Lithium-ion attack on yttrium oxide in the presence of copper powder during Li plating in a super-concentrated electrolyte

Li plating/stripping on Cu and Y2O3 (Cu + Y2O3) electrodes was examined in a super-concentrated electrolyte of lithium bis(fluorosulfonyl)amide and methylphenylamino-di(trifluoroethyl) phosphate. In principle, Li+ ions cannot intercalate into a Y2O3 crystal because its intercalation potential obtained from first-principles calculations is −1.02 V vs. Li+/Li. However, a drastic decrease in the electrode potential and a subsequent constant-potential region were observed during Li plating onto a Cu + Y2O3 electrode, suggesting that Li+ interacted with Y2O3. X-ray diffraction (XRD) patterns and X-ray absorption fine structure (XAFS) spectra of the Cu + Y2O3 electrodes after the Li plating were recorded to verify this phenomenon. The XRD and XAFS results indicated that the crystallinity of Y2O3 crystals was lowered because of attack by Li+ ions or that the Y2O3 crystal structure was broken while the +3 valence state of Y was maintained.


Introduction
Li + -ion batteries (LIBs) are one of the most important rechargeable power devices and are essential to hybrid and electric vehicles. Conventional LIBs composed of lithium cobalt oxide (LiCoO 2 ) and graphite operate at 3.7 V. One of the challenges in the development of next-generation battery technologies is increasing the cell voltage. The use of lithium transition-metal oxides such as LiNi 0.5 Mn 1.5 O 4 , 1-3 LiCoPO 4 , [4][5][6] or LiCoMnO 4 (ref. [7][8][9] instead of LiCoO 2 as a cathode material results in high-voltage cells with an operating voltage of $5 V. By contrast, for the anode active material, only Li metal can be used to increase the cell voltage. Therefore, research into active materials that can exhibit a redox couple at potentials lower than the Li plating/stripping standard level is a challenge in the development of battery technologies. Numerous metal oxides have been investigated as anode materials. 10 For example, niobium(V) oxide (Nb 2 O 5 ) exhibits oxidation/reduction behavior expressed by the equation Nb 2 O 5 + 2Li + + 2e À # Li 2 Nb 2 O 5 at potentials between 2.2 V and 1.6 V vs. Li + /Li. The valence of Nb changes from +5 to +4 during Li + intercalation. 11,12 Thus, for metal oxides to be used as anode materials, they must have more than one valence. However, yttrium oxide (Y 2 O 3 ) has only one valence of +3. 13,14 If Li + ions are intercalated into a Y 2 O 3 crystal, either yttrium ions will be reduced to the metal state or oxygen ions will be oxidized. First-principles calculations show that the intercalation potential of Li + into Y 2 O 3 is À1.02 V (Section 3.2). Therefore, Y 2 O 3 is theoretically inactive toward Li + insertion; Li electrodeposition occurs preferentially.
In our previous reports, 15,16 we examined Li intercalation into graphite in some super-concentrated electrolytes of lithium bis(uorosulfonyl)amide (LiFSA) and self-extinguishing solvents. Li plating/stripping has also been investigated in this electrolyte using an electrochemical cell with Cu and Li metal electrodes. A decrease in the potential of the Cu electrode because of solid electrolyte interphase (SEI) formation was observed at the initial stage of the Li plating test. Aer the SEI had formed, a constant-potential region due to Li electrodeposition appeared. In the later stage of Li plating, a sudden decrease of the Cu electrode potential was observed (Fig. 1). This phenomenon was caused by a large increase in interfacial resistance between the electrolyte and the Cu electrode. When a similar large decrease in potential occurs in a Cu + Y 2 O 3 composite electrode, Li + ions are speculated to interact with Y 2 O 3 . The Li + intercalation should be observed at À1 V vs. Li + /Li if the process occurs as planned (blue dotted line in Fig. 1). In the present paper, we prepared a composite electrode of Cu and Y 2 O 3 powders and examined whether Li + ions insert into the Y 2 O 3 crystallites.

Materials
Cu powder was obtained from Kojundo Chemical (average diameter: 1 mm). Y 2 O 3 nanopowder was purchased from Sigma-Aldrich (average diameter: <100 nm). LiFSA as a Li + salt and methylphenylamino-di(triuoroethyl) phosphate (PNMePh) as a solvent were obtained from Kishida Chemicals and Tosoh Finechem Corp., respectively. LiFSA was used aer drying at 150 C. The water content in the PNMePh liquid was 37 ppm. Vinylene carbonate (VC), an electrolyte additive, 17,18 was purchased from Kishida Chemicals (battery grade). The chemical structure of the three compounds was displayed in the ESI (Fig. S1 †). The X-ray diffraction (XRD) pattern for the Y 2 O 3 powder is shown in Fig. S2. †

Electrochemical measurements
The Cu + Y 2 O 3 composite electrode was prepared as follows. The Cu and Y 2 O 3 powders were mixed with polyvinylidene diuoride (PVdF, #9350, Kureha) and N-methyl pyrrolidone (NMP; Wako Chemicals) using a kneading machine (ARE-310, Thinky Co.). The mixing speed and time were 2200 rpm and 5 min, respectively. The electrode slurry was spread onto a Cu current collector (20 mm thick) using a doctor-blade technique and dried at 150 C under vacuum for 3 h. The nal mixing ratio of Cu, Y 2 O 3 , and PVdF was 89% : 8.5% : 2.5% by weight.
The Cu + Y 2 O 3 electrode sheet was pressed to obtain a 14 mm diameter disk electrode with a thickness of 55 mm (Fig. S3 †). The loadings of the Cu and Y 2 O 3 powders were 28-31 mg cm À2 . A half coin cell (Fig. S4 †) was fabricated using the 14 mm diameter disk electrode, two lter papers (200 mm thick, ADVANTEC, 5C), a Li metal disk 18 mm in diameter and 0.4 mm thick (Honjo Metal), and a super-concentrated electrolyte. The superconcentrated electrolyte was prepared by mixing LiFSA and PNMePh in a molar ratio of 1 : 3. The cells were fabricated using a dilute electrolyte with the molar ratio of 1 : 8, and 1 mol L À1 LiPF 6 in a mixture of ethylene carbonate (EC) and dimethyl carbonate (DMC) (Kishida Chemicals, battery grade) in the same manner. The half coin cell was cycled with a chargedischarge apparatus (Hokuto Denko, HJ1001SM8A). The charge (Li plating) was carried out under a current of 0.05 mA. The Li plating/stripping test was also performed using an electrolyte based on VC at a concentration of 1 vol%. Charging was started under a constant current mode at 0.05 mA for 24 h, and the discharge (Li stripping) to 2.0 V was conducted in constantcurrent mode at 0.05 mA.

Cyclic voltammetry (CV)
The half coin cells were cycled using a potentiometer (IVIUM Technologies, IVIUMSTST-XR) to obtain CV proles of Li plating/stripping and Li intercalation in the highly concentrated electrolyte. A Li-metal anode was used as the counter and reference electrodes. CV of the electrolytes was conducted under a sweep rate of 0.2 mV s À1 at 25 C. The CV curves for the superconcentrated electrolyte are displayed in Fig. S5. †

Electrochemical impedance spectroscopy (EIS)
The Cu + Y 2 O 3 electrode in the coin cell was used as a working electrode, and a Li-metal anode was used as the counter and reference electrodes. EIS was conducted at 25 C in the frequency range from 7 MHz to 20 mHz with a BioLogic SP-300 potentiostat to investigate the formation of a passive layer on the Cu + Y 2 O 3 electrode. It was also conducted between 25 and 51 C with the amplitude of the sinusoidal potential adjusted in 10 mV increments. Impedance spectra were t using the EC-Lab Zt soware.

Analysis
The Cu + Y 2 O 3 electrodes before and aer Li plating/stripping were analyzed by XRD (Rigaku, Ultima IV). The Cu + Y 2 O 3 electrode sample that contained the electrolyte was affixed to a glassy cell holder using Kapton tape in an Ar-lled glove box. The XRD prole was recorded in the 2q range from 10 to 80 at a sweeping rate of 10 min À1 . The morphology of Li plating in the sample was examined by eld-emission scanning electron microscopy (FE-SEM, JEOL, JSM-7000F). The Cu + Y 2 O 3 electrode was rinsed with DMC three times in an Ar-lled glove box and dried under vacuum. Energy-dispersive X-ray spectroscopy was used to identify passivation lms on the Cu, electrodeposited Li, and Y 2 O 3 . 7 Li nuclear magnetic resonance (NMR) spectroscopy was used to examine Li compounds aer the Li plating tests; measurements were performed using a Bruker NMR spectrometer (AVANCE 400). In an argon-lled glove box, the Cu and Y 2 O 3 powders were stripped from Cu foil and washed several times with DMC. Aer drying under vacuum, the powders were packed into a ZrO 2 tube. The NMR reference material was 1 mol L À1 LiCl aqueous solution. The measurement conditions are summarized in Table S1. † X-ray absorption ne structure (XAFS) analysis of the Y 2 O 3 powder aer the Li plating test was conducted through-the-law to identify the valence of Y aer the Li plating. In an Ar-lled glove box, the Cu and Y 2 O 3 powders were stripped from Cu foil, washed with DMC several times, and dried under vacuum. XAFS measurements were carried out at the Toray Research Center Inc (Table S2 †).

Computational details
First-principles calculations were performed for cubic Y 2 O 3 crystals using the Vienna Ab initio Simulation Package (VASP) 19 with the generalized gradient approximation of Perdew-Burke-Ernzerhof (PBE) 20 for the exchange-correlation terms in the density functional theory and using the projector-augmented wave (PAW) method 21 for describing the ion-electron interaction. The cutoff energy for the plane-wave expansion was set to 500 eV. The spin polarization was considered in all calculations because the total number of electrons may become odd upon Li insertion. The insertion sites of Li were estimated using the bond valence sum (BVS), 22 which is one of the indices used to evaluate the stability of atoms (ions) in a crystal structure. Because the valence of the Li + ion in the oxide was assumed to be +1, we inserted Li + ions at sites with a BVS close to 1.

Results and discussion
Observation of potential drop during Li plating Li plating/stripping was studied in each of the investigated electrolytes. In general, the SEI forms at the initial stage of Li plating, and it follows the electrodeposition of Li while a constant voltage is maintained. Electrical shorts due to Li dendrite growth were detected at the end of Li plating. The rst Li plating test in this study was conducted in the superconcentrated electrolyte using an electrode fabricated with only Cu powder (i.e., without Y 2 O 3 ). Its potential-capacity curve during the Li plating process is presented in the ESI (Fig. S6 †). The horizontal axis unit in Fig. S6 † is electrical capacity per weight of Cu powder. The initial stage between A and B reects SEI formation due to reductive decomposition of the PNMePh solvent. The region from B to C corresponds only to Li plating onto Cu powder and Cu foil. A large potential drop was unexpectedly observed when the capacity exceeded 25 mA h g À1 . The potential nally arrived at À5 V (point D). To determine the cause of the potential drop, we conducted EIS measurements corresponding to points A-D in the potential-capacity curve.
EIS is a powerful method for obtaining information about the electrochemical behavior at electrode-electrolyte interphases. The green line in Fig. 2a represents the Nyquist plot corresponding to point A (pristine) in Fig. S6. † One distorted semicircle is observed in the high-frequency range from 7 MHz to 667 Hz, and the linear portion of Z 0 and Z 00 appears in the lowfrequency range. An equivalent circuit composed of R e and two parallel units of resistance and capacity (R SEI |Q SEI , R ct |Q ct ) and Z w in Fig. 2a was used to analyze impedance spectrum A, where R e is the electrolyte resistance, R SEI and Q SEI are the resistance and constant phase element (capacity) 23,24 of the SEI on the metallic Li anode at high frequencies, respectively, R ct and Q ct are the charge-transfer resistance at the Cu electrode-electrolyte interface and the constant phase element for the double-layer capacitance at low frequencies, respectively, and Z w is the Warburg impedance. 25,26 The analysis results for the pristine sample were R e ¼ 0.103 kU, R SEI ¼ 1.596 kU (Q SEI ¼ 5.3 nF), and R ct ¼ 4.421 kU (Q ct ¼ 15.6 nF). The Nyquist plot changed dramatically as Li was plated. The Nyquist plots corresponding to points B and C are represented by yellow and red lines, respectively, in Fig. 2a. The linear portion of Z 0 and Z 00 for the pristine sample disappeared at frequencies less than 667 Hz, and the second semicircle appeared (yellow line). As Li plating proceeded further, the rst semicircle extended horizontally (red line). A third semicircle with a peak at 667 Hz was detected in the medium-frequency range between 3.12 kHz and 1.41 Hz. The sample in which the potential drop occurred exhibited a larger semi-arc at medium frequencies (blue). For the three samples aer Li plating (corresponding to points B, C, and D in Fig. S6 †), an equivalent circuit comprising three components (R SEI |Q SEI , R cu-in |Q cu-in , R ct |Q ct ) was used, where R cu-in and Q cu-in are the charge-transfer resistance and constant phase element in the Cu powder electrode in the medium-frequency range. 27 The potential drop was attributed to the SEI on electrodeposited Li in the Cu electrode. The resistances for the three components (R SEI , R cu-in , and R ct ) are summarized in Fig. 2b, and the constant phase elements for each component are shown in Fig. S7. † The increase in R cu-in with plating capacity y is characteristic. It was greater than 10 kU at point D. We therefore concluded that a large potential drop was associated with the increase in R cu-in . The EIS data for the cell fabricated using only Cu foil as an electrode are shown in the ESI (Fig. S8 †). Its EIS performance was similar to that of the Cu powder electrode.

Calculation of Li + insertion potential
The Li + insertion potential (V) was estimated from the total energy difference of the reaction Y 2 O 3 + 1.5Li + + e À / Li 1 Fig. S9. † The calculated Li + insertion potential was À1.02 V vs. Li + /Li.

Li plating test and XRD analysis
According to the concept presented in Fig. 1, as Li plating proceeds, the potential of the Cu electrode begins to drop. If Li + intercalation is assumed to occur, then the potential will soon become constant when it reaches À1.0 V. To verify this idea, Li plating tests were conducted with cells fabricated using Cu powder electrodes with and without Y 2 O 3 powder. The test results are summarized in Fig. 3a. In the case of the cell without Y 2 O 3 (green line), the potential of the Cu electrode showed a constant value of À0.6 V; it then decreased slowly when the capacity exceeded 20 mA h g À1 , reaching À5 V at a capacity of 28 mA h g À1 . By contrast, the cell with a Cu + Y 2 O 3 electrode exhibited two potential plateaus; the second plateau was   Fig. S10 and S11. † The Li + insertion may be occurring along with Li plating. To examine the Li + insertion process, we collected XRD patterns of the Cu + Y 2 O 3 electrodes aer Li plating to 36.1 mA h g À1 and 49.8 mA h g À1 (Fig. 3b). The two sharp XRD peaks at 2q ¼ 43. 34 and 50.22 are caused by the Cu powder and Cu foil. Five characteristic signals due to Y 2 O 3 cubic crystals are observed at 2q ¼ 29.06 , 33.58 , 39.68 , 48.32 , and 57.58 in the XRD pattern of the pristine sample. These signals are due to diffractions from the (222), (400), (332), (440), and (622) planes of cubic Y 2 O 3 crystals, respectively. 30,31 As the Li plating proceeded, the intensities of these ve peaks decreased. These results suggest that the Y 2 O 3 crystal size decreased or that the crystal structure was destroyed by the insertion of Li + ions. The Li plating tests clearly show that Li + ions attacked and interacted with Y 2 O 3 crystals at potentials below the standard Li plating and stripping potentials. To reconrm this phenomenon, another Li plating/stripping test was conducted in the super-concentrated electrolyte containing VC as an additive. The test with a discharge current of 0.05 mA for 24 h and a charge current of 0.05 mA to 2.0 V was repeated for ten cycles. A potential drop similar to that in Fig. 3a was observed in the discharge step of the seventh cycle (Fig. 4a). XRD patterns of the pristine Cu + Y 2 O 3 electrodes and those aer the h cycle and the eleventh cycle are summarized in Fig. 4b. The ve characteristic XRD peaks for Y 2 O 3 disappeared aer the tenth cycle. Fig. 4b also shows that the intensity of the XRD signals recovered aer Li stripping. Thus, Li + ions were conrmed to have interacted with Y 2 O 3 crystals at potentials less than the Li plating/stripping potentials.

XAFS analysis
The aforementioned XRD studies indicated that Li + ions might have interacted with Y 2 O 3 crystals during Li plating. If Li + -insertion into Y 2 O 3 occurred, the valence of Y or the atomic distance between and Y and O should have changed. XAFS is a powerful method for characterizing local structures in Y 2 O 3 crystals. We conducted XAFS measurements for a sample aer 55 mA h g À1 of Li plating. Fig. 5a shows the Y K-end X-ray absorption near edge structure (XANES) spectra of the sample and Y 2 O 3 crystal powder. The absorption end of Y 2 O 3 was 17 040 eV, 32 and the main absorption peaks were observed in the range from 17 050 to 17 070 eV. In the spectrum of the Cu + Y 2 O 3 electrode aer the Li plating, the absorption end rise was the same as that observed in the spectrum of the Y 2 O 3 powder. This result shows that similar XANES proles were obtained for the Cu + Y 2 O 3 electrode aer Li plating and for the Y 2 O 3 powder even though a slight difference was observed in their main peaks. Therefore, we concluded that the valence of the Y element in the Cu + Y 2 O 3 electrode was +3.
The extended X-ray absorption ne structure (EXAFS) spectrum is located more than 100 eV from the absorption end, and the Fourier transformation of the real spectrum provides information about the local structure, such as atomic distances. Y K-end FT-EXAFS spectra are displayed in Fig. 5b. The peak of the rst close connection, which was associated with Y-O, appeared in the range from 1.2 to 2.0 A in the spectrum of the Y 2 O 3 powder. The peak of the second close connection, which was associated with Y-O-Y, was located between 2.7 and 3.4 A. The EXAFS spectral shape of the Cu + Y 2 O 3 electrode exhibited the same characteristics as that of the Y 2 O 3 powder, and the rst and second close connections were detected in the same position. Therefore, the coordination environment around Y in the Cu + Y 2 O 3 electrode was similar that around Y in the Y 2 O 3 powder. In summary, XANES and EXAFS studies indicated that Li + ions were not inserted into the Y 2 O 3 crystallites during Li plating.

FE-SEM images and EDX analysis
The Li electrodeposition and Y 2 O 3 powder in the Cu + Y 2 O 3 electrode aer the Li plating test to 50 mA h g À1 was visualized by FE-SEM. Fig. S12a † shows an FE-SEM image of the cross section of the Cu + Y 2 O 3 electrode. The microscope magnication was 1000Â. Many needle-like crystals, which may be metallic Li, were observed near the Cu foil (yellow dotted ellipses). An enlarged FE-SEM image of the needle-like crystal is shown in Fig. S12b. † FE-SEM images for the Cu + Y 2 O 3 electrode at 1000Â magnication are displayed in Fig. 6a and b to highlight the Y 2 O 3 . The Y 2 O 3 powders were aggregated, forming a crosslinking mesh with a large hole before the plating. By contrast, the mesh structure aer Li plating was broken, indicating that the Y 2 O 3 powder exerted some inuence during the Li plating.
The SEI lms on the Y 2 O 3 powder, i.e., Li compounds, were further characterized by EDX analysis. We focused on some domains of the Y 2 O 3 powder with a square of less than 1 mm on one side and recorded the elemental distributions for Y, O, C, F, and Cu. The EDX measurement positions on the sample aer Li plating are displayed in Fig. S13a, † and the EDX results are summarized in Fig. S13b. † The EDX results for the sample before Li plating are presented in Fig. S14. † Fig. 7 shows the Y/O ratio for the samples before and aer Li plating. The ratio of the sample before the plating varied between 0.8 and 1.8. By contrast, the Y 2 O 3 domains aer the plating were divided into two ranges of 1.6 # Y/O # 1.7 and 2.1 # Y/O # 3.0. The latter ratio reects the domains aer the interaction between Li + and Y 2 O 3 . We cannot currently explain why the ratio increases. 7 Li-NMR measurement During Li plating tests, many needle-like crystals grew on the Cu powder and Cu foil. These crystals may be metallic Li. The metallic Li reacts with the electrolyte, resulting in the formation of SEIs at the surface of the Cu components. In the same manner, Li + ions, which interact with Y 2 O 3 , can induce reductive decomposition of the electrolyte. To characterize the deactivated Li, we conducted 7 Li-NMR measurements. NMR spectrum A in Fig. 8a corresponds to the sample of Cu + Y 2 O 3 electrode involving Cu foil and LiFSA/PNMePh electrolyte aer the Li plating test. A strong peak is observed at 260 ppm, which, according to Grey et al., 33 is attributable to metallic Li. The broad signal at approximately 0 ppm provided information about the Li + ions in LiFSA and Li compounds resulting from the reductive decomposition of electrolyte by metallic Li. 34-36 NMR spectra B and C correspond to the Cu + Y 2 O 3 electrode materials stripped from Cu foil in measurements conducted without and with rotation of the NMR sample tube. Aer the Cu + Y 2 O 3 electrode was stripped from the Cu foil, the intensity of the peak due to metallic Li decreased sharply, indicating that the metallic Li existed in the vicinity of Cu foil. Because spectrum C is high-resolution NMR data, we carried out waveform deconvolution for the broad NMR peak at $0 ppm. As shown in Fig. 8b, the NMR waveform was deconvoluted into three waves with peaks at À0.63 ppm, À1.32 ppm, or 3.25 ppm. The percentage of each peak was 32%, 14%, and 54%, respectively. The wave with the peak at À1.32 ppm is attributed to Li + ions in the LiFSA/PNMePh electrolyte because it was observed in the spectrum of the pristine sample before the Li plating test (Fig. S15 †). The other two waves reected the reductive decomposition. Letellier et al. studied the SEI in LiPF 6 /(EC + DMC) electrolyte using 7 Li-NMR and identied Li 2 CO 3 , LiF, Li 2 O, LiOH, and ROCO 2 Li as SEI components. 37 Because the 7 Li-NMR signal of LiF appears between 0 and À1 ppm, the wave with the peak of À0.63 ppm in this study may be due to an analogue of LiF. The 7 Li-NMR signal of Li 2 CO 3 was detected at 3.89 ppm; therefore, the Li compound with a peak at 3.25 ppm is likely an oxide.

Conclusions
Research into the electrochemical behavior of certain carriers at potentials less than the standard Li + /Li plating level is Fig. 6 High-resolution FE-SEM images of the cross section of Cu + Y 2 O 3 electrode (a) before and (b) after Li plating. The magnification is 10 000Â, and the scale bar is 1 mm. Fig. 7 Comparison of the Y-to-O ratio determined by EDX elemental analysis of the Cu + Y 2 O 3 electrode before and after Li plating. Fig. 8 (a) 7 Li-NMR spectra for Cu + Y 2 O 3 electrodes with Cu foil (spectrum A), without Cu foil and without rotating (spectrum B), and without Cu foil and with rotating (spectrum C); (b) waveform separation results for spectrum C in the region near 0 ppm. a challenging and unknown technical eld and will lead to immensely useful anode materials for batteries. Li + insertion into cubic Y 2 O 3 crystals is theoretically not possible because the insertion potential is À1.02 V. However, we observed a drastic potential drop of a Cu electrode during Li plating onto Cu in a super-concentrated electrolyte of LiFSA and PNMePh. This unique behavior suggested the possibility of Li + attacking Y 2 O 3 at a potential of À1.02 V. In this study, Li plating tests on Cu were conducted in the presence of Y 2 O 3 cubic powder. The Li plating test results and some spectroscopic analyses, including XRD and XAFS, indicate an interaction between Li + and Y 2 O 3 that leads to a decrease in crystallinity. This unexpected discovery stemmed from our investigation of the decrease in electrode potential in the super-concentrated electrolyte, and we expect similar phenomena to be observed in other highly concentrated electrolyte systems. We speculate that this approach will provide a path for the design of new active materials.

Conflicts of interest
The authors declare no competing nancial interest.