Electrochemical Conversion of Nitrogen Trifluoride as a Gas-to-Solid Cathode in Li Batteries

The MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters.

Citation He, Mingfu et al. "Electrochemical Conversion of Nitrogen Trifluoride as a Gas-to-Solid Cathode in Li Batteries." Journal of Physical Chemistry Letters 9, 16 (July 2018): 4700-4706 © 2018 American Chemical Society

As Published http://dx.doi.org/10.1021/acs.jpclett.8b01897

Publisher American Chemical Society (ACS)

Version Author's final manuscript

Citable link https://hdl.handle.net/1721.1/121525

Terms of Use Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. Electrochemical Conversion of Nitrogen Trifluoride

as a Gas-to-Solid Cathode in Li Batteries

Mingfu He†,Yuanda Li†, Rui Guo†, and Betar M. Gallant†,*

†Department of Mechanical Engineering, Massachusetts Institute of Technology, MA 02139,

United States

*E-mail: [email protected]

1

ABSTRACT: Nonaqueous metal-gas batteries have emerged as a growing family of primary and rechargeable batteries with high capacities and energy densities. We herein report a high- capacity primary Li-gas battery that uses a perfluorinated gas, nitrogen trifluoride (NF3), as the cathode reactant. Gravimetric capacities of ~1100 and 4000 mAh/gC are achieved at 25 and

+ 55°C, respectively (at 20 mA/gC), with discharge voltages up to 2.6 V vs Li/Li . NF3 reduction

- occurs by a 3e /NF3 process, yielding polycrystalline lithium fluoride (LiF) on a carbon cathode.

The detailed electrochemical NF3 conversion mechanism is proposed and supported by solid- and liquid-phase characterization and theoretical computation, revealing the origin of observed discharge overpotentials and elucidating the significant contribution of N-F bond cleavage.

These findings indicate the value of exploring fluorinated gas cathodes for primary batteries; moreover, they open new avenues for future targeted electrocatalyst design and cathode materials synthesis applications benefitting from conformal coatings of LiF.

TOC GRAPHIC

2

Nonaqueous metal-gas batteries have undergone intensive development over the past decade.

Though metal-gas batteries have a long history dating back to early primary Li-sulfur dioxide

1 2 (Li-SO2) and Li-oxygen (Li-O2) batteries, the field experienced a resurgence recently based on

3 renewed enthusiasm for the prospective rechargeability of Li-O2 batteries. With high theoretical

4,5 energy densities (~3500 Wh/kg) compared with Li-ion batteries (~250 Wh/kg), Li-O2 batteries are attractive for electric vehicles if suitable reversibility and cycle life can be achieved.

Unfortunately, their viability is still in doubt owing to persistent challenges, including parasitic side reactions, high charge overpotentials and poor cycle life.5

Regardless, attempts to address these challenges have led to vital new insights and needed innovation in electrochemical technologies. This is evidenced strongly in the recent efforts to

6 7 8 expand the metal-gas family; past years have seen the emergence of the Na-O2, K-O2, Li-CO2,

9 10 11 Li-O2/CO2, Na-CO2, and rechargeable Li-SO2 battery. While most efforts to date have focused on modifying the metal anode for O2 batteries, there is significant opportunity to explore novel chemistries for gas cathodes, specifically those that offer prospects of high capacity and voltage. As many gas cathodes are not necessarily rechargeable at the outset, near-term applications include high-energy density primary batteries for remote and backup power, autonomous underwater vehicles, and military and space applications. However, such investigations can contribute essential new understanding to electrochemical conversion mechanisms of gases and also help to pave the way forward towards longer-term development of rechargeable systems. Our own recent efforts have shown that gas cathode reactions need not be limited to common O2, CO2 or SO2, but can also involve nominally ‗inert‘, highly fluorinated gases.12

3

The present work explores the viability of a highly-fluorinated gas, nitrogen trifluoride (NF3), as a gas cathode in a Li battery. NF3 is a colorless, nonflammable gas with routine usage in the

13 microelectronics industry. When paired with a Li anode, NF3 can theoretically undergo a three- electron reduction, yielding stoichiometric LiF and N2 gas. This reaction requires two NF3

- + molecules, viz. 2 NF3 + 6 e + 6 Li = 6 LiF + N2, and N-N bond formation. Given the stability of the products, LiF and N2(g), the 6-electron reaction has a highly negative Gibbs free energy of reaction, ΔrG°, of -3351 kJ/mol, and corresponds to a remarkable theoretical electrochemical potential of 5.79 V vs. Li/Li. Moreover, the theoretical discharge capacity, normalized to the weight of reactants, is 876 mAh/g. If this free energy can be harnessed in a battery, a theoretical

-1 gravimetric energy density of 5970 Wh kg can be achieved. However, NF3 is generally considered chemically inert, as reactions usually need to be initiated by high energy,14–16 and to the best of our knowledge, its electrochemical activity has not been investigated.

We thus herein studied the use of NF3 as the reactive gas in a Li battery employing a glyme- based electrolyte and a catalyst-free carbon cathode (Scheme 1). First, to determine the suitability of using NF3 in a nonaqueous Li environment, the electrochemical behavior was screened in typical battery electrolytes consisting of 0.07 - 0.1 M lithium perchlorate (LiClO4; concentration adjusted in accordance with solubility) in diethylene glycol diethyl ether

(diglyme), tetraethylene glycol dimethyl ether (TEGDME), acetonitrile, propylene carbonate

(PC), and dimethyl sulfoxide (DMSO). Swagelok-type Li-NF3 cells (Figure S1) were constructed within an argon glovebox using a PC electrolyte-stabilized Li metal foil as the anode and Vulcan

Carbon (VC) as the cathode. In acetonitrile only, due to its reactivity with Li, LiFePO4 was used as the anode.17 Additional details can be found in Supporting Information (SI). Galvanostatic discharges at 100 mA/gC are shown in Figure S2. While varying degrees of activity were

4

a Scheme 1. Schematic of a primary Li-NF3 battery.

+ a. Li, Li -containing nonaqueous electrolyte, and carbon/dissolved NF3(g) are used as the anode, electrolyte, and cathode, respectively. During discharge, two NF3 molecules are nominally - reduced to form LiF solids and liberate N2(g) by an overall six-electron process, 2NF3 + 6e + + 6Li 6LiF + N2.

observed in all solvents, the discharge profile in acetonitrile exhibited sloping behavior and did not exhibit a clear end of discharge within the tested voltage window. Meanwhile, the capacities in PC and DMSO were relatively low (< 300 mAh/gC); furthermore, the separator using DMSO turned brown after discharge, indicating electrolyte decomposition. In contrast, discharge capacities in glyme-based solvents yielded moderate capacities of 460 - 500 mAh/gC with evident voltage quasi-plateaus around 2.10 - 2.18 V vs. Li/Li+; thus, diglyme was selected as the solvent for continued analysis.

Physicochemical properties of the NF3-LiClO4 (diglyme) system were next investigated in detail. The solubility of NF3 in H2O was previously reported to be as low as 0.8 mM in H2O at the partial pressure of 14.70 psi (1 atm) and 298 K,18 but solubilities in organic solvents have not been widely reported. Our attempts to measure NF3 solubility in diglyme using a method adapted from that reported previously12 were challenged due to low solubilities. We did, however, verify

19 the presence of NF3 in the electrolyte from F-NMR measurement (Figure S3) and estimated its

5 solubility, given instrument resolution, to be below 1 mM. Although this solubility is lower than those typically observed in a metal-gas battery (e.g., 1 - 10 mM for O2 in nonaqueous electrolytes), it did not impede electrochemical measurement, as discussed further below. The chemical compatibility of NF3 with electrolyte components was also investigated. NF3 is an oxidizer; although highly electronegative fluorides reduce the nucleophilicity of the nitrogen lone pair, parasitic reactivity of NF3 with ether oxygens or salt anions was a possible concern.

1 Thus, H NMR and Raman measurements were conducted on electrolytes bubbled with NF3 and

1 maintained for 24 hours under an NF3 headspace. As shown in Figure S4, H NMR peaks at  =

3.38, 3.56, and 3.65 ppm are assigned to diglyme, and the peak positions and integral ratios

(3:2:2, respectively) remained identical after NF3 treatment, indicating that no reaction occurred.

The electrolyte stability is further corroborated by identical Raman spectra (Figure S5) before and after NF3 dissolution.

Galvanostatic discharge behavior of Li-NF3 batteries was next investigated. As shown in

Figure 1a, upon discharge at low currents (20 mA/gC, 25 °C), the voltage decreased from OCV and reached a quasi-plateau of 2.10 - 2.25 V, which was retained throughout discharge to a

+ capacity of 950 mAh/gC before the voltage decreased to the cutoff of 1.6 V vs. Li/Li . The eventual observed ―sudden death‖19 of the battery discharge is likely due to the accumulation of electrically insulating products on the cathode, mainly composed of polycrystalline LiF, as discussed later. We note that OCVs of the Li-NF3 battery during the discharge, obtained from the galvanostatic intermittent titration technique (GITT) measurement, were significantly higher than the discharge voltage and also decreased steadily from 2.98 – 2.84 V (Figure S6a and b), likely due to gradual coating of the carbon by these discharge products. Further increasing the current density from 20 mA/gC to 500 mA/gC under direct galvanostatic conditions led to a decrease in

6

Figure 1. Galvanostatic discharge profiles of Li-NF3 batteries. (a) Rate capability (20 – 500 mA/gC) at 25 °C; (b) Temperature effect (25 – 55 C) at 50 mA/gC; (c) Rate capability at 55°C; and (d) Comparison of attainable discharge capacities from panels (a) and (c).

both the discharge voltage and full discharge capacity, e.g. 497 mAh/gC at 100 mA/gC and 199 mAh/gC at 500 mA/gC. In comparison to possible theoretical (thermodynamic) potentials (Table

- + S2 and S3) for reduction of NF3 and precipitation of F with Li , the obtained discharge voltages below 2.3 V suggest the occurrence of high overpotentials (larger than 1.9 V). These overpotentials are unlikely to originate from the Li anode given its low discharging overpotentials in glymes,20 and are thus presumed to arise from the cathode. To evaluate whether concentration losses were a likely significant contributor to overpotentials due to the low solubility of NF3, a discharging Li-NF3 cell was intentionally stopped and relaxed to OCV for

7 one hour, until the OCV ceased changing significantly (< 1 mV/min) and concentration gradients were equalized. Upon resumption of the same current, the discharge voltage was similar to that before (Figure S6c), indicating minimal effects of the concentration polarization.5 This was further confirmed by rotating disk electrode measurements, as presented later. We conclude that concentration is not the most significant source of overpotential, which are most likely due to sluggish kinetics (discussed later in the text).

To investigate whether the performance of the Li-NF3 battery could be improved, the battery temperature was increased from 25 °C to 35 and 55 °C (Figure 1b). In response, the discharge voltage increased modestly, from 2.19 V at 25 °C to 2.25 V at 55 °C . However, the most significant changes occurred with the discharge capacity, which increased substantially by almost three-fold (547 mAh/gC to 1547 mAh/gC) at 50 mA/gC (Figure 1b). The full rate capability at 55 °C is shown in Figure 1c and compared in Figure 1d, with a maximum capacity of 3966 mAh/gC obtained at 20 mA/gC and 55 °C . The increased capacity trend is similar to those observed in Li-O2 batteries and can likely be attributed to enhanced solubility of reduction intermediates in the electrolyte.21,22 Electrode architectures were not optimized in these studies, and these metrics hold potential to be improved through engineering of the cathode substrate.

We next turned to mechanistic investigation of the electrochemical reactions underlying the significant observed reductive activity. First, to validate that the discharge reaction involves

- consumption of NF3 and to determine the Faradaic ratio (e /molecule of NF3 reacted), the pressure within the cell headspace was monitored using a transducer during galvanostatic discharge (additional details in the SI). The discharge curve at 50 mA/gC and the corresponding pressure are shown in Figure 2a. After a minimal delay (< 60 mAh/gC), the cell pressure

8

Figure 2. (a) Galvanostatic discharge of a Li-NF3 battery at a current density of 50 mA/gC (yellow line) and the corresponding cell pressure (gray circles) at 25°C. The solid black line is - the expected pressure change for a Li-NF3 battery following a 3e /NF3 overall reaction of 6 Li + 2 NF3 = 6 LiF + N2. (b) X-ray diffraction patterns of the fully discharged and pristine cathode. The LiF (PDF#04-0857) XRD reference pattern is also included. (c) Scanning electron microscopy image and (d) X-ray photoelectron spectra survey scan of a pristine and a fully discharged cathode (20 mA/gC, 25 °C, 1130 mAh/gC). (e) High-resolution F 1s and (f) N 1s X- ray photoelectron spectrum of the fully discharged cathode.

decreased linearly during discharge, confirming that NF3 was consumed at a constant rate. The

-4 rate of pressure decrease was 1.9210 psi/(mAh/gC), which agrees accurately with the predicted

-4 pressure decrease (1.9210 psi/(mAh/gC)) for the overall reaction of 6 Li + 2 NF3(g) = 6 LiF +

N2(g). Results from three separate measurements yielded similar results within 4% average error

(Table S1). The near-exact agreement indicates that the reaction occurs with high Faradaic efficiency indicative of minimal side reactions.

To further verify the proposed overall reaction, qualitative and quantitative evidence for the formation of LiF in the discharged phase was sought. Figure 2b compares the X-ray diffraction

(XRD) pattern of a pristine and fully discharged cathode. All extra peaks from the fully

9 discharged cathode could be indexed to be LiF phase (cubic, PDF#04-0857). Comparison of cathodes before and after discharge (Figure 2c) by scanning electron microscopy (SEM) revealed that LiF assumed the shape of cubic-like structures with lateral lengths of ca. 100 nm. To quantify the LiF formed, fully discharged cathodes were soaked in D2O to dissolve LiF, and resulting solutions were subjected to quantitative 19F-NMR measurement. A typical 19F-NMR spectrum (Figure S7) shows a peak at  = -122.8 ppm corresponding to LiF.23 From three separate samples, the LiF amount was quantified as (98 ± 12)% of the expected LiF amount if the Li-NF3 battery‘s overall reaction is formulated as 6 Li + 2 NF3 = 6 LiF + N2. X-ray photoelectron spectroscopy (XPS) of discharged cathodes (Figure 2d) further confirmed the presence of copious LiF and negligible nitrogen-containing species in discharged electrodes. In comparison to the pristine cathode, the discharged cathode contained significant amounts of Li

(43.9 % in atomic ratio) and F (38.7 %), and significantly reduced C (8.6 % in compared to 59.1

% of the pristine electrode), which indicated that the carbon particles had been coated with LiF particles during discharge. The dominance of LiF on the cathode surface was further corroborated by the high-resolution F 1s spectrum (Figure 2e), which only exhibited a peak at

685.1 eV, assignable to LiF species and distinct from the F 1s peak at 689 eV originating from the fluorinated binder in pristine electrodes (Figure S8). Although the N 1s spectrum (Figure 2f) showed a peak of finite area at 399.1 eV, surface N-containing species were negligible compared to the amounts of LiF (1.2 % atomic concentration compared to the integrated F 1s peak associated with LiF). The very small quantities of N-containing species could result from incompletely reacted NFx reduction products. Taken together, these results provide strong evidence that the overall reaction is 6 Li + 2 NF3 = 6 LiF + N2.

10

Figure 3. (a) Cyclic voltammogram traces of NF3-saturated 0.1 M LiTFSI/diglyme solution at scan rates ranging from 2 – 50 mV/s, and (b) the cathodic peak current dependence on the scan + + rate from (a). (c) Cyclic voltammogram traces of NF3-saturated 0.1 M MTFSI (M = Li , Na , and K+) electrolytes with different alkali cations, at a scan rate of 10 mV/s.

To gain further insight into the reaction mechanism and thus the limiting behaviors of the NF3 reduction reaction, cyclic voltammetry (CV) of NF3 was carried out in electrolysis-type cells

11 containing excess electrolyte with dissolved NF3 and a glassy carbon working electrode. For CV measurements only, alkali bis(trifluoromethane)sulfonamide (TFSI), instead of perchlorate salts, were used because of the higher solubilities than alkali perchlorate in diglyme, specifically for

+ - - non-Li cations. The CV responses of NF3 in solutions containing either TFSI or ClO4 -based salt had similar reduction onset potentials (Figure S9), which supports this approach. Figure 3a shows cyclic voltammograms of NF3-saturated 0.1 M LiTFSI/diglyme solution on the first scan, for scan rates ranging from 2 to 50 mV/s. Only cathodic peaks were observed with peak potentials occurring at approximately 1.79 to 1.48 V vs Li/Li+, while no anodic peaks were observed on the subsequent positive scan, indicating that NF3 reduction is irreversible. The cathodic peak currents were found to be proportional to the square root of the scan rate (Figure

3b), indicating that freely diffusing species in the electrolyte, rather than surface-adsorbed species, were reduced to generate cathodic currents.24 These cathodic currents are clearly attributable to the electrochemical reduction of dissolved NF3, since CV curves of Ar-saturated electrolyte only showed capacitive behaviors (Figure S10). The onset potentials of these CV traces were within 2.0-2.4 V vs. Li/Li+, which agreed well with the galvanostatic discharge voltages of Li-NF3 batteries (Figure 1a). Moreover, the subsequent CV scan showed negligible cathodic currents (Figure S11), due to passivation of the working electrode by LiF from the first scan.

Given the large overpotentials associated with NF3 reduction, cyclic voltammetry was also employed to provide further insight into the voltage regimes over which the kinetics limit the discharge behavior. Rotating disk electrode measurements were thus conducted to attempt to isolate the kinetic behavior. At 1600 RPM (Figure S12), no significant difference in onset potential (< 200 mV) was observed compared to the stationary condition, supporting the fact that

12 kinetics, and not NF3 transport or reactant depletion, is responsible for the large overpotentials.

We thus further looked into an important factor describing the electrochemical reduction kinetics, which is the electron transfer coefficient, .25 Typically,  is taken to be 0.5 for a reversible, outer-sphere reaction; however, our results show that NF3 is highly irreversible due to

N-F bond cleavage, and thus it is interesting to evaluate the link between irreversibility and the electrochemical kinetics. A theoretical model, i.e. dissociative electron transfer, describing bond-breaking-coupled electrochemical reactions has been developed by Savéant,26 and is summarized in the SI. This model has been found to apply to halide-containing species.27–29 In these cases,  values are typically within ~0.2 - 0.4; the smaller  value in the dissociative electron transfer process results from the contribution of bond dissociation energy to the standard activation Gibbs free energy, and an  value much smaller than 0.5 can also indicate a concerted, as opposed to stepwise, bond breaking process.26 By analogy to previous halogenated systems above, we expect that NF3 also undergoes a dissociative electron transfer process upon discharge, given the extensive experimental evidence for LiF formation. In accordance with

Costentin et al.,28  was obtained from the peak and half-peak potential separation in CV measurements, i.e.,  = 1.856 RT (Ep-Ep/2 )/F = 47.8 mV /(Ep-Ep/2), at 298 K, and was thus calculated to be ~0.16 - 0.19, with minimal variation depending on the scan rate, from Figure 3a.

The obtained NF3-Li significantly smaller than 0.5 indicates a concerted dissociative electron transfer process and that excess overpotential needs to be applied to impart a favorable kinetic landscape for electron transfer to NF3. To provide further validation of the relevance of the concerted dissociative electron transfer model, which predicts rapid dissociation of NF3 into NF2

- .- and F upon electron transfer, we rationalized that an intermediate NF3 , if present (and not predicted by the model), would exhibit a sensitivity of the onset potential to the identity of the

13

30 alkali cation, similar to observations in the O2 reduction reaction (ORR). As shown in Figure

3c, however, nearly identical CVs were obtained for reduction of NF3 with different cations,

+ + + namely Li , Na , and K . This finding suggests that the F2N-F bond dissociation is concerted with the electron transfer. Moreover, it shows that the alkali cation does not determine the reduction potential, likely arising from sluggish ion transfer compared to a rapid fluoride ion expulsion step. Taken together, these findings indicate that the first reduction step, and likely the followings steps, should be considered as an electrochemical-chemical process, and that the precipitation step of LiF should in fact not be considered in the free energy analysis when estimating the expected thermodynamic potential.

The stepwise thermodynamic landscape was thus more accurately evaluated based on this dissociative electron transfer model, which describes the electrochemical potentials of F- formation rather than LiF formation, in contrast to the balanced reactions obtainable from standard thermochemical data (Table S3). This was achieved by constructing a series of

Bordwell thermodynamic cycles (Figure 4a; additional details in Tables S2-S5 and associated discussion). Figure 4b shows three plausible stepwise reaction pathways, and their computed free energy landscapes, consistent with the reaction stoichiometry determined in this work. As can be seen in Figure 4b, the first step common to all pathways involves the single-electron reduction of

- + NF3 and N-F bond breaking, yielding an NF2 radical and F anion. As stated above, as the Li transfer is decoupled from the electron transfer, precipitation of LiF is not reflected in the electrode potential (inclusion of the free energy of precipitation in each step would only further lower each free energy by 0.97 eV per LiF). The corresponding electrochemical reduction

- + potential, E°(NF3/NF2 + F ), was calculated to be 3.67 V vs. Li/Li (Figure 4c, also see Table

S5). As all pathways begin with this step, the first electron transfer pins the electrode potential,

14

Figure 4. (a) A Bordwell thermodynamic cycle linking the homolytic N-F bond dissociation reaction between the gas and solution phase to obtain the stepwise standard reduction potential of NF3 or any of its lower fluorides. ‗RX‘ is used to represent a generic nitrogen fluoride, e.g. NF3 or any of its lower fluorides; ‗R‘ represents nitrogen and ‗X‘ represents the fluorine that is dissociated in each single-electron transfer. More details about notations can be found in Scheme S1. (b) Relative Gibbs free energy diagram, based on stepwise electron transfer reactions, - corresponding to full reduction of 2 NF3 to 6 F and N2 (stoichiometric equivalents of LiF, if formed at each step, would further lower each intermediate and the product N2 by 0.97 eV per LiF). The x-axis follows the overall reaction progress as a function of electron transfer number. (c) Standard reduction potentials, E°(RF/R.+F-), computed for each intermediate in (b) following the dissociative electron transfer model. The blue numbers above each step indicate the standard ǂ activation energy in the gas phase, ΔG0 , of each reaction intermediate. In both panels, dashed black lines indicate electrochemical steps, whereas dashed gray lines (connecting the two pathways) indicate chemical steps.

- which also occurs elsewhere in multi-electron transfer reactions; for example, with the CO2/CO2

31,32 reduction potential in aqueous CO2 reduction. For the Li-NF3 system, this value of 3.67 V vs.

Li/Li+ helps to temper interpretation of the apparently excessive overpotentials compared to the

15 nominal potential (5.79 V vs. Li/Li+); this 6-electron potential, which reflects an average potential of sequential steps, cannot be attained in a stepwise reaction. Moreover, any attainable potential will be lower by about 0.97 V without the additional free energy contributed by precipitation of LiF. The gas-phase standard activation energy of NF3 was also computed (Figure

4c) to be 0.66 eV, which helps to further understand why the observable reduction potentials

(~2.0 - 2.4 V vs. Li/Li+) are significantly lower than 3.67 V vs. Li/Li+. Combining the thermodynamic potential and estimated kinetic overpotential, a reasonable expectation of the discharge potential is ~3 V for a reaction pinned by the first electron transfer and F- dissociative step, and will be further lowered by the low value of  in practice. These results indicate that

- future work to catalyze this first step, either by promoting NF3 adsorption or stabilizing the NF3 for concerted Li+ transfer, could prove fruitful.

Following the initial electron transfer, three subsequent pathways are possible, as shown in

Figure 4b. These three pathways reflect homogeneous coupling between radicals at different stages of reduction. As all three pathways are highly thermodynamically downhill, kinetics will determine which pathway is dominant in practice; we note that coupling between two NF2 radicals, forming N2F4, is both thermodynamically possible and kinetically facile because the

33–35 reverse reaction, i.e. N2F4= 2 NF2, is highly unfavorable according to previous studies. The kinetics of other coupling reactions are not known, and further determining the pathway is speculative at this point. However, the experimental evidence indicates that, regardless of pathway, the reaction goes to completion. Thus, in spite of some high kinetic barriers of intermediates (i.e. NF2, N2F4, and NF, also have large kinetic barriers of 0.64 – 0.79 eV, Figure

4c and Table S6), these appear to be surmountable at the electrode potentials applied herein (~2.1

– 2.4 V vs. Li/Li+). After these high-kinetic-barrier intermediates, all three pathways are both

16 thermodynamically feasible and kinetically facile with minimal kinetic barriers (0 – 0.21 eV).

Taken together, the insights provided by this analysis provide a highly useful framework for evaluating the observed reduction potentials for NF3 as well as future reactions likely to involve electron transfer-coupled bond-breaking events for battery cathodes.

In conclusion, we have demonstrated the feasibility of a primary nonaqueous Li-NF3 battery in which the electrochemical reduction of NF3 at a catalyst-free carbon cathode undergoes a multiple-electron (3 e-/molecule) reduction process, leading to the complete cleavage of all three

N-F bonds and generation of N2 and stoichiometric LiF as the final products. Our study sheds light on the origin of the observed reduction potentials, and provides a framework for evaluating the attainable potentials of multi-ligand, halogenated gases in alkali metal environments relevant to development of future batteries. Efforts are still needed to improve the Li-NF3 battery and realize its intrinsic potential for energy delivery (i.e., primary) applications, including designing tailored electrocatalysts to facilitate the N-F bond breaking process, screening of wider electrolyte systems to improve NF3 solubility, and/or exploring strategies to control the growth of the insulating LiF discharge products. Overall, the reduction process is remarkably well- behaved, yielding clean LiF layers with high Faradaic efficiency; thus, this work also opens new opportunities for applications where conformal LiF coatings may prove beneficial in electrode materials synthesis, such as advanced Li-ion battery cathodes.36,37

ASSOCIATED CONTENT

Supporting Information. Experimental section, thermodynamics and kinetics calculations, and additional figures and tables.

AUTHOR INFORMATION

17

Corresponding Author

*[email protected] (B.M.G)

ACKNOWLEDGMENT

The authors greatly appreciate the financial support from an MIT Bose Fellowship research grant. This work made use of the MRSEC Shared Experimental Facilities at MIT, supported by the National Science Foundation under award number DMR-14-19807. We also acknowledge the help of Aliza Khurram on the data processing of in situ pressure measurement results.

REFERENCES

(1) Linden, D.; McDonald, B. The Lithium—Sulfur Dioxide Primary Battery — Its Characteristics, Performance and Applications. J. Power Sources 1980, 5 (1), 35–55.

(2) Abraham, K. M.; Jiang, Z. A Polymer Electrolyte-Based Rechargeable Lithium/Oxygen Battery. J. Electrochem. Soc. 1996, 143 (1), 1.

(3) Peng, Z.; Freunberger, S. A.; Chen, Y.; Bruce, P. G. A Reversible and Higher-Rate Li-O2 Battery. Science. 2012, 337 (6094), 563–566.

(4) Lu, J.; Li, L.; Park, J.-B.; Sun, Y.-K.; Wu, F.; Amine, K. Aprotic and Aqueous Li-O2 Batteries. Chem. Rev. 2014, 114 (11), 5611–5640.

(5) Luntz, A. C.; McCloskey, B. D. Nonaqueous Li-Air Batteries: A Status Report. Chem. Rev. 2014, 114 (23), 11721–11750.

(6) Hartmann, P.; Bender, C. L.; Vračar, M.; Dürr, A. K.; Garsuch, A.; Janek, J.; Adelhelm, P. A Rechargeable Room-Temperature Sodium Superoxide (NaO2) Battery. Nat. Mater. 2013, 12 (3), 228–232.

(7) Ren, X.; Wu, Y. A Low-Overpotential Potassium–Oxygen Battery Based on Potassium Superoxide. J. Am. Chem. Soc. 2013, 135 (8), 2923–2926.

(8) Xu, S.; Das, S. K.; Archer, L. A. The Li–CO2 Battery: A Novel Method for CO2 Capture and Utilization. RSC Adv. 2013, 3 (18), 6656–6660.

18

(9) Takechi, K.; Shiga, T.; Asaoka, T. A Li–O2/CO2 Battery. Chem. Commun. 2011, 47 (12), 3463–3465.

(10) Hu, X.; Li, Z.; Zhao, Y.; Sun, J.; Zhao, Q.; Wang, J.; Tao, Z.; Chen, J. Quasi–solid State

Rechargeable Na-CO2 Batteries with Reduced Graphene Oxide Na Anodes. Sci. Adv. 2017, 3 (2), e1602396.

(11) Lim, H.-D.; Park, H.; Kim, H.; Kim, J.; Lee, B.; Bae, Y.; Gwon, H.; Kang, K. A New

Perspective on Li-SO 2 Batteries for Rechargeable Systems. Angew. Chemie Int. Ed. 2015, 54 (33), 9663–9667.

(12) Li, Y.; Khurram, A.; Gallant, B. M. A High-Capacity Lithium–Gas Battery Based on Sulfur Fluoride Conversion. J. Phys. Chem. C 2018, 122 (13), 7128–7138.

(13) Golja, B.; Barkanic, J. A.; Hoff, A. A Review of Nitrogen Trifluoride for Dry Etching in Microelectronics Processing. Microelectronics J. 1985, 16 (1), 5–21.

(14) Hoffman, C. J.; Neville, R. G. Nitrogen Fluorides and Their Organic Derivatives. Chem. Rev. 1962, 62 (1), 1–18.

(15) Evans, P. J.; Tschuikow-Roux, E. Thermal Decomposition of Nitrogen Trifluoride in Shock Waves. J. Chem. Phys. 1976, 65 (10), 4202–4209.

(16) Vileno, E.; LeClair, M. K.; Suib, S. L.; Cutlip, M. B.; Galasso, F. S.; Hardwick, S. J. Thermal Decomposition of NF3 by Ti, Si, and Sn Powders. Chem. Mater. 1995, 7 (4), 683–687.

(17) Yamada, Y.; Furukawa, K.; Sodeyama, K.; Kikuchi, K.; Yaegashi, M.; Tateyama, Y.; Yamada, A. Unusual Stability of Acetonitrile-Based Superconcentrated Electrolytes for Fast-Charging Lithium-Ion Batteries. J. Am. Chem. Soc. 2014, 136 (13), 5039–5046.

(18) Dean, C. R. S.; Finch, A.; Gardner, P. J. The Aqueous Solubilities of Nitrogen Trifluoride and Dinitrogen Tetrafluoride. J. Chem. Soc. Dalt. Trans. 1973, 0 (23), 2722–2725.

(19) Knudsen, K. B.; Nichols, J. E.; Vegge, T.; Luntz, A. C.; McCloskey, B. D.; Hjelm, J. An Electrochemical Impedance Study of the Capacity Limitations in Na-O 2 Cells. J. Phys. Chem. C 2016, 120 (20), 10799–10805.

(20) Adams, B. D.; Carino, E. V.; Connell, J. G.; Han, K. S.; Cao, R.; Chen, J.; Zheng, J.; Li, Q.; Mueller, K. T.; Henderson, W. A.; et al. Long Term Stability of Li-S Batteries Using High Concentration Lithium Nitrate Electrolytes. Nano Energy 2017, 40, 607–617.

(21) Liu, B.; Xu, W.; Zheng, J.; Yan, P.; Walter, E. D.; Isern, N.; Bowden, M. E.; Engelhard, M. H.; Kim, S. T.; Read, J.; et al. Temperature Dependence of the Oxygen Reduction Mechanism in Nonaqueous Li–O 2 Batteries. ACS Energy Lett. 2017, 2 (11), 2525–2530.

19

(22) Luntz, A. C.; Viswanathan, V.; Voss, J.; Varley, J. B.; Nørskov, J. K.; Scheffler, R.; Speidel, A. Tunneling and Polaron Charge Transport through Li 2 O 2 in Li–O 2 Batteries. J. Phys. Chem. Lett. 2013, 4 (20), 3494–3499.

(23) Nasybulin, E.; Xu, W.; Engelhard, M. H.; Nie, Z.; Burton, S. D.; Cosimbescu, L.; Gross,

M. E.; Zhang, J.-G. Effects of Electrolyte Salts on the Performance of Li–O 2 Batteries. J. Phys. Chem. C 2013, 117 (6), 2635–2645.

(24) Bard, A. J.; Faulkner, L. R. Electrochemical Methods : Fundamentals and Applications; Wiley, 2001.

(25) Savéant, J. Single Electron Transfer at an Electrode. In Elements of Molecular and Biomolecular Electrochemistry; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2006; pp 1–77.

(26) Savéant, J. Electron Transfer, Bond Breaking and Bond Formation. In Elements of Molecular and Biomolecular Electrochemistry; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2006; pp 182–250.

(27) Pause, L.; Robert, M.; Savéant, J.-M. Reductive Cleavage of Carbon Tetrachloride in a Polar Solvent. An Example of a Dissociative Electron Transfer with Significant Attractive Interaction between the Caged Product Fragments. J. Am. Chem. Soc. 2000, 122 (40), 9829–9835.

(28) Costentin, C.; Robert, M.; Savéant, J.-M. Successive Removal of Chloride Ions from Organic Polychloride Pollutants. Mechanisms of Reductive Electrochemical Elimination in Aliphatic Gem-Polychlorides, α,β-Polychloroalkenes, and α,β-Polychloroalkanes in Mildly Protic Medium. J. Am. Chem. Soc. 2003, 125 (35), 10729–10739.

(29) Rosokha, S. V.; Lukacs, E.; Ritzert, J. T.; Wasilewski, A. Mechanism and Thermodynamics of Reductive Cleavage of Carbon–Halogen Bonds in the Polybrominated Aliphatic Electrophiles. J. Phys. Chem. A 2016, 120 (10), 1706–1715.

(30) Laoire, C. O.; Mukerjee, S.; Abraham, K. M.; Plichta, E. J.; Hendrickson, M. A. Elucidating the Mechanism of Oxygen Reduction for Lithium-Air Battery Applications. J. Phys. Chem. C 2009, 113 (46), 20127–20134.

(31) Lamy, E.; Nadjo, L.; Savéant, J. M. Standard Potential and Kinetic Parameters of the Electrochemical Reduction of Carbon Dioxide in Dimethylformamide. J. Electroanal. Chem. Interfacial Electrochem. 1977, 78 (2), 403–407.

(32) Rosen, B. A.; Salehi-Khojin, A.; Thorson, M. R.; Zhu, W.; Whipple, D. T.; Kenis, P. J. A.; Masel, R. I. Ionic Liquid-Mediated Selective Conversion of CO₂ to CO at Low Overpotentials. Science. 2011, 334 (6056), 643–644.

20

(33) Johnson, F. A.; Colburn, C. B. The Tetrafluorohydrazine-Difluoroamino Radical Equilibrium. J. Am. Chem. Soc. 1961, 83 (14), 3043–3047.

(34) Modica, A. P.; Hornig, D. F. Kinetics of the Thermal Dissociation of N2F4 in Shock Waves. J. Chem. Phys. 1968, 49 (2), 629–637.

(35) Tsckuikow-Roux, E.; MacFadden, K. O.; Jung, K. H.; Armstrong, D. A. Kinetics of the Thermal Dissociation of Tetrafluorohydrazine. J . Phys. Chem 1973, 77 (6), 734–742.

(36) Zhang, L.; Chen, G.; Berg, E. J.; Tarascon, J.-M. Triggering the In Situ Electrochemical Formation of High Capacity Cathode Material from MnO. Adv. Energy Mater. 2017, 7 (8), 1602200.

(37) Jung, S. K.; Kim, H.; Cho, M. G.; Cho, S. P.; Lee, B.; Kim, H.; Park, Y. U.; Hong, J.; Park, K. Y.; Yoon, G.; et al. Lithium-Free Transition Metal Monoxides for Positive Electrodes in Lithium-Ion Batteries. Nat. Energy 2017, 2 (2), 4–12.

21