oriented in different directions inside a single particle. (c) The SAED pattern along the [001] zone axis shows the trigonal fundamental reflections without any cation ordering (no spots visible at the expected locations for Li/TM ordering, dashed arrows). A ‘forbidden’ reflection in the trigonal (O3) symmetry is highlighted in the center of the triangle by the arrow.

In addition to the bulk structural rearrangements, changes in surface chemistry are known to play a crucial role for the stability of cathode materials and the loss of capacity.^13–17 The changes in surface chemistry probed at various cycling stages by XPS are presented in Fig. 7, and corresponding surface concentrations for the elements and functional groups are provided in Tables S1 and S2.† The surface is mostly comprised of LiPF₆ salt decomposition products (LiF, Li_x′PO_y′F_z′, Li_xPF_y). The surface has little C–O (286.5 eV in C1s, below 3.0 at%), O–C=O (289 eV in C1s, below 1.6 at%), or carbonate groups (289.5–290 eV), which is significantly different from the surface chemistry of the high voltage spinel LiMn_1.5Ni_0.5O₄ thin films that showed significantly more organic species.¹³ This difference is attributed to the lower voltage operation of the Li-rich NMC cathode compared to the high voltage spinel, and indicates that the organic solvent molecules do not decompose as extensively with Li-rich NMC cathodes. The ethers (C–O) are possibly related to polyethylene oxide (PEO) and lithium alkoxides (LiOR) whereas the O–C=O groups may result from carboxylates or esters.^13,15–17 Moreover, the amount of O–C=O groups eventually becomes nearly null at 35 cycles. In contrast, the amount of F is high and stable around 25 at%; at cycle 65 and 184 the extra F is possibly due to the presence of Li₂SiF₆-like species (see later). This F has to originate from the salt due to the lack of fluorinated binder. More than half of the F is in the form of LiF (685 eV in F1s) up to 35 cycles but the relative amount of LiF reduces as cycling further progresses. The concentration of F in Li_xPF_y (∼688 eV in F1s, ∼136 eV in P2p) increases to 8–9 at% at cycles 64 and 185. The amount of F in Li_x′PO_y′F_z′ (686.5–687 eV in F1s, 134.5–135 eV in P2p) is stable near 3 at% up to cycle 35. Upon further cycling an overlap with Li₂SiF₆ species, evidenced by an extremely high Si2p binding energy 105.3 eV (Fig. S3†) and a large binding energy for Li1s above 57 eV is measured. The P2p spectra show a progression from O-containing Li_x′PO_y′F_z′ species towards O-free Li_xPF_y species. These results also indicate that only small amounts of P and F are related to species assigned to Li_x′PO_y′F_z′ at ∼687 eV for cycles 65 and 184, and further supports the assignment of the strong signal at ∼687 eV in F1s to Li₂SiF₆-like species. In summary, the formation of a thickening SEI layer is likely to promote increasing internal impedance and capacity decay.

Fig. 7 X-ray photoelectron spectra (C1s, O1s, Li1s, Mn3p, P2p, F1s and Mn2p core levels) ofr electrodes cycled 12, 35, 64 and 184 times. The spectra of the starting thin film material are inserted when applicable. Assignments are indicated in the figures. *Assignment to Li₂SiF₆-like species are for samples cycled 65 and 184 times.

The presence of species similar to Li₂SiF₆ is further confirmed by the measurement of a reference Li₂SiF₆ powder (Fig. S4†) showing binding energies of 687.2 eV for F1s, 105.2 eV for Si–F⁻ in Si2p and 57.2 eV for Li1s – note that the O1s signal and Si2p signal around 102.5 eV result from the surface oxidation of the powder into sub-stoichiometric SiO_x species. Compounds similar to Li₂SiF₆, such as Na₂SiF₆³⁶ and K₂SiF₆³⁷ have Si2p3/2 binding energies of 104.3 and 104.6 eV, respectively. These Si species are likely resulting from the Al₂O₃ (99.6%) substrate material, which is known to contain amorphous silica at the grain boundaries (c.f. MSDS). The substrate is exposed to the electrolyte during cycling since the films show cracks formed after post-annealing (Fig. S5†). Interestingly, the Si content slowly increases upon cycling (Fig. S3 and Table S2†), possibly related to the reaction of SiO₂ with traces of HF generated in the electrolyte. The presence of these highly oxidized Si species highlights that Si oxides may convert upon prolonged exposure to the electrolyte into species similar to Li₂SiF₆. Recently, Philippe et al. studied the reaction of Si anode during the reaction with Li.³⁸ They argued that the native oxide converts during electrochemical cycling, and also during chemical aging, into species with a Si2p binding energy around 105–106 eV. On the basis of Si2p data, they attributed this signal to SiO_xF_y compounds resulting from the partial fluorination of SiO₂ by HF.³⁸ It is worthwhile to emphasize that Li1s, F1s (Fig. 7) as well as Si2p (Fig. S3†) binding energies, as well as the atomic concentrations (Tables S1 and S2†) are correlated to the formation of the high binding energy Si2p species. Moreover, the amount of O attributed to O = X (∼531.5 eV in O1s) for the samples cycled 65 and 184 times is too large to account for the P signal around 135 eV related to Li_x′PO_y′F_z′. The overall stoichiometry of the complex reaction product of SiO₂ is thereby tentatively written as Li₂SiF_6−2yO_y in a F rich chemistry.

The evolution of the metals concentrations and binding energies are discussed next. The concentrations measured for the pristine material are in fairly good agreement with the expected Li_1.2Mn_0.55Ni_0.15Co_0.1O₂ composition. The concentration of Mn decreases upon cycling. At cycle 1, Co and Ni are evidenced (not presented) but their concentrations, which are omitted in Table S2,† are less than 1 at%. The omission is due to the fact that measuring with Al Kα radiation induces a strong overlap of Ni2p with F Auger lines. For consistency, all concentrations are reported based on the quantification of survey scans measured with Al Kα to ensure that the same escape depth information is provided. As of cycle 12, no Co and Ni are evidenced on the surface whereas Mn concentration is still around 7 at%. Given the higher propensity of Mn for being dissolved in the electrolyte, this result suggests that Ni and Co migrate into the bulk structure. This modification of the surface may contribute to the fade in capacity and voltage of the electrode material. In addition, Mn⁴⁺ partly converts to Mn³⁺ during the first discharge, as suggested by both Mn2p and 3p core levels. The progression to Mn³⁺ is further confirmed by the Mn3s energy separation (not presented) which increases from 4.30 eV for the pristine electrode to 5.15 eV at 1 cycle and 5.20 eV at 12 cycles. These values are consistent with the co-existence of Mn³⁺ and Mn⁴⁺.¹³ Upon further cycling, Mn3s energy separations of 5.6 eV obtained, which suggests that all Mn is present on the surface as Mn³⁺. As known from earlier studies,¹⁵ Mn³⁺ may disproportionate into Mn⁴⁺O₂ and Mn²⁺O with the latter being possibly dissolved by residual HF from the electrolyte. Regarding the changes in Li chemistry, the signal of Li⁺ from the lattice has become very weak after 12 cycles. On the other hand, species related to LiF are evidenced, and in a smaller proportion Li is also present with C–O and O–C=O groups, as measured by Li1s ∼ 55.5 eV. At 35 cycles, Li1s signal can be well fitted with a single peak at 56 eV corresponding to LiF. Upon further cycling at 65 cycles, a signal at higher binding energies (∼57 eV), attributed to Li₂SiF₆-like species discussed above, is measured and overlaps with the signal of LiF whereas cycling 184 times shows dominantly the response of the former species.

4. Conclusions

Lithium-rich NMC thin film cathodes have been successfully prepared by RF magnetron sputtering in pure Ar. The obtained films exhibit a large reversible capacity of 260 mA h g⁻¹ and the expected voltage profiles. The investigation of the 2p, 3p and 3s XPS core levels reveals that the surface is mostly composed of Mn⁴⁺, Co³⁺ and Ni²⁺ cations in an O²⁻ environment, as previously reported and expected for the bulk material. The microstructure studied by TEM/SAED reveals that a trigonal layered structure is formed in which Li/Mn ordering is partially achieved within the TM layers. The voltage decay inherent to Mn-rich cathodes is also observed for the films. After prolonged cycling, no evidence for Li/TM ordering is measured by inspecting with TEM/SAED more than a dozen particles, which suggests that the initial Li₂MnO₃ domains are irreversibly converted. However, evidence for the presence of stacking faults and/or the presence of TM in the Li layer is found, which possibly contributes to the declines in capacity and voltage. The SEI of cycled discharged cathodes is rich in inorganic compounds (LiF, Li_x′PO_y′F_z′, Li_xPF_y), along with small amounts of species with C–O and O–C=O groups such as PEO, LiOR, carboxylates or esters. The signals for Ni and Co are not visible after few cycles. This suggests that Ni and Co migrate into the bulk, which may contribute to the instability of the electrode structure and the capacity fade. In the case of Mn, the reduction of Mn⁴⁺ into Mn³⁺ is clearly evidenced, as expected from the activation of Li₂MnO₃ domains.

Acknowledgements

Dr Daniel Abraham, ANL, is thankfully acknowledged for providing the powder material. Drs Sergiy Kalnaus and Lynn Trahey are gratefully acknowledged for fruitful discussions. This work was support by the U.S. Department of Energy (DOE), Basic Energy Sciences (BES), Materials Sciences and Engineering Division (LB, GMV, NJD – Synthesis, XRD, electrochemistry, XPS) and by the Vehicle Technologies Office (program managers Peter Faguy and David Howell) of the Office of Energy Efficiency and Renewable Energy at the U.S. Department of Energy (DM, CD, RAM, DLW – XRD, Microscopy). Microscopy research was supported via a user project supported by ORNL's Shared Research Equipment (ShaRE) User Program, which is also supported by DOE-BES.

References

1. J.-M. Tarascon and M. Armand, Nature, 2008, 451, 652–657.
2. C. S. Johnson, N. Li, C. Lefief, J. T. Vaughey and M. M. Thackeray, Chem. Mater., 2008, 20, 6095–6106.
3. M. M. Thackeray, S.-H. Kang, C. S. Johnson, J. T. Vaughey, R. Benedek and S. A. Hackney, J. Mater. Chem., 2007, 17, 3112–3125.
4. M. M. Thackeray, C. S. Johnson, J. T. Vaughey, N. Li and S. A. Hackney, J. Mater. Chem., 2006, 18, 2257–2261.
5. Y. Li, M. Bettge, B. Polzin, Y. Zhu, M. Balasubramanian and D. P. Abraham, J. Electrochem. Soc., 2013, 160, A3006–A3019.
6. D. Kim, G. Sandi, J. R. Croy, K. Gallagher, S. H. Kang, E. Lee, M. D. Slater, C. S. Johnson and M. M. Thackeray, J. Electrochem. Soc., 2013, 160, A31–A38.
7. J. R. Croy, K. G. Gallagher, M. Balasubramanian, Z. Chen, Y. Ren, D. Kim, S. H. Kang, D. W. Dees and M. M. Thackeray, J. Phys. Chem. C, 2013, 117, 6525–6536.
8. D. Mohanty, S. Kalnaus, R. A. Meisner, K. J. Rhodes, J. Li, E. A. Payzant, D. L. Wood III and C. Daniel, J. Power Sources, 2013, 229, 239–248.
9. S. K. Martha, J. Nanda, G. M. Veith and N. J. Dudney, J. Power Sources, 2012, 199, 220–226.
10. S. K. Martha, J. Nanda, G. M. Veith and N. J. Dudney, J. Power Sources, 2012, 216, 179–186.
11. D. Mohanty, A. S. Sefat, S. Kalnaus, J. Li, R. A. Meisner, D. P. Abraham, E. A. Payzant, D. L. Wood III and C. Daniel, J. Mater. Chem. A, 2013, 1, 6249–6261.
12. S. K. Martha, J. Nanda, Y. Kim, R. R. Unocic, S. Pannala and N. J. Dudney, J. Mater. Chem. A, 2013, 1, 5587–5595.
13. L. Baggetto, N. J. Dudney and G. M. Veith, Electrochim. Acta, 2013, 90, 135–147.
14. L. Baggetto, R. R. Unocic, N. J. Dudney and G. M. Veith, J. Power Sources, 2012, 211, 108–118.
15. K. Edström, T. Gustafsson and J. O. Thomas, Electrochim. Acta, 2004, 50, 397–403.
16. S. Verdier, L. El Ouatani, R. Dedryvère, F. Bonhomme, F. Biensan and D. Gonbeau, J. Electrochem. Soc., 2007, 154, A1088–A1099.
17. R. Dedryvère, D. Foix, S. Franger, S. Patoux, L. Daniel and D. Gonbeau, J. Phys. Chem. C, 2010, 114, 10999–11008.
18. H. Yim, W. Y. Kong, Y. C. Kim, S.-J. Yoon and J.-W. Choi, J. Solid State Chem., 2012, 196, 288–292.
19. B. Yan, J. Liu, B. Song, P. Xiao and L. Lu, Sci. Rep., 2013, 3, 3332.
20. D. Mohanty, P. Paudel and H. Gabrisch, Solid State Ionics, 2010, 181, 914–920.
21. N. J. Dudney, J. B. Bates, R. A. Zuhr, S. Young, J. D. Robertson, H. P. Jun and S. A. Hackney, J. Electrochem. Soc., 1999, 146, 2455–2464.
22. J. B. Bates, N. J. Dudney, B. Neudecker, A. Ueda and C. D. Evans, Solid State Ionics, 2000, 135, 33–45.
23. J. Fischer, C. Adelhelm, T. Bergfeldt, K. Chang, C. Ziebert, H. Leiste, M. Stüber, S. Ulrich, D. Music, B. Hallstedt and H. J. Seifert, Thin Solid Films, 2013, 528, 217–223.
24. Z. Lu and J. R. Dahn, J. Electrochem. Soc., 2002, 149, A815–A822.
25. A. R. Armstrong, M. Holzapfel, P. Novak, C. S. Johnson, S. H. Kang, M. M. Thackeray and P. G. Bruce, J. Am. Chem. Soc., 2006, 128, 8694–8698.
26. J. B. Bates, N. J. Dudney, B. J. Neudecker, F. X. Hart, H. P. Jun and S. A. Hackney, J. Electrochem. Soc., 2000, 147, 59–70.
27. P. S. Bagus, E. S. Ilton and C. J. Nelin, Surf. Sci. Rep., 2013, 68, 273–304.
28. A. J. Nelson, J. G. Reynolds and J. W. Roos, J. Vac. Sci. Technol., A, 2002, 18(4), 1072–1076.
29. H. W. Nesbitt and D. Banerjee, Am. Mineral., 1998, 83, 305–315.
30. V. M. Pessa, J. Phys. C: Solid State Phys., 1975, 8, 1769–1775.
31. H. W. Nesbitt, D. Legrand and G. M. Bancroft, Phys. Chem. Miner., 2000, 27, 357–366.
32. M. Lenglet, F. Hochu, J. Dürr and M. H. Tuilier, Solid State Commun., 1997, 104, 793–798.
33. H. Yu, R. Ishikawa, Y.-G. So, N. Shibata, T. Kudo, H. Zhou and Y. Ikuhara, Angew. Chem., 2013, 52, 5969–5973.
34. J. Li, L. Baggetto, S. K. Martha, G. M. Veith, J. Nanda, C. Liang and N. J. Dudney, Adv. Energy Mater., 2013, 3, 1275–1278.
35. D. Mohanty, A. Sefat, J. Li, R. A. Meisner, R. A. Justine, E. A. Payzant, D. P. Abraham, D. L. Wood and C. Daniel, Phys. Chem. Chem. Phys., 2013, 15, 19496–19509.
36. V. I. Nefedov, Y. V. Salyn, G. Leonhardt and R. Scheibe, J. Electron Spectrosc. Relat. Phenom., 1977, 10, 121–124.
37. W. E. Morgan and J. R. Van Wazer, J. Phys. Chem., 1973, 77, 964–969.
38. B. Philippe, R. Dedryvere, M. Gorgoi, H. Rensmo, D. Gonbeau and K. Edström, Chem. Mater., 2013, 25, 394–404.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4ra03674c

---