Room-temperature cycling of metal fluoride electrodes: Liquid electrolytes for high-energy fluoride ion cells

Working toward fluoride batteries

Owing to the low atomic weight of fluorine, rechargeable fluoride-based batteries could offer very high energy density. However, current batteries need to operate at high temperatures that are required for the molten salt electrolytes. Davis et al. push toward batteries that can operate at room temperature, through two advances. One is the development of a room-temperature liquid electrolyte based on a stable tetraalkylammonium salt–fluorinated ether combination. The second is a copper–lanthanum trifluoride core-shell cathode material that demonstrates reversible partial fluorination and defluorination reactions.

Science, this issue p. 1144


Fluoride ion batteries are potential “next-generation” electrochemical storage devices that offer high energy density. At present, such batteries are limited to operation at high temperatures because suitable fluoride ion–conducting electrolytes are known only in the solid state. We report a liquid fluoride ion–conducting electrolyte with high ionic conductivity, wide operating voltage, and robust chemical stability based on dry tetraalkylammonium fluoride salts in ether solvents. Pairing this liquid electrolyte with a copper–lanthanum trifluoride ([email protected]) core-shell cathode, we demonstrate reversible fluorination and defluorination reactions in a fluoride ion electrochemical cell cycled at room temperature. Fluoride ion–mediated electrochemistry offers a pathway toward developing capacities beyond that of lithium ion technology.

The search for batteries that offer the high energy density necessary to meet emerging energy storage needs is increasingly focused on alternatives to lithium ion electrochemistry (1). Energy density is dictated by the number of electrons transferred in the reaction (n), the cell voltage or potential difference between cathode and anode (), Faraday’s constant (F), and the cell volume (ΣMi):Embedded Image(1)Exploitation of multi-electron redox processes (n > 1) is an attractive route to achieve improved energy density. Next-generation lithium electrochemical systems, such as lithium-air and lithium-sulfur (Li/S), offer high theoretical energies due to multi-electron reactions at the cathode, but they use lithium metal anodes that have poor cycle life and raise safety concerns (2, 3). Magnesium anode cells can afford two-electron transfer per Mg2+ ion, but they are limited at their cathodes by the poor mobility and the large structural reorganization imposed by this cation with high charge density in a solid lattice (4).

Alternatively, multi-electron electrochemistries can use anionic species as the charge carrier. With mono-anionic charge carriers, reversible multi-electron reactions may be facilitated by moving several singly charged anions per reaction step, rather than one highly polarizing, multiply charged cation. In general, anions have a lower charge density and a lesser degree of solvation than cations of the same absolute charge, and therefore display greater mobilities. High-power devices that use anion-mediated multi-electron redox reactions in aqueous alkaline solution are well established—for example, nickel-cadmium and nickel-zinc batteries, although their energy density is limited (<300 Wh/liter) (5). For higher-energy devices, nonaqueous electrolytes are required to operate reversibly at potentials beyond that of the thermodynamic voltage window of water (1.23 V at 298 K).

In this context, fluoride ion batteries (FIBs) are of considerable interest (6). In contrast to the intercalation/deintercalation mechanisms that operate for most lithium ion batteries, FIBs function by conversion processes. FIB conversion reactions occur, for example (upon discharge), when a high-valent metal fluoride cathode (MFn, Eq. 2) becomes reduced to the metal concomitant with the oxidation of a low-potential metal anode (M′, Eq. 3) to a metal fluoride:Embedded ImageCathode: (2)Embedded ImageAnode: (3)Multivalent fluoride conversion reactions have high thermodynamic reaction potentials (>3 V) and volumetric capacities (>1000 Ah/liter). Hence, batteries based on these materials offer theoretical energy densities up to 5000 Wh/liter (7), ≥8 times the theoretical values for current lithium ion technologies. Rechargeable FIB devices were reported in 2011 (8), following principles established for primary galvanic cells in the 1970s (6). FIBs can only operate at high temperatures (≥150°C) because they use solid-state fluoride electrolytes with very limited ionic conductivity at room temperature (10–6 S/cm) (6, 8). A few electrochemical cells have been reported that exhibit fluoride-mediated conversion reactions in a liquid electrolyte (9–12); however, these cells either contain bifluoride ions (HF2–) with a very small electrolyte voltage window (~0.7 V) (11) or operate through irreversible F– complexation to the metal charge carrier (12). Hence, electrolyte limitations present a serious challenge for designing FIBs at room temperature.

Metal fluoride electrolyte salts are universally insoluble in organics at concentrations exceeding 0.05 M (13). Organic fluoride electrolyte salts, like those with tetraalkylammonium (R4N+) cations (14), could have improved solubility; however, these are nontrivial to synthesize in truly anhydrous form because decomposition of F– to HF2– occurs readily through elimination processes at elevated temperatures (15–17). Commercial anhydrous tetramethylammonium fluoride (TMAF) is of limited solubility in organics. We therefore selected neopentyl-substituted (Np, or 2,2-dimethylpropyl-) alkylammonium salts, as the Np chain is both branched (to improve solubility) and lacks β-hydrogens (to inhibit decomposition upon drying). Both dry N,N,N-trimethyl-N-neopentylammonium fluoride (Np1F) and N,N,N-dimethyl-N,N-dineopentylammonium fluoride (Np2F) were synthesized in batches as large as 100 g using an HF-titration procedure (18–20) and demonstrated greatly improved solubility relative to TMAF (Fig. 1A).

Fig. 1 Physical and electrochemical properties of nonaqueous, fluoride ion–conducting liquid electrolytes (25°C).

(A) Influence of tetraalkylammonium cation structure on fluoride salt solubility in 3-fluorobenzonitrile. Solubility (mol/liter) represents the approximate salt saturation concentration. (B) Np1F solubility in BTFE, acetonitrile (ACN), 3-methoxypropionitrile (MeOPN), and propionitrile (PN). Inset shows 19F NMR spectra in the bifluoride region for Np1F dissolved in each solvent. Reactivity of F– with solvent protons results in HF2– formation (doublet, –146.6 ppm); F– reacts with CD3CN NMR solvent to produce DF2– in all cases (triplet, –147.0 ppm). (C to E) Molecular dynamics simulations for ions in BPFE, BTFE, and diglyme. (C) Comparison of ion solvation free energies. Data are means ± SD. (D) Comparison of F– radial distribution functions calculated with respect to the H (or F) atoms bonded to the α-CX2 moiety of the solvent. (E) Simulated solvation shell of BTFE molecules surrounding F– (pink sphere). (F) Ionic conductivity of Np1F (black) or Np2F (red) in liquid BTFE electrolyte solutions as a function of concentration. (G) F– transport numbers (t–) from PFG-SE 1H and 19F NMR experiments, and ionic conductivity values (σAC) from AC impedance measurements, for 0.75 M Np1F in BTFE and 0.75 M Np2F in BTFE. (H) Stokes radii (Rion) for electrolyte ions in 0.75 M Np1F/BTFE (blue) or 0.75 M Np2F/BTFE (red). The degree of ion dissociation (α) is also shown (SD = ±0.003) (20). (I) Linear sweep voltammograms for 0.75 M Np1F in BTFE, BTFE:DME (3:2 v/v), and BTFE:TEGDME (3:2 v/v) collected using a scan rate of 1 mV/s. DME, 1,2-dimethoxyethane; TEGDME, bis[2-(2-methoxyethoxy)ethyl] ether.

Initial screening of these NpF salts revealed three broad classes of organic solvents: (i) NpF insoluble; (ii) NpF soluble (at appreciable electrolyte concentrations, >0.5 M) but F– reactive; or (iii) NpF soluble, F– stable (but at very limited concentrations). Certain ionic liquids, such as 1-methyl-1-propylpyrrolidinium bis(trifluorosulfonyl)imide (MPPy-TFSI), were found in the third class (fig. S1); however, the associated expense and high viscosity of ionic liquids, and additional complications for battery operation that might arise from the presence of other anions (e.g., TFSI) in the mixture, kept our focus on simple organic solvents. Inspired by studies of Li/S battery systems, we turned to partially fluorinated ethers (21). Bis(2,2,2-trifluoroethyl) ether (BTFE) is the only organic solvent we found to dissolve NpF at substantially high concentrations (>2.2 M) while not reacting with F– (Fig. 1B and fig. S2). Long-term 19F nuclear magnetic resonance (NMR) monitoring confirmed that the solution is stable at room temperature for more than 3000 hours (fig. S3) (20). This contrasts with, for example, the stability of F– in dichloromethane, which shows complete decomposition to HF2– within 12 hours (22).

Molecular dynamics (MD) simulations (20) were used to characterize the solvation free energies of F–, TMA+, Np1+, and Np2+ in BTFE, bis(perfluoroethyl) ether (BPFE), and bis(2-methoxyethyl) ether (diglyme) solvents (Fig. 1C). The F– solvation free energy was much less favorable in BPFE (ΔGS,F = –8 kcal/mol) than in diglyme (ΔGS,F = –59 kcal/mol) and BTFE (ΔGS,F = –63 kcal/mol), which suggests that the α-CH2 feature plays an important role in dissolving F–. The solvation free energies for each R4N+ are similar, indicating that the introduction of bulkier alkyl substituents onto the cation should favorably reduce the salt lattice energy with little effect on the cation solubility.

Analysis of the radial distribution function for solvated F– (for which the largest probability of F– interaction occurs about 2 Å from the H atom of the α-CH2 moiety in BTFE) (Fig. 1D), in concert with quantum chemical calculations to characterize the partial charge distribution in BTFE and diglyme (fig. S4), presents a consistent physical picture of F– solvation. MD simulations reveal the innermost coordination shell of F– in liquid BTFE (Fig. 1E); on average, at least one BTFE molecule has both α-CH2 groups oriented toward the anion. Overall, F– solvation is facilitated by interactions of F– with partial positive charges on the α-CH2 moiety that are enhanced by adjacent electron-withdrawing groups (as such, these are greatest for BTFE). Thus, ethereal media possessing these structural features, such as BTFE and to a lesser extent diglyme, may be considered as useful solvents for room-temperature fluoride ion electrolytes.

Baseline measurements of NpF/BTFE electrolytes indicate ionic conductivities (Fig. 1F) that are comparable to those of lithium ion battery electrolytes (10–3 to 10–2 S/cm) (23). To fully characterize the electrolyte solution properties, we carried out pulsed-field gradient spin-echo (PFG-SE) 1H and 19F NMR and AC impedance measurements for these electrolyte formulations (tables S1 and S2). High F– transport numbers (t– > 0.5) support the assignment of F– as the major contributor to the ionic conductivity and charge mobility in BTFE solutions (Fig. 1G). The degree of ion dissociation (α) for the Np1F/BTFE and Np2F/BTFE electrolytes was estimated from NMR and impedance data to be α = 0.087 and 0.108, respectively, suggesting a considerable degree of ion pairing in solution (20). Stokes radii of the ions (Rion) at room temperature reveal that three BTFE molecules diffuse together with Np2+, whereas only two BTFE molecules diffuse together with Np1+ (Fig. 1H). The increased degree of ion separation and the greater solvation of the Np2+ cation are likely contributing factors for improved conductivity in Np2F/BTFE electrolytes.

Linear sweep voltammetry (LSV) of 0.75 M Np1F/BTFE reveals a cathodic voltage limit of +0.7 V versus Li+/Li and an anodic voltage limit of +4.8 V versus Li+/Li, for an overall electrolyte voltage window of 4.1 V (Fig. 1I and fig. S5). Cathodic stability can be extended up to 400 mV by blending BTFE with a range of straight-chain glycol ethers, motivated by the MD simulations. These wide voltage windows offer the potential to support interfacial redox chemistry for a variety of metal cathode materials in FIB cells. For example, metals such as bismuth [Embedded Image(Bi3+/Bi) = +3.4 V versus Li+/Li], lead [Embedded Image(Pb2+/Pb) = +2.9 V versus Li+/Li], and copper [Embedded Image(Cu2+/Cu) = +3.4 V versus Li+/Li] should be fully compatible as cathode materials with this liquid electrolyte. These metals have previously demonstrated fluoride conversion reactions with limited cycling in high-temperature FIBs (6, 8, 24). We achieved electrochemical cycling of Bi, Pb, and Cu electrodes in a three-electrode cell at room temperature using our liquid electrolytes (Fig. 2A), whereby up to 10 cycles were carried out (Fig. 2, B to E, and fig. S6). In all cases, however, considerable metal dissolution into the electrolyte was found (20). Similar studies of cerium anodes [Embedded Image(Ce3+/Ce) = +0.7 V versus Li+/Li] and calcium anodes [Embedded Image(Ca2+/Ca) = +0.2 V versus Li+/Li] indicated that electrolyte breakdown was considerable (fig. S7), consistent with their potentials at the edge of the electrolyte cathodic window. However, performance was improved by modification of the metal with a fluorinated solid-electrolyte interphase (SEI) layer (25). Pretreatment of Ca or Ce electrodes with 1H,1H,2H,2H-perfluorooctyltriethoxysilane (FOTS) additive (fig. S8) (20) allowed for the desired formation of a CFn-containing SEI layer on the metal surface, whereby reversibility of the Ce-to-CeF3 conversion reaction was improved (Fig. 2, F and G).

Fig. 2 Room-temperature performance of metal electrode materials reversibly cycled in nonaqueous, fluoride ion–conducting liquid electrolytes.

(A) Schematic of external electron flow, electrolyte ion shuttling, and redox reactions occurring at FIB electrodes during charge or discharge cycles. (B and C) Data collected for Bi. (D and E) Data collected for Pb. (F and G) Data collected for Ce with an SEI layer formed from FOTS additive. Voltage profiles of Bi (B), Pb (D), and Ce (F) electrodes were collected during electrochemical cycling in a three-electrode cell (20). Ionic liquid (IL) = 0.1 M TMAF in MPPy-TFSI; BTFE = 0.1 M Np1F in BTFE. Inset in (F) shows expanded view of charge cycle. Ce was cycled in 0.75 M Np1F/BTFE. pXRD patterns were obtained for Bi (C), Pb (E), and Ce (G) electrodes in pristine condition (black), after first charge or fluorination (red), and after final discharge shown or defluorination (blue). Asterisks indicate new peaks corresponding to BiF3, β-PbF2, or CeF3 due to metal fluorination after charge.

To mitigate challenges associated with cathode metal dissolution, we designed composite cathode materials featuring a core-shell nanostructure with an inert thin shell around the active material (essentially an artificial SEI) (26). This would serve to (i) protect this active core from dissolution, (ii) protect the electrolyte from decomposition, (iii) restrict volume expansion and maintain structural integrity of the core, and (iv) selectively percolate electrolyte ions into the core. We selected Cu and LaF3 for the core and shell, respectively, because of the high theoretical specific capacity of CuF2 [528 mAh/g (27)] and the inert nature and highly selective F– conductivity of LaF3 (6), which we assumed would lead to facile F– diffusion between the liquid electrolyte and the Cu core. Figure 3A shows transmission electron microscopy (TEM) images of pristine [email protected] core-shell nanoparticles isolated after synthesis. These spherical nanoparticles are composed of a core 50 nm in diameter and a shell 5 nm thick (Fig. 3A, inset). Energy-dispersive x-ray spectroscopy (EDS) confirmed the elemental composition and stoichiometry of the core and shell regions of the material (Fig. 3B and fig. S9).

Fig. 3 Characterization of [email protected] core-shell cathode materials and their electrochemical cycling at room temperature.

(A) TEM image of pristine core-shell nanoparticles. Inset shows a high-resolution TEM (HR-TEM) image of the thin LaF3 shell encasing the thicker Cu core. (B) Image of pristine [email protected] nanoparticles obtained via EDS showing elemental distribution map of Cu (green), La (blue), and F (red). (C) Electrochemical charge and discharge curves for a three-electrode cell with [email protected] cathode in 1 M Np1F/BTFE, cycled at 10 μA. (D) pXRD of [email protected] cathode material in pristine condition, after first charge (fluorinated), and after seventh discharge (defluorinated). (E to G) EELS data in the scanning TEM mode, with quantification results for shell, interface, and core regions of [email protected] cathode nanoparticles after first charge (fluorinated). (E) HR-TEM image of a fluorinated [email protected] nanoparticle. (F) Representative EELS spectra showing F K, La M5,4, and Cu L3,2 edges obtained for the fluorinated sample in the shell, interface, and core regions. Insets show the F K-edge and Cu L3,2-edge from the graph below. (G) Plot of averaged elemental percent of Cu, La, and F obtained from 36 different EELS spectra, 12 each from shell, interface, and core regions of multiple fluorinated particles. (H) Cyclic voltammogram of Cu-LaF3 thin-film electrode in 0.1 M TMAF/MPPy-TFSI for 10 cycles. (I) X-ray photoelectron spectroscopy (XPS) depth profiles for Cu-LaF3 thin-film electrodes in pristine condition (dashed curves) and after charge (solid curves).

Electrodes fabricated with these [email protected] nanoparticles were cycled reversibly at room temperature in a three-electrode cell for seven cycles (Fig. 3C). Reversible conversion of Cu to CuF2 is evidenced by powder x-ray diffraction (pXRD) (Fig. 3D), which suggests that the LaF3 shell permits passage of F– to enable CuF2 formation as intended. The desired protective nature of this shell is also clear, as inductively coupled plasma mass spectrometry of the electrolyte solution after cycling found no evidence for Cu or La dissolution within the limits of instrument detection (<10 μg), unlike for the uncoated Cu electrodes discussed above. A control experiment using these [email protected] particles in electrolyte containing no F– confirmed that the LaF3 shell does not act as the F– source for the formation of CuF2; rather, the conversion of Cu to CuF2 does indeed occur via F– diffusion from the liquid electrolyte to the nanoparticle core and vice versa (fig. S10).

Detailed analysis of the fluorinated [email protected] nanoparticles (i.e., after charging) was carried out using electron energy-loss spectroscopy (EELS) by scanning TEM (Fig. 3, E to G) (20). In contrast to the pristine [email protected] nanoparticles, the nanoparticle structure after the first charge resembles a yolk-shell structure, where distinct void spaces (or interface) can be seen between the shell and the core (Fig. 3E and fig. S11). Representative EELS spectra show discrete F K, La M5,4, and Cu L3,2 edges for three regions at different depths within the fluorinated nanoparticle (Fig. 3F). Averaged elemental compositions show Cu-only cores, Cu- and F-containing interfaces, and La- and F-containing shells (Fig. 3G). In a limited number (2/13) of spectra taken from the interface region, quantitative analysis revealed a >2:1 ratio of F to Cu, indicating that more F– is available in the interface region than Cu is available to react with it. These results suggest that F– diffusion into the core (rather than through the shell) is a limiting factor, and if so, the limited total electrode capacity observed may be due to this bottleneck.

This hypothesis was confirmed through studies of thin-film structures of similar core-shell composition (20). Cyclic voltammetry of the thin film (Fig. 3H) reveals behavior similar to that of the nanoparticle material, with distinct peaks for the Cu-to-CuF2 conversion process evident (oxidation at ~3.2 V, reduction at ~2.3 V); the relative asymmetry of the anodic peak indicates that the CuF2 formation process is the most difficult electrochemical process, consistent with limited F– diffusion into the Cu core film. X-ray photoelectron spectroscopy (XPS) depth-profile studies of thin-film structures confirm this assessment (Fig. 3I), indicating that F– penetrates through the entire shell but through only ~4.9 nm of the Cu core, for a total F– diffusion length of ~8.4 nm. On the basis of this observed diffusion length, [email protected] nanoparticles with a shell thickness of 2 nm and a core diameter of 12 nm might allow for complete conversion of the Cu core to CuF2 upon first charge, and much higher practical use of the material upon cycling.

Using a simple yet robust liquid electrolyte with high fluoride ion conductivity and wide voltage window, we have demonstrated reversible electrochemical cycling of metal fluoride electrodes at room temperature whereby F–, not the metal cation, is the active ion shuttle. These results point toward FIBs that operate at room temperature. In particular, optimization of a [email protected] core-shell cathode and its pairing with an electropositive metal anode, such as Ce, offers a path toward developing a high-energy device.

Acknowledgments: This work is dedicated to the memory of Nebojša Momčilović. Funding: The research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with NASA. V.K.D. thanks the NSF Graduate Research Fellowship Program for support under grant NSF-DGE-1650116. T.F.M. acknowledges NSF under DMREF award NSF-CHE-1335486. M.A.W. acknowledges the Resnick Sustainability Institute. This research used computational resources from the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy (DOE) under contract DE-AC05-00OR22725. This research also used resources of the National Energy Research Scientific Computing Center (NERSC), a DOE Office of Science User Facility supported by the DOE Office of Science under contract DE-AC02-05CH11231. STEM/EELS work at the Molecular Foundry was supported by the DOE Office of Science, Office of Basic Energy Sciences, under contract DE-AC02-05CH11231. We acknowledge support from the Beckman Institute of the California Institute of Technology to the Molecular Materials Research Center. Author contributions: The project was conceptualized by S.C.J. and supervised by K.O., C.J.B., R.H.G., and S.C.J.; fluoride salt synthesis procedures were developed and performed by V.K.D., C.M.B., N.M., I.M.D., and N.G.N.; salt characterization, analysis and solvent screening was performed by V.K.D., C.M.B., and N.M.; ionic conductivity and voltage window studies were performed by V.K.D. and K.J.B.; PFG-SE NMR experiments and analysis was carried out by W.J.W. and V.K.D.; computational studies were performed by B.M.S. and M.A.W. under the supervision of T.F.M.; electrochemical cells were built and tested by Q.X., N.H.C., and K.O.; cathode materials were synthesized by R.K.M.; TEM, EDS, and pXRD was performed by Q.X.; EELS was performed by S.A., D.R., and M.A.; V.K.D. and A.H. performed XPS measurements; V.K.D. made all figures in the main paper and in the supplement; and the manuscript was written by V.K.D. with input from S.C.J. and all authors. Two patent applications have been filed by Caltech and Honda Motor Co. Ltd.: US 15/228,876 (inventors: S.C.J., V.K.D., C.M.B., N.M., B.M.S., M.A.W., T.F.M., R.H.G., C.J.B., and K.O.) and US 15/844,079 (inventors: N.H.C., K.O., R.K.M., Q.X., C.J.B., S.C.J., I.M.D., and Hongjin Tan). Competing interests: All authors declare that they have no competing interests. Data and materials availability: Force-field parameters and example inputs for the MD simulations performed are available for download (data S1). All other data are available in the manuscript or the supplementary materials.

Comments are closed.