Investigation of Mixtures of BF₃ Carbonates and LiX (X = OCH₂CF₃, OC(H)(CF₃)₂, CO₂CF₃) as Novel Electrolyte Systems for Lithium Ion Batteries

All syntheses and measurements were done in an inert argon atmosphere using standard Schlenk techniques or glove boxes (Jacomex GPT4 or MBraun, H₂O and O₂ content ≤ 1 ppm). BF₃ (UHP 2.5, Linde GmbH) and carbonates (battery grade, BASF SE) were used for the adduct syntheses as well as for electrolyte preparation. BF₃·D (with D = dimethyl carbonate (DMC), ethylene carbonate (EC), ethyl methyl carbonate (EMC) and propylene carbonate (PC)) was synthesized by introducing an excess of gaseous BF₃ into the corresponding carbonate as recently described.⁵² BF₃·O(CH₂CH₃)₂ (>98%, Merck) was distilled under vacuum. The solvent free lithium alkoxides Li[OCH₂CF₃] and Li[OC(H)(CF₃)₂] were synthesized by reaction of the alcohols HOCH₂CF₃ (99+%, AlfaAesar) and HOC(H)(CF₃)₂ (99+%, AlfaAesar) with LiH (> 97%, Merck) in n-Hexane (99%, Sigma/Aldrich).⁵³ The commercial lithium salt Li[CO₂CF₃] (97%, AlfaAesar) was dried prior to use under dynamic vacuum (10^–3 mbar at 150 °C).

For concentration-dependent conductivity measurements, 1:1 stoichiometric amounts of LiX (with X = OCH₂CF₃, OC(H)(CF₃)₂, CO₂CF₃) and BF₃·D were weighed into a flask and dissolved in EC:EMC (3:7 wt%, battery grade, water content ≤ 50 ppm, L57, BASF SE) and stirred overnight. The concentration-dependent conductivity measurements were carried out using a Mettler Toledo InLab710 conductivity electrode and a Mettler Toledo S30 SevenEasy^TM conductometer under argon atmosphere inside a glove box (Jacomex GPT4, H₂O and O₂ content ≤ 1 ppm). NMR measurements of all mixtures were carried out before conductivity measurements with a Bruker Avance II + WB 400 MHz and Bruker Avance III HD 300 MHz spectrometer in 5 mm NMR tubes with the deuterated solvent CD₃CN. The corresponding spectra were analyzed and visualized by the software package Topspin 4.0.6. Further experimental details, procedures and NMR spectra of the mixtures are deposited in the supplementary information (SI) (available online at stacks.iop.org/JES/167/080507/mmedia).

DFT optimizations were carried out with TURBOMOLE^54,55 (Version 7.2) in the highest possible point group at 0 K using the BP86^56,57 functional with RI-J approximation^58,59 in combination with the def2-TZVPP^60–62 or def-TZVP⁶³ basis set in combination with D3(BJ) dispersion correction. The frequency analyses for the calculations of vibrational frequencies and zero point energies were executed with AOFORCE.^64,65 The absence of imaginary vibrational frequencies was checked. The reaction enthalpies and the Gibbs reaction energies were corrected by zero point energies and thermal corrections of the enthalpy and entropy at 298 K. Equation 9 (see Table III) below is the only non-isodesmic investigated reaction, which in addition produces the much smaller, and thus much better solvated ion [BF₄]^– together with very polar carbonate molecules. Only here, Δ_rG⁰ was corrected for solvation by the COSMO model⁶⁶ at (RI-)BP86(BJ)/def-TZVP level by using the published permittivity of L57⁶⁷ of 18.5.

The cyclic voltammograms were recorded with a Metrohm Autolab PGSTAT101 potentiostat inside a glove box (Jacomex GPT4, H₂O and O₂ content ≤ 1 ppm) with a three-electrode setup of the electrochemical cell. The working electrode consisted of platinum (Metrohm 6.1204.190 Pt, 1 mm diameter) or glassy carbon (Metrohm 6.1204.600 GC, 2 mm diameter). The reference and counter electrode consisted of lithium (lithium foil, BASF SE). All measurements were carried out with a sweep rate of 0.01 V·s^–1. The spectra were analyzed by the software package NOVA 2.0.

CR 2032 coin cells were prepared in an argon filled glove box (MBraun, H₂O and O₂ content ≤ 1 ppm). Positive electrode tapes consisting of Li[Ni_0.33Co_0.33Mn_0.33O₂] (93% active material with 3% conductive carbon Super C65 and 4% poly(vinylidenedifluoride) binder (PVDF), 14 mm diameter, NCM, BASF SE) and negative electrode tapes consisting of graphite (95.7% graphite with 0.5% conductive carbon Super C65 and 3.8% binder composed of sodium carboxymethyl cellulose (CMC) and styrene butadiene rubber (SBR), 15 mm diameter, BASF SE) were used. A three-layered system was used as separator to ensure the extraction of the electrode of the cells after cycling for ex situ surface analysis. The sequence of separators was: Celgard 2500 (19 mm diameter) / GF/D (19 mm diameter, Whatman)/ Celgard 2500 (19 mm diameter). LiX and BF₃ carbonates (1:1 stoichiometric ratio) were mixed and stirred overnight as electrolytes and had a lithium ion concentration of 1.0 M in L57 (3:7 wt%, EC:EMC, battery grade, water content ≤ 50 ppm, BASF SE).

An Arbin BT2000 battery cycler was used for constant current charge/discharge measurements in a potential range between 3.0 V to 4.2 V at 25 °C. The cycling protocols started with an open circuit voltage (OCV) measurement, followed by five formation steps (specific nominal capacity of 150 mAh·g⁻¹ per step. Formation: one step with a current rate (C-rate) of C/20, two steps with C/10, followed by two steps with C/5) and cycling for 50 steps with a constant charge/discharge C-rate of C/2. All cells were prepared in duplicate to confirm reproducibility of the cycling behavior. Representative cycling data are presented below.

After the 55 cycles, at 0% state of charge (SOC) electrochemical impedance spectroscopy (EIS) measurements were performed on a Bio-Logic potentiostat with a perturbation of 10 mV and in the frequency range of 300 kHz–10 mHz. EIS measurements were carried out at 25 °C and repeated three times (with an OCV step before each EIS measurement).

All cells were transferred and disassembled in an argon filled glove box (MBraun, H₂O and O₂ content ≤ 1 ppm) after EIS measurements. The positive and negative electrodes were rinsed three times with ethyl methyl carbonate (3 × 500 μL) to remove residual electrolyte from the electrode surface and dried overnight in the glove box prior to surface analysis. X-ray photoelectron spectroscopy (XPS) was acquired with a Thermos K-Alpha system using monochromatized Al K α radiation (hν = 1486.6 eV) under ultra-high vacuum (8.1·10⁻⁸ Pa). Therefore, samples were transferred into the XPS chamber with a vacuum transfer vessel from Thermo Scientific. The spectra obtained were analyzed by using the Thermos Advantage software (Version 5.988). The binding energy was corrected based on the C 1s orbital energy of hydrocarbon at 284.8 eV for all spectra.⁶⁸ The assignments of the binding energy of the components are from the NIST XPS database.⁶⁹

The pure lithium salts LiX (X = OCH₂CF₃, OC(H)(CF₃)₂, CO₂CF₃) display poor ionic conductivity (0.05 mS·cm^–1 to 0.46 mS·cm^–1 at 1.5 mol·L^–1) in mixed carbonate solvents like L57 (3:7 wt% EC:EMC, EC = ethylene carbonate; EMC = ethyl methyl carbonate). For comparison, LiBF₄ or LiPF₆ in L57 with a concentration of 1.0 mol·L^–1 have conductivities of 3.5 mS·cm^–1 or 7.5 mS·cm^–1. The reason for the low ionic conductivity of pure LiX presumably is the good accessibility of the negative charge of the anions X^–, which leads to ion-tuple formation/association with Li⁺ and to a lower mobility of the lithium ions.

Upon addition of commercially available BF₃OEt₂ in a 1:1 or 1:2 stoichiometric ratio to LiX as a pre-test, all measured conductivities increased significantly (Table I). The ionic conductivity is increasing from 0.05 mS·cm^–1 to 1.44 mS·cm^–1 for Li[OCH₂CF₃], 0.21 mS·cm^–1 to 3.22 mS·cm^–1 for Li[OC(H)(CF₃)₂] and 0.46 mS·cm^–1 to 3.08 mS·cm^–1 for Li[CO₂CF₃] for the stoichiometric ratio of 1:1. The doubling of LiX from 1 to 2 equivalents led only for Li[OC(H)(CF₃)₂] to an increase of the ionic conductivity (from 3.08 mS·cm^–1 for 1:1 to 4.69 mS·cm^–1 for 2:1), while the others display lower conductivities due to the maximum of solubility of LiX in L57.

Since several studies have shown that ethers are not suitable for applications in LIBs,⁷ BF₃ adducts containing organic carbonates, which are anyway present as solvent in LIBs, were synthesized according to our recently published protocol.⁵² The exemplarily tested composite electrolytes prepared from BF₃·EMC and LiX in L57 showed significantly higher ionic conductivities in both stoichiometric ratios (1:1 and 1:2) than the pure lithium salts. The highest ionic conductivity was observed for Li[CO₂CF₃] and BF₃·EMC (1:1 ratio) with 5.21 mS·cm^–1. In addition, and used in 1:1 ratio, the carbonate donor EMC induced considerably higher conductivities than the ether donor. Since the ionic conductivity differences between 1:1 and 1:2 mixes were not too drastic and for keeping the electrolytes economic, in the following measurements always a stoichiometric ratio of 1:1 was used for the combination of LiX with BF₃·D.

Starting from the concentration of 1.5 mol·L^–1, all composite electrolytes containing LiX (with X = OCH₂CF₃, OC(H)(CF₃)₂, CO₂CF₃) and BF₃D (with D = dimethyl carbonate (DMC), ethylene carbonate (EC), ethyl methyl carbonate (EMC), propylene carbonate (PC), diethyl ether (OEt₂)) in L57 as solvent were diluted stepwise until the final concentration of 0.1 mol·L^–1 was reached. Moreover, all starting mixtures (1.5 mol·L^–1) were investigated by NMR spectroscopy before measuring the conductivities. In the SI, tables including all measured conductivities (Tables SI–SIII) are listed and the respective NMR spectra are displayed (Figs. S1–S15).

The overall ionic conductivity is increasing as [OCH₂CF₃]⁻ < [OC(H)(CF₃)₂]⁻ < [CO₂CF₃]⁻, which follows the increasing strengths of the corresponding acids H−X as evaluated by their aqueous pK_a-values, i.e. 12.40 < 9.30 < 0.23.^73–75 For all electrolyte mixtures, the ionic conductivities displayed in Fig. 1 were lower, if the ether adduct BF₃·OEt₂ was used in conjunction with any LiX. The spread of the ionic conductivity values, if using composite electrolytes of BF₃·D (with D = carbonate) with LiX, were astonishingly rather widespread as a function of D. For the least favorable composite electrolytes of BF₃D (with D = carbonate) and Li[OCH₂CF₃], the conductivities were similar within 0.25 mS·cm^–1 (Fig. 1, left). The highest conductivities reached rather low 2.06 mS·cm^–1 to 2.31 mS·cm^–1 at a concentration of 1.5 mol·L^–1.

Zoom In Zoom Out Reset image size

By contrast, and changing to the superior Li[OC(H)(CF₃)₂] and Li[CO₂CF₃] salts, their composite electrolytes with BF₃·D showed a larger spread of the conductivities as a function of D that may reach about 1 mS·cm^–1. The highest electrical conductivities with Li[OC(H)(CF₃)₂] and BF₃·D were found at the concentration of 1.1 mol·L^–1 (3.43 mS·cm^–1 to 4.41 mS·cm^–1, Fig. 1, middle), while those with Li[CO₂CF₃] and BF₃D showed the highest conductivities between 1.3 mol·L^–1 and 1.5 mol·L^–1 and reached 5.03 mS·cm^–1 to 5.81 mS·cm^–1 (Fig. 1, right). For both salts, the mixtures with BF₃·EC, including the organic carbonate with the highest dielectric constant (ε_r = 90)⁷⁶ of all used carbonates, led to the highest conductivities. The combination of Li[CO₂CF₃] with BF₃·D (with D = carbonate) in L57 showed the highest conductivities of all the investigated mixtures. Starting with a concentration of 0.5 mol·L^–1, they are consistently higher than that of a 1.0 molL^–1 solution of LiBF₄ in L57 used as a reference.

To investigate the dismutation according to Eqs. 4–7, NMR measurements were done. The ¹¹B NMR spectra display three borates in all composite electrolytes (Table II). Of those, [BF₄]⁻ and the single substituted [BF₃X]^– are with 24%–65% and 8%–67% typically the main components. Only for Li[OC(H)(CF₃)₂], the anion [BF₂X₂]⁻ is with 22%–26% a dominant part of the mixture. For all LiX/BF₃·D systems, the product distribution is in equilibrium at room temperature and chemical exchange between different borate species in solution occurs according to ¹⁹F,¹⁹F NOESY NMR spectroscopy (Figs. S23–S31).⁷⁷ Only the mixtures with Li[OCH₂CF₃] furnished a larger portion of the neutral borane BX₃ that may have formed according to Eq. 8,

$$\begin{eqnarray}&&{{\left[{{\rm{BFX}}}_{3}\right]}^{-}}_{\left({\rm{solv}}.\right)}\rightleftharpoons {{\rm{BX}}}_{3\left({\rm{solv}}.\right)}+{{{\rm{F}}}^{-}}_{\left({\rm{solv}}.\right)}\end{eqnarray} \tag{ 8 }$$

or a related path. Since the neutrals are not participating in the lithium ion transfer and the fluoride F^– would react further or ion pair with Li⁺, Eq. 8 probably accounts for the low ionic conductivity of the respective composite electrolytes including Li[OCH₂CF₃] that are all lower than that of pure LiBF₄. The threefold substituted [BFX₃]⁻ was observed only for Li[OC(H)(CF₃)₂] at a low relative concentration of 3%–4%.

Although the ionic conductivity values using one type of LiX vary considerably as a function of D = carbonate, the typical compositions of the electrolytes are rather stable and just differ by a few percent. Exceptions present only D = OEt₂ and PC combined with Li[CO₂CF₃]. The Li[CO₂CF₃]/BF₃D combination (with D = carbonate), shows ionic conductivities that are significantly higher than that for LiBF₄ in L57. The main component in these mixtures is according to the NMR analysis the [BF₃X]^– anion, so that one may suggest that this component might contribute positively to the conductivity. However, the ionic conductivity of LiPF₆ in L57 is 7.5 mS·cm^–1 at 1.0 mol·L^–1 and still higher than that of the best composite electrolyte formed by Li[CO₂CF₃] and BF₃EC in L57 (5.81 mS·cm^–1).

To investigate the individual ionic conductivities of the lithium borates like Li[BF₃X], Li[BF₂X₂] or Li[BFX₃] (with X = OCH₂CF_3, OC(H)(CF₃)₂, CO₂CF₃) several reactions were carried out.^78–81 However, all attempts of purifying these lithium borates failed for all investigated LiX. These reactions are described in the SI (Chapter 5). This is in agreement with the chemical exchange observed for all electrolytes in the NMR spectra. Furthermore, the compositions of each electrolyte are stable over months at ambient temperature. It suggests that it might be impossible to prepare the pure lithium borates under all tested stoichiometric ratios and solvent combinations.

The ¹¹B NMR investigations revealed for lithium salts combined with BF₃D (with D = carbonate) the presence of several borates [BF_xX_4-x]^– (with X = OCH₂CF_3, OC(H)(CF₃)₂, CO₂CF₃). In order to examine the suitability of these anions as electrolytes in LIBs quantum chemical calculations at (RI-)BP86(BJ)/def2-TZVPP level were performed.

The calculated molecular orbital energies shown in Fig. 2 are related to the stability of the anions against reduction and oxidation. The lower the HOMO energy, the more stable are the anions against oxidation of the positive electrode. The HOMO energies are lowering according to [BF_x(OCH₂CF₃)_4-x]^– > [BF₄]^– > [BF_x(OC(H)(CF₃)₂)_4-x]^– > [BF_x(CO₂CF₃)_4-x]^–, while the energies of the LUMO are decreasing in the order of [BF₄]^– > [BF_x(OCH₂CF₃)_4-x]^– > [BF_x(OC(H)(CF₃)₂)_4-x]^– > [BF_x(CO₂CF₃)_4-x]^–. The HOMO-LUMO gap is decreasing in the order of [BF₄]^– > [BF_x(OC(H)(CF₃)₂)_4-x]^– > [BF_x(OCH₂CF₃)_4-x]^– > [BF_x(CO₂CF₃)_4-x]^–. These energies are indicating the stability of the anions against reduction and the least stable in this respect appear to be the borates [BF_x(CO₂CF₃)_4-x]^–.

Zoom In Zoom Out Reset image size

The influence of the electron withdrawing substituents X^– and their electrostatic potentials are shown in Fig. 2. The strengths of the electron withdrawing group decreases in the order of [CO₂CF₃]⁻ > [OC(H)(CF₃)₂]⁻ > [OCH₂CF₃]⁻ (cf. to the pK_a-values mentioned above). The delocalization of the negative charge apparently should lead to reduced interactions between cation and anion and to a better lithium ion mobility. This observation is with the exception of X = OCH₂CF₃ confirmed by the ionic conductivity measurements.

Quantum chemical investigations of the reaction enthalpies (Δ_rH⁰) and Gibbs reaction energies (Δ_rG⁰) of the dismutation reactions of [BF₃X]⁻, [BF₂X₂]⁻ and [BFX₃]⁻ according to Eqs. 4–7 were performed (Table III). The calculations of Δ_rH⁰ and Δ_rG⁰ were carried out at the (RI-)BP86(BJ)/def2-TZVPP level in the gas phase, include zero-point energy corrections and were corrected for thermal contributions to enthalpy and free energy at 298 K. The dismutation reactions are isodesmic, the same number and type of bonds were formed/broken and the total charge of the compounds is not changing during the reaction. Thus, we expect that they are therefore close to the expectation values to be observed in the electrolyte. The Gibbs reaction energies provide information, if the dismutation reaction can thermodynamically take place (Δ_rG⁰ < 0, exergonic) or not (Δ_rG⁰ > 0, endergonic).

Table III.  Quantum chemical investigations of Δ_rH⁰ and Δ_rG⁰ of the dismutation reactions calculated at level (RI-)BP86(BJ)/def2-TZVPP.^a) For details to the calculations with each carbonate see SI (Chapter 7).

			ΔrH0 [kJ·mol–1]	ΔrG0 [kJ·mol–1]
(4)	2 [BF3X]− → [BF2X2]– + [BF4]–
	with X =	OCH2CF3	5.1	12.5
		OC(H)(CF3)2	9.7	18.4
		CO2CF3	20.7	29.2
(5)	2 [BF2X2]− → [BFX3]− + [BF3X]–
	with X =	OCH2CF3	–11.0	–10.4
		OC(H)(CF3)2	2.3	8.2
		CO2CF3	7.7	5.5
(6)	2 [BFX3]− → [BF2X2]− + [BX4]−
	with X =	OCH2CF3	10.4	15.6
		OC(H)(CF3)2	–6.9	–8.4
		CO2CF3	8.7	12.9
(7)	2 [BF2X2]− → [BX4]− + [BF4]–
	with X =	OCH2CF3	–6.5	7.3
		OC(H)(CF3)2	7.4	26.4
		CO2CF3	44.9	53.1
(9)	BF3D + [BFX3]– → BX3 + [BF4]– + D		with D = carbonateb)
	with X =	OCH2CF3	43.2 to 51.5	3.3 to 16.8 (–76.2 to –52.7)b)
		OC(H)(CF3)2	66.0 to 74.3	15.8 to 29.3 (–73.9 to –50.0)b)
		CO2CF3	177.4 to 185.7	136.0 to 149.5 (45.7 to 69.6)b)

^a)All calculations approximate the situation in the gas phase, include zero-point energy corrections and were corrected for thermal contributions to enthalpy and free energy at 298 K. ^b)Equation 9 is the only non-isodesmic reaction which in addition produces the much smaller, and thus much better solvated ion [BF₄]^– together with very polar carbonate molecules. Here it appeared wise to include a solvation model to the calculation. Δ_rG⁰ corrected for solvation by the COSMO model at the (RI-)BP86(BJ)/def-TZVP level is included in parenthesis. The permittivity of L57 was recently measured as ε_r = 18.5 and used.⁶⁷

The calculated thermodynamics collected in Table III are in a good agreement with the borate species, which were observed in the NMR spectra. For Li[OCH₂CF₃]/BF₃D no [BF₂X₂]^– species in the ¹¹B NMR spectra were observed. In agreement with this, the dismutation reaction Eq. 5 of the [BF₂X₂]^– ion was calculated to be exergonic by Δ_rG⁰ = –10.4 kJmol^–1. For this composite electrolyte system primarily [BF₄]^– and BX₃ were detected. The neutral BX₃ may have formed from the [BFX₃]^– intermediate in a follow-up reaction according to Eq. 9. Yet, this is the only non-isodesmic reaction in Table III, which in addition produces the much smaller, and thus much better solvated ion [BF₄]^– together with very polar carbonate molecules. Thus, here it appeared wise to include a solvation model to the calculation. Therefore, Δ_rG⁰ was corrected for solvation by the COSMO model⁶⁶ at the (RI-)BP86(BJ)/def-TZVP level, included in parentheses with Table III. Δ_rG⁰ now turns exergonic for X = OCH₂CF₃ (–76.2 kJmol^–1 to –52.7 kJmol^–1) and OC(H)(CF₃)₂ (–73.9 kJmol^–1 to –50.0 kJmol^–1; see SI Chapter 7 for detailed information)—which is in good agreement to the observations by NMR spectroscopy for X = OCH₂CF₃. The formation of neutral B(OCH₂CF₃)₃ is favored by the lowest Lewis acidity of all BX₃ molecules, as evidenced by its low fluoride ion affinity (FIA)⁸² compared to the other BX₃ molecules (FIA = 369 of B(OCH₂CF₃)₃/398 of B(OC(H)(CF₃)₂)₃/505 of B(CO₂CF₃)₃). The increased FIA in comparison to B(OCH₂CF₃)₃ is the origin that no neutral BX₃ were observed for X = OC(H)(CF₃)₂.

Furthermore, in the Li[OC(H)(CF₃)₂]/BF₃D electrolyte system the minor component is [BFX₃]⁻ with 3%–4%, which is also in a good agreement with the calculations (Δ_rG⁰ = –8.4 kJmol^–1 for Eq. 6). Here the dismutation of [BF₃X]⁻ and [BF₂X₂]⁻ (dismutation to [BX₄]⁻ and [BF₄]⁻) display the most endergonic Gibbs energies. This is also the case for the composite electrolyte of Li[CO₂CF₃]/BF₃D, which is in a good agreement with the ¹¹B NMR spectra were primarily [BF₃X]⁻ and [BF₄]⁻ were detected.

To investigate the stability of the composite electrolytes, solutions of LiX with BF₃·DMC (ratio 1:1) were prepared (0.1 mol·L^–1 in 1:1 wt% EC:DMC, L30), and multi-sweep cyclic voltammetry (CV) experiments between 0 V-3 V (Figs. 3a–3c) and 3 V-5 V (Figs. 3d–3f) vs Li⁺/Li were performed.

Zoom In Zoom Out Reset image size

Both, Li[OCH₂CF₃] and Li[OC(H)(CF₃)₂] combined with BF₃DMC show no waves in the cathodic scans and the different scans hardly differ from each other (Figs. 3a and 3b). The current density increases for both composite electrolytes at potentials below 0.5 V (around 0 V: –98 μA·cm^–2 for Li[OCH₂CF₃] with BF₃DMC and –250 μA·cm^–2 for Li[OC(H)(CF₃)₂] with BF₃·DMC). In the first anodic scan of Li[OCH₂CF₃] and BF₃·DMC (Fig. 3d) two waves at around 3.1 V and 3.3 V occur with low current densities (up to 8.5 μA·cm^–2). In every further scan, the wave at 3.1 V disappears and the wave at 3.3 V is shifted to 3.4 V with an almost equal current density for every scan. At potentials more positive than 4.3 V the current density increases up to 37 μA·cm^–2. A similar observation was made for the composite electrolyte with Li[OC(H)(CF₃)₂] and BF₃DMC showing no waves in the range of 3.0 V to 4.2 V (Fig. 3e).

The composite electrolyte of Li[CO₂CF₃] and BF₃DMC differs in the cathodic as well as the anodic scan (Figs. 3c and 3f). In the first cathodic sweep two peaks could be observed diminishing to one wave with decreasing current density in the subsequent sweeps. The same is true for the anodic sweeps: two waves in the first sweep with decreasing current density in the following. Experiments with regenerated electrode diffusion layer show no difference to the last sweep of the forerunning experiment (not shown). This indicates that the origin of the observed waves is related to changes of the electrode surface state and not the electrolytes stability. The waves shape and the current density also support this assumption.

Due to the actually rather small current densities observed for all three composite electrolyte systems, they can be considered as stable within the potential range of 0.5 V to 4.5 V vs Li⁺/Li (marking the onset of electrolyte decomposition at about 15 μA·cm^–2). Actually, the electrolyte containing Li[CO₂CF₃] and BF₃·DMC is the most stable one (0.1 V-4.7 V vs Li⁺/Li).

The composite electrolytes containing LiX (with X = OCH₂CF₃, OC(H)(CF₃)₂, CO₂CF₃) and BF₃·D (with D = DMC, EC, EMC, PC) (1.0 mol·L^–1, ratio 1:1) were studied as electrolytes in NCM/graphite cells with L57 as solvent. After the formation cycles, galvanostatic cycling was carried out with a C-rate of C/2 between 3.0 V and 4.2 V at 25 °C. For comparison, cells with the conducting salts LiPF₆ (1.0 mol·L^–1) and LiBF₄ (1.0 mol·L^–1) were built additionally using the same materials.

The specific discharge capacities of cells containing the composite electrolytes are provided in Fig. 4 (top). The initial discharge capacity for cells cycled with the composite electrolytes (D = carbonate) are lower than for cells containing LiPF₆ (155 mAh·g⁻¹) or LiBF₄ (141 mAh·g⁻¹). The lowest initial discharge capacities show cells with Li[CO₂CF₃] and BF₃D (94 mAhg⁻¹−103 mAh·g⁻¹) and similarly cells with Li[OCH₂CF₃] and BF₃·D (94 mAhg⁻¹−96 mAh·g⁻¹). By contrast, the initial discharge capacities of cells including Li[OC(H)(CF₃)₂] and BF₃D as electrolytes (134 mAh·g⁻¹−137 mAhg⁻¹) are almost as high as for cells containing LiBF₄ (141 mAh·g⁻¹), but still lower as that for cells containing LiPF₆ (155 mAh·g^–1).

Zoom In Zoom Out Reset image size

The capacity retention from the 6^th to the 55^th cycle is slightly higher for cells cycled with Li[OC(H)(CF₃)₂] and BF₃·D (with D = carbonate) (92%–94%) than for cells containing LiBF₄ (92%), whereas cells with LiPF₆ (98%) display the highest capacity retention. The observations of the concentration-dependent conductivity measurements are inversed for the full-cell cycling of the adducts BF₃·D (with D = carbonate) with Li[OC(H)(CF₃)₂]. Thus, the electrolytes containing the cyclic adducts BF₃·EC and BF₃PC showed slightly increased ionic conductivity compared to BF₃·DMC and BF₃·EMC. Yet, cells containing BF₃·EC and BF₃·PC display a slightly lower specific discharge capacity compared to cells containing BF₃·DMC and BF₃·EMC.

The replacement of a CF₃ group by a hydrogen atom follows the trend of the concentration-dependent conductivity measurements and leads to lower capacity retentions for Li[OCH₂CF₃] and BF₃·D (with D = carbonate) used as electrolyte (80%–84%). However, the electrochemical behavior of the cells with the respective electrolyte system of Li[OCH₂CF₃]/BF₃·D (with D = carbonate) is very similar to their concentration-dependent conductivities.

Like observed for the initial discharge capacity, cells cycled with Li[CO₂CF₃] and BF₃·D (with D = carbonate) display significantly lower capacity retentions (< 3%) from the 6^th to the 55^th cycle. Even though these composite electrolytes exhibited the highest measured conductivities (up to 5.81 mS·cm^–1) that are higher than the conductivity of LiBF₄ in L57.

The trend of the discharge capacities is confirmed by the coulombic efficiencies (Fig. 4, bottom) which are decreasing in the following order (after the 8^th cycle): LiPF₆ (≥ 99.7%) > Li[OC(H)(CF₃)₂]/BF₃·D (≥ 99.6%) > LiBF₄ (≥ 99.3%) > Li[OCH₂CF₃]/BF₃·D (≥ 99.0%) ≫ Li[CO₂CF₃]/BF₃·D (≥ 85.5%).

For a better understanding of the processes during cycling, the 1^st and 10^th cycle of the full-cells were investigated further. While the first cycle with a C-rate of C/20 provides information of the changes during the initial cycle (e.g. formation of the SEI), the 10^th cycle with a C-rate of C/2 depictures the cell in a proceeded phase. Therefore, the voltage profiles of these two cycles from cells cycled with LiX and BF₃·D and their corresponding differential capacity plots are provided in Figs. 5 and 6. Again, cells cycled with LiPF₆ and LiBF₄ as electrolyte were added for comparison.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The voltage profiles and the corresponding differential capacity plots differ for each LiX, but the differences are minimal for the various adducts using the like LiX and with exception of Li[CO₂CF₃] and BF₃·PC. For cells cycled with LiPF₆ and LiBF₄ as electrolyte, the capacities are significantly higher for the 1^st and 10^th cycle than for cells containing LiX and BF₃·D. Cells cycled with Li[OC(H)(CF₃)₂]/BF₃·D and Li[CO₂CF₃]/BF₃·D display also the typical voltage profile of the NCM electrode with two anodic peaks at 3.5 V and 3.7 V in the first cycle, which are dedicated to the Ni redox pairs (Ni²⁺/Ni³⁺ and Ni³⁺/Ni⁴⁺).⁸³ For cells cycled with LiBF₄ or LiPF₆ both peaks are pronounced in the first cycle. Apparently, this typical voltage profile is not observable for cells cycled with Li[OCH₂CF₃] and BF₃D as electrolyte. The observable peaks display a significant decrease, which indicates a loss of active material. Furthermore, an irreversible oxidation peak between 3.3 V and 3.0 V was detected. We suggest that neutral BX₃ might react on the negative electrode to a radical anionic species [BX₃]^·–, which could dimerize and form a surface layer that could be related to the capacity loss.

All cells cycled with LiX and BF₃·D as electrolyte display small irreversible oxidation peaks in the differential capacity plot of the first cycle. For the composite electrolyte Li[OCH₂CF₃]/BF₃·D and Li[OC(H)(CF₃)₂]/BF₃·D these peaks are in the range of 2.2 V–2.6 V and 2.3 V–2.6 V. Furthermore, two irreversible oxidation peaks occur for Li[CO₂CF₃] and BF₃D in the range of 1.6 V–1.9 V and 2.8 V–3.1 V. Recently, we reported small irreversible oxidation peaks in the first cycle in a similar range for cells cycled with BF₃·D (with D = carbonate) as additive (1 wt%) with LiPF₆ in L57 as electrolyte. We assumed that the formation of a partially boron based surface layer in the first cycle is assigned to these peaks. The XPS spectra of the electrodes of the cells cycled with BF₃·D (with D = carbonate) as additive confirmed the incorporation of boron in the surface layers.⁵² To study, if this is also the case for the composite electrolyte systems, the cycled electrodes were investigated via XPS (see section below). For cells with LiBF₄ or LiPF₆ no oxidation peaks in this voltage area could be observed, which led to higher capacities in the first cycle.

Even if cells containing Li[OCH₂CF₃] and BF₃·D as electrolyte display the most significant difference in the voltage profile of the first cycle, these cells show an increased initial discharge capacity compared to cells cycled with Li[CO₂CF₃] and BF₃·D.

After the 10^th cycle, the capacities of cells with the composite electrolytes are still lower than for cells with LiBF₄ or LiPF₆ (Fig. 6). The differential capacity plot of the 10^th cycle of cells cycled with Li[OC(H)(CF₃)₂] and BF₃D display the typical reversible voltage profile of the NCM electrode material, while for cells cycled with Li[OCH₂CF₃] and BF₃·D only one peak at 3.7 V appeared. After the 10^th cycle, the reversibility of the composite electrolyte Li[CO₂CF₃] with BF₃D is no longer achieved.⁸³ The loss of the reversibility leads to a rapidly decreasing specific discharge capacity, which suggests that the composite electrolyte of Li[CO₂CF₃] with BF₃·D (with D = carbonate) is not suitable for NCM/graphite cells.

EIS measurements were applied at a constant potential of 0 V after the 55^th discharge cycle to compare the impedance induced by the electrolytes. The sum of the full-cell impedance includes contributions of the electrolyte and the electrodes (surface film and bulk material). The Nyquist plots for cells containing the electrolytes LiX/BF₃·DMC, LiBF₄ and LiPF₆ are provided in Fig. 7.

Zoom In Zoom Out Reset image size

For cells containing LiX/BF₃·DMC or LiBF₄,⁸⁴ only one semicircle could be observed, while for cells with LiPF₆ as conductive salt two semicircles were found. According to the literature these semicircles are representing the interphase film impedance (region ii) and the charge transfer resistance (region iii).^85–87 Reasons for observing only one semicircle can be similar capacitances of the surface film and the electrical layer⁸⁸ or the presence of a constant phase element.⁸⁷

The resistance of the lithium conduction of the liquid electrolyte (region i) is represented by the region of high frequencies and decreases in the order LiPF₆ > Li[OC(H)(CF₃)₂]/BF₃·DMC > LiBF₄ > Li[OCH₂CF₃]/BF₃·DMC > Li[CO₂CF₃]/BF₃·DMC.⁸⁹ The sum of the resistances of the interphase film and the charge transfer display an increase of the NCM/graphite cell impedance in the order of LiBF₄ < Li[OC(H)(CF₃)₂]/BF₃·DMC < LiPF₆ < Li[OCH₂CF₃]/BF₃·DMC ≪ Li[CO₂CF₃]/BF₃·DMC. For cells including LiBF₄ and Li[OC(H)(CF₃)₂]/BF₃DMC as electrolyte a stroke line in the low-frequency region represents a low Warburg resistance corresponding to lithium ion diffusion in the active material compared to cells cycled with LiPF₆ or Li[OCH₂CF₃]/BF₃·DMC.⁹⁰

The sum of the resistances is consistent with the capacity retention of the cells cycled with the composite electrolytes. The cells cycled with Li[OC(H)(CF₃)₂] and BF₃·DMC as electrolyte display the lowest resistance and the highest discharge capacity. However, the substitution of a CF₃ group of Li[OC(H)(CF₃)₂] through a hydrogen atom leads to higher impedance and inferior cycling performance. By contrast, cells cycled with Li[CO₂CF₃] and BF₃·DMC as an electrolyte show an extremely high resistance and the lowest discharge capacity.

The relative atomic concentrations of elements detected on the surface of both, fresh electrodes, NCM and graphite, and electrodes cycled with LiX and BF₃·D as electrolyte are shown in Fig. 8. The fresh NCM electrode has almost the same relative atomic concentration distribution as the electrode cycled with Li[OC(H)(CF₃)₂] and BF₃·EMC. The concentrations of boron on the surface of the electrodes cycled with LiX and BF₃·D as electrolyte are increasing in the order of Li[OC(H)(CF₃)₂]/BF₃·EMC < Li[OCH₂CF₃]/BF₃·EMC < Li[CO₂CF₃]/BF₃·PC. At the same time the concentrations of fluorine and carbon on the surfaces are decreasing in the order of Li[OC(H)(CF₃)₂]/BF₃·EMC > Li[OCH₂CF₃]/BF₃·EMC > Li[CO₂CF₃]/BF₃·PC. The NCM electrode cycled in the presence of Li[CO₂CF₃] with BF₃·PC, has a significantly higher oxygen concentration than the other electrodes.

Zoom In Zoom Out Reset image size

Similar results were observed for the concentrations on the surface of graphite electrodes cycled with LiX and BF₃·D. The concentrations of boron, oxygen and carbon are increasing in the order of Li[OC(H)(CF₃)₂]/BF₃·EMC < Li[OCH₂CF₃]/BF₃·EMC < Li[CO₂CF₃]/BF₃·PC, while the concentrations of lithium and fluorine are decreasing in the same order.

The Li 1s, B 1s, C 1s, O 1s and F 1s spectra of fresh graphite and graphite cycled with LiX and BF₃·EMC or BF₃·PC are displayed in Fig. 9. The C 1s spectrum of fresh graphite is dominated by C–O (287.0 eV), C–H (286.2 eV) and C–C (284.8 eV) species, which are related to conductive carbon and CMC/SBR binder.⁹¹ For the electrodes cycled with LiX and BF₃·D, peaks at 290.0 eV could be observed, which are characteristic for Li₂CO₃ or lithium alkyl carbonates from the reduction.⁹² Furthermore, all electrodes cycled with LiX and BF₃·D display an elevated content of C–O (287.0 eV) and C=O (288.0 eV) species. The weak peak of CF₃ groups at 293.0 eV indicates the decomposition of LiX. The O 1s spectra of graphite electrodes cycled with LiX and BF₃·D show a broad peak of oxygenated species in the 536 eV–530 eV binding energy range and confirm the presence of C–O (534.5 eV), C=O (532.0 eV) and M=O (530.5 eV) species on the surface. In the F 1s spectra, LiF (685.0 eV) and LiBF₄ (687.5 eV) are present on every electrode cycled with LiX and BF₃·D as electrolyte. An increased intensity of LiF was observed especially for Li[OCH₂CF₃] and BF₃·EMC, while Li[OC(H)(CF₃)₂]/BF₃·EMC and Li[CO₂CF₃]/BF₃·PC have a higher content of C–F species (688.0 eV) concentration on the surface. This observation is consistent with the Li 1s spectra of LiX and BF₃·D. In the B 1s spectra the intensity of B–F and B–O species is increasing on the graphite surface in the order of Li[OC(H)(CF₂)₃]/BF₃·EMC < Li[OCH₂CF₃]/BF₃·EMC < Li[CO₂CF₃]/BF₃·PC.

Zoom In Zoom Out Reset image size

The B 1s, C 1s, O 1s and F 1s spectra of fresh NCM and NCM cycled with LiX and BF₃·EMC or BF₃·PC are provided in Fig. 10. The C 1s spectrum of the fresh electrode is dominated by peaks of CF₂ (291.0 eV) and CH₂ (286.2 eV) from PVDF and C–C (284.8 eV) from conductive carbon.⁹³ These peaks were found in the C 1s spectra of all cycled NCM electrodes. The C 1s spectra of electrodes cycled with LiX and BF₃·D as electrolyte, contain C–O (287.0 eV) and C=O (288.0 eV) species. The C 1s spectra of Li[CO₂CF₃] and BF₃·PC display an increased intensity of C=O bonds and Li₂CO₃/RLiCO₃ species (290.0 eV) than the surface of the electrodes cycled with the two other electrolytes. Furthermore, a weak peak of CF₃ groups (293.0 eV) could be found on the surface of the NCM electrodes cycled with Li[CO₂CF₃] and BF₃·PC as electrolyte, which indicates the decomposition of [CO₂CF₃]^–.

Zoom In Zoom Out Reset image size

The O 1s spectrum of the fresh positive electrode is dominated by the metal oxide peak at 529.5 eV and two peaks, which are related to Li₂CO₃ at 534.5 eV (C–O) and 532.0 eV (C=O).⁹³ These species could also be observed on the surface of all cycled electrodes. The intensity of C–O and C=O species on the surface of the electrode cycled with Li[CO₂CF₃] and BF₃·PC is significantly higher than for the electrodes containing the two other LiX with BF₃·EMC. The presence of C–O and C=O bonds on the surfaces are increasing in the order of Li[OC(H)(CF₃)₂]/BF₃EMC < Li[OCH₂CF₃]/BF₃·EMC ≪ Li[CO₂CF₃]/BF₃·PC and indicate the decomposition of LiX.

The F 1s spectrum of the fresh electrode contains C–F groups (688.0 eV) of PVDF. All cycled electrodes have also C–F groups on the surface, which are indicating again the decomposition of LiX. The presence of LiF (685.0 eV) on the surface is decreasing in the order of Li[OC(H)(CF₂)₃]/BF₃·EMC > Li[OCH₂CF₃]/BF₃·EMC > Li[CO₂CF₃]/BF₃·PC. The B 1s spectra with Li[OCH₂CF₃]/BF₃·EMC and Li[CO₂CF₃]/BF₃·PC display a peak at 193.0 eV, which could be assigned to B–F and B–O bonds.^88,93 The B 1s spectrum of the electrode from the cell with Li[OC(H)(CF₂)₃] and BF₃·EMC as electrolyte showed no boron compounds on the surface of the NCM electrode.

The results of the XPS spectra of the surfaces of the NCM electrodes are consistent with the observations made during the cycling of LiX with BF₃·D. The XPS spectra are suggesting a decrease of the CEI in the order of Li[CO₂CF₃]/BF₃·PC > Li[OCH₂CF₃]/BF₃·EMC > Li[OC(H)(CF₃)₂]/BF₃·EMC, which is accompanied by a better cycling performance and lower impedance built up of the electrolytes. The incorporation of boron species in the surface layer of the electrodes was observed with different concentrations. According to the cycling performance the presence of boron species on the graphite electrode has no significant effect of the reversibility of the cells (cf. boron content on the surface of the graphite electrode of Li[OC(H)(CF₃)₂]/BF₃·EMC). On the NCM electrode the incorporation of boron species seems to be dominant: the higher the concentration, the lower the reversibility and the capacity retention of the cells. In agreement with this, only the best performing electrolyte of this series, Li[OC(H)(CF₃)₂] with BF₃·EMC, virtually shows no boron-uptake into its surface.