Properties of trifluoromethylated lithium borates for lithium-ion battery electrolytes

The structures of the synthesized lithium borates were confirmed by high-resolution mass spectrometry and ¹¹ B, ¹⁹F, ¹³C, and ⁷Li nuclear magnetic resonance (NMR) spectroscopy (table 1). The experimentally obtained mass (M_exp) agreed with the calculated mass (M_cal). The mass difference (|M_exp− M_cal|/M_exp × 10⁶) for each lithium borate (PFP-F₂, PFP-Ox, HHIB₂, HHIB-F₂, and HHIB-Ox) was 1.39, 4.83, 3.74, 2.95, and 1.89 ppm, respectively, confirming the successful synthesis of the lithium borates. The ¹¹B, ¹⁹F, ¹³C, and ⁷Li NMR spectra further confirmed the structure and high purity of the synthesized lithium borates (figures S2–S5). The ¹H NMR showed no signals from impurities or by-products, except for solvated ethylmethylcarbonate (EMC) and the residual protonated acetonitrile-d3 methyl group (CD₃CN; figure S1). The ¹³C NMR spectra showed characteristic peaks from C=O (140–147 ppm), CF₃ (112–113 ppm), and (CF₃)₂ C (79–84 ppm), as well as peaks from the solvated EMC (figure S2). The (CF₃)₂ C peaks for the HHIB group (HHIB₂, HHIB-F₂, and HHIB-Ox) and PFP group (PFP₂, PFP-F₂, and PFP-Ox) were observed at ca. 79 and 84 ppm, respectively. The (CF₃)₂ C peak chemical shift was affected by the C=O within the structure. The CF₃ peak chemical shifts corresponded well with the number of electron-withdrawing CF₃ groups attached to the beta carbon: PFP-group (PFP₂, PFP-F₂, and PFP-Ox) > HHIB-group (HHIB₂, HHIB-F_2, and HHIB-Ox). The C=O peaks from the solvated EMC were detected for most of the lithium borates, except for HHIB-F₂, Ox_2, and Ox-F₂. The C=O (ca. 140 ppm) chemical shifts for the solvated EMC were slightly shifted, while the CH₃, CH₃CH₂, and CH₃ CH₂ chemical shifts for the solvated EMC were similar to those observed in the EMC solution. The C=O peak shift suggests a strong interaction between the EMC and Li⁺ via the carbonyl moiety, making it difficult to remove EMC from lithium borates. An additional C=O feature was detected for PFP-Ox, HHIB₂, HHIB-F₂, HHIB-Ox, Ox₂, and Ox-F₂ at 141–147 ppm. The C=O peaks from the HHIB group can be seen at the lowest magnetic field (ca. 146 ppm), indicating that they are strongly affected by the alpha carbon CF₃ group.

Table 1. Structure-related parameters obtained from NMR and high-resolution mass spectrometry.

Lithium borate		19F NMR [ppm]	11B NMR [ppm]	13C NMR [ppm]	7Li NMR [ppm]	Mexp Mcal| |Mexp− Mcal|/Mexp × 106
Structure	Abbreviation
	PFP2 (LiBPFPB)	δ 94.1	δ 12.7	δ 112.5 δ 83.5	δ 0.3	— — —
	PFP-F2	δ 94.1 δ 16.6	δ 6.4	δ 112.7 δ 83.9	δ 0.2	380.9768 380.9773 1.39
	PFP-Ox	δ 94.2	δ 10.7	δ 142.7 δ 112.2 δ 84.0	δ 0.2	430.9581 430.9602 4.83
	HHIB2	δ 88.2	δ 11.4	δ 146.3 δ 111.8 δ 79.1	δ 0.3	430.9586 430.9602 3.74
	HHIB-F2	δ 88.1 δ 15.4	δ 5.7	δ 147.3 δ 112.1 δ 79.1	δ 0.2	258.9811 258.9818 2.95
	HHIB-Ox	δ 88.2	δ 10.1	δ 146.1 δ 142.3 δ 111.6 δ 79.1	δ 0.2	308.9641 308.9647 1.89
	Ox2 (LiBOB)	—	δ 8.8	δ 141.8	δ 0.2	— — —
	Ox-F2 (LiDFOB)	δ 10.3	δ 4.4	δ 143.0	δ 0.1	— — —
	LiBF4	δ 10.3	δ 0.0	—	δ 0.0	— — —

NMR analysis suggests that the effect of the electron density reduction (shift to a lower magnetic field) caused by the two CF₃ groups is stronger than that of the carbonyl group. The ¹⁹F NMR shows the CF₃ group within the PFP-group and HHIB-group at 94.1 and 88.2 ppm, respectively. The PFP CF₃ group signal is located at a lower field than that in HHIB, likely because C(CF₃)₂, which has greater electron-withdrawing properties than carbonyl, is attached to the β-position of the CF₃ group. The result indicates that the CF₃ group peak position varies significantly depending on the chelate ring substituent. A similar trend was observed for the F atom peak that is directly attached to the boron center. The F peak position was significantly affected by the opposite substituent (PFP, HHIB, oxalate, or F), specifically 16 ppm for PFP-F₂ and HHIB-F₂ and 10 ppm for Ox-F₂ and LiBF₄ (figure S3). However, the peaks arising from the CF₃ groups within PFP₂, PFP-F₂, and PFP-Ox were similar, indicating negligible induction effects of PFP, F, and Ox bound to boron on the opposite CF₃ group in the PFP chelate ring. The same trend was observed for the CF₃ groups within the borate-containing HHIB group (HHIB₂, HHIB-F_2, and HHIB-Ox). In the ¹¹B NMR spectra, the boron peaks were downshifted with an increasing number of CF₃ groups, indicating that the electron-withdrawing CF₃ group delocalized the negative charge on the boron center (figures S4(a) and (b)). In contrast, the ⁷Li NMR suggested almost no Li peak position dependence on the structure of the lithium borates or electron density of the boron center, probably because of the dominant effect of the solvated deuterated acetonitrile around Li⁺ (figure S5).

2.2.1. Thermal stability

2.2.2. Hydrolytic stability

2.2.3. Electrochemical stability

The electrochemical stability of each lithium borate was evaluated via linear sweep voltammetry (LSV) with a glassy carbon electrode (GCE) in an EMC- ethylene carbonate (EC) 2:1 (vol/vol) solution. The electrochemical windows of the electrolytes containing PFP-F₂, PFP-Ox, HHIB₂, HHIB-F₂, and HHIB-Ox were comparable to or even wider than those of Ox₂, Ox-F₂, and BF₄-containing electrolytes. The result confirmed the applicability of the newly prepared lithium borates as electrolytes for lithium-ion batteries (figure 1).

Zoom In Zoom Out Reset image size

A large reductive current was observed below 0.5 V_Li, which corresponds to the EC reduction (figure 1(a)). During the initial potential sweep, a small reductive current was observed at 0.5–2.0 V_Li for most of the salts except for LiBF₄. The onset potential of the reduction peak is in the following order; HHIB₂ < PFP-F₂ < HHIB-F₂ ∼ PFP₂ < Ox-F₂ < Ox₂ ∼ PFP-Ox < HHIB-Ox. To further clarify the origin of the reduction peak, we performed LSV in borate-containing 1,2-dimethoxyethane (DME) solvent, with higher reduction stability compared to EC (figures S8(a) and (b)). The LSV showed a similar small reduction peak, and the onset potential trend was independent of the solvent. In order to confirm the origin of this small reductive peak, LSV was performed in an EMC-EC 2:1 (vol/vol) solution containing potential impurities, such as PFP, HHIB, and Ox (figure S8(c)). Although a reduction current appeared at ca. 1.3V_Li for PFP, Ox, and HHIB-containing EMC-EC electrolytes, no reductive peak corresponding to those in figure 1(a) was observed, suggesting a negligible impurity contribution to the reductive peaks. Therefore, we conclude that the reductive peak observed in figure 1(a) was originated from partial salt decomposition. Although the strong electron-withdrawing nature of the CF₃ group generally decreases the LUMO level [43], the presence of the CF₃ group did not decrease the reduction stability of PFP₂, PFP-F₂, HHIB_2, or HHIB-F₂. Furthermore, borates containing Ox ligands, including Ox₂ and OxF₂, showed relatively low reduction stability (high onset potential) among the salt tested, which is in agreement with the previous studies suggesting Ox-contained borates easily decompose via reduction reactions and generate solid electrolyte interphase (SEI) on the anode surface [15, 20, 44–46]. The energy level calculations of highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) for each borate (figure S9) suggest that the LUMO level is in the order of PFP₂ > PFP-F₂ ≫ PFP-Ox and HHIB₂ > HHIB-F₂ ≫ HHIB-Ox, which gave the following general trend in the LUMO level; borates with CF₃ group > B–F bond > Ox. Although the presence of solvents significantly affects the redox potentials of the molecule [47], the observed LUMO-level trend is partially reflected in the LSV results, where borates with Ox ligands have relatively low reduction stability.

We here propose that the partial decomposition of the borates form an SEI layer on the GCE surface, inhibiting the further decomposition of the borates and solvent. Our hypothesis was supported by the absence of reductive current at 0.5–2.0 V_Li during the second potential sweep in both EC/EMC and DME system, indicating the inhibition of the further borate reduction by the SEI formed during the initial cycle (figures 1(b) and S8(b)). Furthermore, the onset potential of the EC and DME reduction reaction shifted slightly lower in the second cycle compared to that of the initial cycle, suggesting that the SEI can also prevent EC and DME decomposition.

Oxidation stability was evaluated for all prepared lithium borates, confirming that all borates were electrochemically stable up to ca. 4.8 V_Li. This range covered the usual operating voltage of the lithium-ion battery (up to ca. 4.2 V_Li [4, 5, 7]; figures 1(c) and (d)). The oxidation stability of bis- and mono-chelated lithium borates depends on the ligands; the oxidation stability followed the order: oxalate, HHIB, and PFP. The strong electron-withdrawing nature of the CF₃ group lowered the HOMO level and improved oxidation resistance [25]. With the exception of Ox-F₂, the oxidation current decreased in the second cycle, suggesting that oxidative decomposition products deactivate the GCE. Despite the presence of easily oxidized oxalate groups, PFP-Ox and HHIB-Ox showed excellent oxidation resistance. Thus, the observed trend indicates that CF₃-and/or oxalate-containing oxidized products can effectively prevent further decomposition of the borates.

The ionic conductivity for novel lithium borate electrolyte can be tuned by introducing an electron-withdrawing CF₃ group to the anion structure (improved Li⁺ dissociation) and by reducing the molecular weight and anion size (improved Li⁺ diffusion). HHIB-F₂, containing 2 CF₃ groups together with small and light F atoms, optimally balanced the above-mentioned design requirements, resulting in the highest ionic conductivity among the newly prepared lithium borates (table 3).

The ionic conductivities of 0.8 M lithium borate in EC-EMC (1:2 vol/vol) solution at 30 °C are listed in table 3. All newly synthesized lithium borates showed better ionic conductivity than conventional LiBF₄. The conductivity followed the order: HHIB-F₂ > Ox-F₂ > HHIB-Ox > Ox₂ > PFP-F₂ > HHIB₂ > PFP-Ox > PFP₂, suggesting that the anion size and electron-withdrawing nature of the ligand played significant roles in determining ionic conductivity.

Lithium borates with large PFP ligands (PFP₂, PFP-Ox, and PFP-F₂) and HHIB₂, which contains two relatively large HHIB ligands, showed low ionic conductivity. However, the introduction of an electron-withdrawing group (CF₃ or F) improved the ionic conductivity (replacing the carbonyl group with the CF₃ group: HHIB-Ox > Ox₂, HHIB-F₂ > Ox-F₂, replacing the Ox group with F: HHIB-F₂ > HHIB-Ox, Ox-F₂ > Ox₂). The ionic conductivity can be determined by the dissociation degree of the salt and/or the mobility of anions and cations [6, 48, 49]. To clarify the effect of the dissociation degree on the ionic conductivity, ¹¹B NMR and Raman spectroscopy were performed. Although the degree of Li dissociation was dictated by the boron atom electron density, no clear relationship between Li dissociation and ionic conductivity was confirmed (figure 2).

Zoom In Zoom Out Reset image size

The Li dissociation degree (α) was estimated from the Raman intensity ratio of the EC-related peaks (I_bound), considering the difference in the EC molar number (table S2). The Raman spectra of 0.8 M lithium borate in EC-EMC (1:2) solution at 850–960 cm⁻¹ are shown in figure 2(a). The peak at approximately 880–900 cm⁻¹ was assigned to C–O symmetric stretching of the non-solvated EC (denoted as free-EC) [50–52]. The small peak at approximately 905 cm⁻¹ corresponds to the C–O symmetric stretching vibration from EC interacting with the Li cation (denoted as Li⁺-EC). The Li dissociation degree (α) was estimated using the following method: the peak intensities of free-EC and Li⁺-EC (denoted as I_EC and I_Li+-EC, respectively) were calculated by integrating the deconvoluted free-EC and Li⁺ -EC peak areas [50, 52]. From the ratio of I_EC to I_Li+-EC and the amount of EC used to prepare the 0.8 M solution, the molar number of EC interacting with Li⁺ was calculated. A solvation number of 4 was assumed for Li⁺ [52–54] (EC to Li⁺ is 3, and EMC to Li⁺ is 1), allowing for the determination of the molar number of Li⁺ solvated by EC. Considering that the solvated Li⁺ is dissociated, it can be divided by the total molar number of Li⁺ in the solution to obtain an estimated α. The lithium borates showed an estimated α of 0.65–0.76, slightly higher than that of LiBF₄. The Li⁺ dissociation degree was in accordance with the number of electron-withdrawing CF₃ groups within the molecule: PFP₂ > HHIB₂ > PFP-Ox > HHIB-Ox > Ox₂ for bis-chelated lithium borates and PFP-F₂ > HHIB-F₂ > Ox-F₂ for the mono-chelated lithium borates.

Herein, we propose that the electron density of the boron atom varies with the number of electron-withdrawing groups, which also dictates the Li dissociation degree. This is supported by the chemical shift analysis of the ¹¹B NMR spectra, which can be used to estimate the boron atom electron density (figure 2(b)). For the bis-chelated lithium borates (PFP₂, HHIB₂, PFP-Ox, HHIB-Ox, and Ox₂), the chemical shifts of borate were large towards a low magnetic field (PFP₂ > HHIB₂ > PFP-Ox > HHIB-Ox > Ox₂) in accordance with the number of CF₃ groups, indicating lower boron atom electron density in the same order. The same trend was observed for the mono-chelated lithium borates (PFP-F₂ > HHIB-F₂ > Ox-F₂). The difference in the boron electron density for the bis- and mono-chelated lithium borates (bis-chelated < mono-chelated) can be explained by considering the resonance effect, which delocalizes the negative charge depending on the ligand structure. Furthermore, the cyclic structures of the bis- and mono-chelated borates benefited from the resonance effect compared to the simple BF₄ ligand.

A good correlation between the ¹¹B NMR chemical shifts and α values was confirmed in both the mono- and bis-chelated borates (figure 2(c)). This suggests that the boron atom electron density, which was affected by the number of electron-withdrawing groups and resonance effect from the ligand structure, dictated the degree of Li dissociation. Thus, the results support our hypothesis, and the ¹¹B NMR chemical shifts can be used as a descriptor of the Li dissociation degree. However, the Li dissociation degree did not show a clear correlation with ionic conductivity (figure 2(d)), suggesting that the ionic conductivity was influenced by different parameters such as ion mobility.

The relationship between ionic conductivity and Li-ion mobility (self-diffusion coefficient of Li⁺, D_Li) was studied by calculating the self-diffusion coefficient (D) of lithium cations using the ⁷Li signals from PFG NMR [55]. The ionic conductivity was tied to D_Li, and was significantly affected by the solution viscosity, which was determined by the anion molecular weight and size (figure 3).

Zoom In Zoom Out Reset image size

To estimate the mobility of the ionic species and solvents, the self-diffusion coefficient (D) of borate anions, solvents, and lithium cations were calculated from ¹⁹F (except for non-fluorinated BOB, for which ¹¹B was used), ¹H, and ⁷Li signals of PFG NMR, respectively (table S3). The validity of the obtained self-diffusion coefficient (D) was confirmed by the inverse proportionality of the Li⁺ self-diffusion coefficient (D_Li) to the viscosity (figure 3(a)), in agreement with the Stokes–Einstein equation expressed as follows:

$$\begin{align*} D = (KT)/(6 \pi \eta r_s) \end{align*}$$

where κ is the Boltzmann constant, η is the viscosity, and r_s is the Stokes radius of the diffused species. The observed linear relationship between D_Li and the self-diffusion coefficient of the solvents (D_EC, D_EMC) provides further support (figure 3(b)) and is consistent with the literature [56].

Figure 3(c) reveals that the ionic conductivity correlates well with the Li⁺ self-diffusion coefficients (D _Li ), suggesting that D _Li can be used as a descriptor for ionic conductivity. The large deviation in LiBF₄ likely originates from the significantly lower degree of Li⁺ dissociation. Based on the above equation, D _Li can be tuned by the electrolyte viscosity and Stokes radius of Li⁺. The anion viscosity and molecular weight exhibited a linear relationship (figure 3(d)), suggesting that the molecular weight determines the solution viscosity. For example, BF₄, which has the lowest molecular weight, showed the lowest viscosity, whereas the highest molecular weight, PFP₂, showed the highest viscosity. Here, the difference in the complex structure (bis- or mono-chelated) had a negligible effect on the correlation between viscosity and molecular weight. It was also found that reducing the anion size (PFP₂ vs. HHIB₂, HHIB-F₂) decreased the viscosity and effectively improved ionic conductivity.

Based on the above-described results, HHIB₂ and HHIB-F₂ were chosen as electrolytes and their charge–discharge performances were evaluated using a LiNi_0.33Mn_0.33Co_0.33O₂ cathode and artificial graphite anode. After stabilizing the cells by pre-charge/discharge at a 0.2 C rate at 25 °C (figure S10), we conducted cycle tests at a 3 C rate at 60 °C to confirm the performance of each electrolyte under severe conditions. The newly synthesized lithium borates showed notably larger initial capacity and better cycling stability compared to conventional LiBF₄, suggesting their applicability as electrolytes in lithium-ion batteries (figure 4).

Zoom In Zoom Out Reset image size

Figures 4(a)–(c) shows the charge–discharge profiles of the 1st, 50th, and 100th cycles at 60 °C. It should be noted that the charge–discharge profile was obtained after SEI formation via pre-cycling treatment (see figures S10(a)–(c)). The capacities of the 1st cycle for cells prepared with HHIB₂ (155.8 mAh g⁻¹) and HHIB-F₂ (154.8 mAh g⁻¹) were notably larger than the cell with the LiBF₄ electrolyte (138.0 mAh g⁻¹). The difference in the 1st cycle capacity is largely due to undesirable side reactions (solvent and electrolyte decomposition), which consume Li⁺ and decrease capacity. Therefore, the notably lower 1st cycle capacity of LiBF₄ compared to HHIB₂ and HHIB-F₂ is due to the progressive loss of Li⁺ through side reactions, in agreement with the understanding that LiBF₄ cannot form a good SEI to inhibit side reactions [57–59].

Further support comes from the cycling stability test, where the capacity fades from the 1st to 100th cycle was as follows: HHIB₂ (19.6%) < HHIB-F₂ (23.7%) < LiBF₄ (28.6%). Relatively high cycling stability was observed for HHIB₂ due to the high anionic thermal stability. Although HHIB-F₂ and LiBF₄ have similar thermal stability, there is a large difference in cycle capacity. Figure 4(d) shows that LiBF₄ has low coulombic efficiency up to 30 cycles, suggesting the side reaction is not sufficiently suppressed. In contrast, HHIB-F₂ showed a coulombic efficiency of >99% after the 10th cycle, confirming the formation of the SEI by HHIB-F_2, which can effectively suppress side reactions. The effect of SEI from borates was further confirmed by the AC impedance spectroscopy using coin cells (figure S10(d)). HHIB₂ shows the largest semi-circle, indicating high resistance to SEI due to reductive decomposition products. Conversely, HHIB-F₂ exhibited the smallest semi-circle, indicating the SEI component that increases resistance has been reduced and/or the Li⁺ conductivity of SEI has been improved by HHIB-F₂. To confirm the cause of the difference in resistance, x-ray photoelectron spectroscopy (XPS) was performed to analyze the surface of the negative electrode after the pre-charge/discharge (figure S11). HHIB-F₂-derived SEI contains a trace amount of CF₃ and B-F, while HHIB₂-derived SEI mainly consists of Li₂O. Although the details are still not clear, this difference in the composition may be responsible for the resistance of SEI, which may affect the battery performance.

Both mono- (BF₂ type; PFP-F₂, HHIB-F₂, Ox-F₂) and bis-chelated lithium borates (PFP₂, HHIB₂) were synthesized using LiBF₄ (KISHIDA CHEMICAL, battery grade) and corresponding bidentate ligands (perfluoropinacol, 2-hydroxy-3,3,3,3ʹ,3ʹ,3ʹ-hexafluoroisobutiric acid, oxalic acid) as starting materials. Perfluoropinacol (>98% purity) was purchased from TOKYO CHEMICAL INDUSTRY and oxalic acid was obtained by drying oxalic acid dihydrate (>99%), purchased from FUJIFILM Wako Pure Chemical, at 90 °C under reduced pressure. In addition, 2-hydoxy-3,3,3,3ʹ,3ʹ,3ʹ-hexafluoroisobutiric acid was synthesized according to a literature procedure [60]. Anhydrous alkyl carbonates such as EMC or DMC (KISHIDA CHEMICAL, battery grade) were selected as reaction solvents. Trimethylchlorosilane (Me₃SiCl, TOKYO CHEMICAL INDUSTRY, >98%) was used as a defluorination reagent to minimize residual chlorine. The molar ratio between Me₃SiCl and the ligand dictates the final product structure; bis-chelated lithium borates were mainly obtained using four equivalents of Me₃SiCl and two equivalents of ligand, while mono-chelated lithium borates were obtained using two equivalents of Me₃SiCl and one equivalent of ligand. Asymmetric bis-chelated lithium borates (PFP-Ox, HHIB-Ox) were synthesized by treating Me₃SiCl with Ox-F₂ and the corresponding ligand (perfluoropinacol, 2-hydoxy-3,3,3,3ʹ,3ʹ,3ʹ-hexafluoroisobutiric acid). The general synthetic method for lithium borate is shown in Scheme 1. The individual synthesis steps are shown in the SI, and all manipulations were performed under nitrogen atmosphere with a dew point of ⩽−50 °C. The lithium borate Ox₂ (LiBOB) was not synthesized and was purchased from Albemarle with a purity of ⩾99%.

Zoom In Zoom Out Reset image size

The ¹H, ⁷Li, ¹¹B, ¹³C, and ¹⁹F NMR spectra used for characterization were recorded using a JEOL JNM-ECZ400S spectrometer in deuterated acetonitrile (CD₃CN) at ambient temperature. HRMS were recorded using an LTQ Orbitrap Discovery (Thermo Fisher Scientific) mass spectrometer in ESI mode. TG and DTA were performed with equipment from Rigaku (Thermo plus TG8120) using 10 mg samples at a heating rate of 10 °C min⁻¹ under a nitrogen atmosphere. LSV data were collected at a scan rate of 5 mV s⁻¹ with an electrochemical analyzer (ALS-604E, BAS). Glassy carbon was used as the working electrode and lithium metal was used as both the counter and reference electrodes. The ionic conductivity was measured using a conductivity meter with a submersible-type cell (HORIBA, ES-51/3552-10D) at 30 °C. The cell constant was calibrated before use using a potassium chloride standard solution (TAKEMURADENKISEISAKUSHO, 1,413 μS cm⁻¹). Raman spectra were collected using equipment from RENISHAW (inVia Raman microscope WiRE5) equipped with an argon laser emitting a 514.5 nm line. The samples were placed in quartz cells and self-diffusivity data were measured using the pulsed field gradient spin-echo diffusion method [48, 61, 62]. PGSE NMR measurements were performed using a JEOL JNM-ECZ400S spectrometer. The diffusion coefficients of the anions, lithium, and solvents were measured at 30 °C using ¹⁹F, ⁷Li, and ¹H NMR, respectively. The interval between the gradient pulses (Δ) was 50 ms and the maximum magnetic field gradient (g) was 0.3 T m⁻¹. Tubes were purchased from SIGEMI (BMS-005J) and used to seal the samples. ¹¹B NMR was used to measure the lithium bis(oxalato)borate anions. However, the echo signal did not decay under these conditions and attempts to obtain the self-diffusivity value were unsuccessful. Therefore, only the bis(oxalato)borate anion was measured using a JEOL JNM-ECA400 spectrometer with Δ = 25 ms and maximum g = 12.0 T m⁻¹. A rheometer MCR102 (Anton Paar) was used for viscosity measurements, and calibration was performed before use using a standard calibration solution (Nippon Grease, JS2.5, and JS10). The HOMO and LUMO energy levels were determined using density functional theory with the functional B3LYP with the 6–31Gdp basis set in Gaussian 09 program. XPS (PHI 5000 VersaProbe II, ULVAC-PHI, Japan) measurements were carried out using Al Kα radiation (1486.6 eV) under ultrahigh vacuum. The samples were transferred from the Ar-filled glove-box to the XPS chamber using a transfer vessel to minimize exposure to air. A charge neutralizer was employed for all measurements, and the graphite peak at 284.3 eV was used as a reference for energy scale adjustment. The obtained spectra were analyzed by Multipack software (v. 9.6.0.15). Line syntheses of elemental spectra were conducted using Gaussian–Lorentzian (80:20) curve fitting with Shirley background subtraction.

Battery-grade EC was purchased from KISHIDA CHEMICAL. The electrolyte consisted of carbonate solutions EC/EMC (1/2 vol/vol) and electrolytes (PFP-F₂, HHIB-F₂, HHIB₂, and LiBF₄). The electrolyte concentration was 1.0 M, and the composite NMC111 cathode consisted of 90.2 wt% LiNi_0.33Mn_0.33Co_0.33O₂ active material (MX-6; Umicore), 3.8 wt% conductive carbon (HS-100; Denka), and 6.0 wt% polyvinylidene difluoride (PVDF) binder (L#7208; KUREHA). The composite graphite anode contained 90.0 wt% graphite (MAG-D; Hitachi Chemical) along with a 10.0 wt% PVDF binder (L#9130; KUREHA). A cellulose separator (TF40-30) was purchased from NIPPON KODOSHI. The active mass loading of the cathodes and anodes was 12.2 and 6.7 mg cm⁻², respectively. Graphite/NMC111 full 2032-type coin cells were built with a cathode (d = 10.0 mm), anode (d = 12.0 mm), separator (d = 16.0 mm), and 40 μl of electrolyte solution in each cell prepared in a glove box filled with argon (content of H₂O below 1 ppm) for electrochemical performance measurements. The assembled graphite/NMC111 cells were charged at a constant current 0.2 C to 4.3 V (CC–CV mode) at 25 °C. After resting for 1 h, the charged cells were discharged in CC mode (0.2 C) to 3.0 V at 25 °C. This process is called the pre-charge/discharge. The cycling tests were conducted under the following conditions. The cells were charged to 4.3 V in CC–CV mode (3C) at 60 °C, and after resting for 1 min at 60 °C, the charged cells were discharged to 3.0 V in CC mode (3 C) at 60 °C. The charge/discharge cycle was repeated 100 times. After the pre-charge/discharge (i.e. before cycling), the cells were probed by electrochemical impedance spectroscopy (EIS; ALS-660 C electrochemical analyzer, BAS, Japan) at 25 °C and at a state of charge of 100% (25 °C, 0.2 C charging) using an amplitude of 10 mV and a frequency range of 100 kHz to 10 mHz.