Characterisation of Deuorination Reaction and Solid Electrolyte Interphase in Fluoride-Shuttle Battery Anode MgF2

Fluoride-shuttle batteries using uoride ion transfer have been extensively investigated for producing post-lithium-ion batteries. One of the key issues hampering the development is the low capacity utilisation rate in anodes, causing a signicant reduction in battery performance. To improve the utilisation rate, it is necessary to clarify the unexplained parts regarding the (de)uorination behaviour to optimise the electrode design. Here, we demonstrated the characterisation of Mg metal formations and the solid electrolyte interphase (SEI) in deuorinated MgF 2 anode. Mg was mainly formed in the region with electron and F ion conductivities, implying that these conductivity paths must be eciently increased to further improve the utilisation rate. Nanosized Mg metals were formed even in the region with poor electron conductivity, implying that an ecient (de)uorination process could be achieved by designing the electrode conguration or/and elemental composition. The inuence of SEI in battery performance have currently been neglected because the problems of low utilisation rate are more serious. However, as research progresses, the control of the SEI composition and its properties should be an important investigation to further improve uoride-shuttle battery performances.


Introduction
Fluoride-shuttle battery, which utilises uoride ion (F -) transfer between electrodes, is one of the candidates as a next generation battery due to its potential to achieve higher energy density than the lithium-ion battery 1,2 . The concept of this battery was reported in 1970s 3,4 . Reddy and Fichtner in 2011 2 demonstrated the rst reversible cycling performance in an all-solid-state uoride-shuttle battery even at elevated temperature, and it has attracted signi cant attention ever since [5][6][7][8][9][10][11][12] . There are many types of metal/metal uorides, thus making it possible to achieve a high voltage cell by exploring the suitable combinations of cathode and anode materials with an appropriate uoride ion conductivity electrolyte.
Then, in the de uorination process based on conversion reaction of metal/metal uorides in the electrodes, multiple electron reactions can be promoted by moving single charged anions of Fin electrolyte. Therefore, theoretical energy density of uoride-shuttle battery can be expected to be up to 5000 Wh/L 1 .
However, uoride-shuttle batteries require extensive research and development to improve their electrochemical performance for practical applications. One of the critical problems of the battery is its low-capacity utilisation rate or low utilisation of active materials 2,[13][14][15] . The most critical problem lies with the construction of the anode. This is because typical anode materials such as AlF 3 , CaF 2 , LaF 3 and MgF 2 have wide band gaps (E g) of ~10.8 eV 16 , ~11.8 eV 17,18 , ~10.5 eV 19 , and ~11.8 eV 20 , respectively.
The band gaps of the anodes are over twice as large as typical cathode materials, for example, ~2 eV of CuF 2 21,22 and ~4 eV of BiF 3 23,24 . The large band gap of anode materials results in poor electronic conductivity, making it di cult for F ions to exchange electrons at the interfaces between a metal and a metal uoride inside active anodic materials during the (de) uorination process. Since the formation enthalpy of a metal uoride for an anode is essentially much lower than that of a metal 25 , removal of F ions from the metal uorides of an active anode is more di cult compared to that of an active cathode. A nearly complete utilisation in the rst discharge capacity (de uorination) has been reported in the case of a CuF 2 cathode using a liquid electrolyte 8 and a BiF 3 cathode using a solid electrolyte 7 at room temperature. Meanwhile, regardless of various efforts such as improving conductivity 14 and a composited electrode (Mg-MgF 2 ) 15,26 , the capacity utilisation rate of the active anode materials are currently very low. Therefore, the active anode material needs to be present in excess to the cathode active materials for electrochemical testing 2,5,8,14,15,26,27 . In other words, the enhancement of the capacity utilisation rate in an anode can directly lead to the improvement of the energy density and the performance of uoride-shuttle batteries. To improve the low utilisation rate of anode materials, it is necessary to understand the mechanisms of (de) uorination reaction and optimise the electrode design.
However, there is a lack of information regarding the (de) uorination behaviour in anodes. Moreover, useful analyses of anode materials have not yet been established due to the nanoscale dimensions of particles and the low amount of reaction from the material.
In this study, imaging techniques using scanning transmission electron microscopy (STEM) and electron energy loss spectroscopy (EELS) were developed to analyse the MgF 2 anode, which is a promising anode material owing to its high theoretical capacity of 860 mAhg -1 , before/after electrochemical de uorination. Advanced STEM EELS imaging technique successfully visualised the distribution of Mg metal in the de uorinated MgF 2 anode. Furthermore, the interphase, i.e., solid electrolyte interphase (SEI) in the case of lithium-ion battery anode 28 , was also con rmed in de uorinated MgF 2 , on the surface or the interface between the electrolyte and the anode.

Results
Electrochemical measurement of MgF 2 anodes. Figure 1 shows the charge/discharge curves of the MgF 2 anode at room temperature in a half-cell test. The MgF 2 anode shows a capacity of 493 mAhg -1 in the charge (de uorination) process. This capacity is 57 % of the MgF 2 theoretical speci c capacity, 860 mAhg -1 . After charge, the discharge ( uorination) capacity is 271 mAhg -1 . Coulombic e ciency of the rst cycle is around 55 %. This performance is not enough for practical use but is more favourable than previous reports 14,15,26 .
Previously, most uoride-shuttle batteries have been investigated using solid electrolytes. In such systems, degradation of cycling performance is mainly attributed to the loss of contact between an electrolyte and the active material. The interface dissociation is caused by the large volume change of active materials during (de) uorination. MgF 2 , used as the active anode material in the present study, exhibits a large volume shrinkage (-71 %) from MgF 2 (V = 65.18 Å 3 ) to Mg (V = 46.46 Å 3 )). In contrast, liquid electrolytes proposed for uoride-shuttle batteries 8,[29][30][31][32][33][34][35] can maintain contact with active materials. Therefore, liquid electrolytes are considered to be one of the main reasons why the Coulombic e ciency reported in this study is better than those of previous reports using solid electrolytes 14,15,26 .
In addition, the electrochemical performance of the present MgF 2 anode can be also explained by the con guration of additive Fe particles and MgF 2 . The X-ray diffraction (XRD) pattern obtained from the pristine anode shows only the re ections from Fe particles (see Fig. S1). Re ections from MgF 2 do not appear because of their ne particle size, approximately several nanometers, as a result of mechanical milling process. The scanning electron microscopy (SEM) images (see Fig. S2) show that the nanosized MgF 2 covers the Fe particles. The exposed surface of the Fe particles and the existence of ne Fe particles in the MgF 2 layer are preferable microstructures from the viewpoint of improving electron conductivity. However, the utilisation of MgF 2 anode after electrochemical de uorination was still at 57 %.
This result indicates that the morphology of the composite anode in this study was not su cient to enhance the exchange reaction of F ions between the anode and the liquid electrolyte.
Characterisation of MgF 2 anodes before/after de uorination. Compositional analysis using SEM and energy dispersive X-ray spectroscopy (EDS) was carried out to observe Mg metal distribution in the de uorinated MgF 2 anode. However, we did not observe clear changes in the MgF 2 active materials before/after de uorination (Fig. S3). The results indicate that the size of Mg metal formed by electrochemical de uorination is too small to be detected due to the overlapping of Mg metal and MgF 2 .
Therefore, in order to clarify the distribution of Mg metal in de uorinated MgF 2 anode, STEM observation and EELS analysis were performed. . This result indicates that the pristine MgF 2 is not amorphous but crystalline.
After de uorination, it can be con rmed that a lot of voids are formed in MgF 2 anode as shown in Figs 2d and 2e. These voids are thought to result from volume shrinkage associated with de uorination from MgF 2 to Mg.On the other hand, the electron diffraction pattern of the de uorinated MgF 2 (Fig. 2f) indicates the similar pattern to the pristine MgF 2 (Fig. 2c) although each re ection was slightly broader.
Then, it is di cult to con rm the re ection from Mg metal in the electron diffraction pattern of Fig. 2f. Most of MgF 2 was not de uorinated and the partially formed Mg metal particles are supposed to be very small and/or unevenly distributed in the anode.
Visualisation of Mg metal distribution in MgF 2 anodes. EELS analysis was carried out to visualise the distribution of Mg metal formed after de uorination in the anode. It should be noted that MgF 2 is prone to release uorine when triggered by electron beam radiations 38 . Therefore, it is necessary to minimise the exposure time and incident beam intensity during recording EEL spectra to avoid the formation of Mg caused by electron beam. In order to overcome this problem, we examined the low energy loss regions below 100 eV, i.e. low-loss EEL spectra with plasmon excitation. Recording of low-loss EEL spectra can be achieved faster than that of core-loss EEL spectra up to 300 eV. This analysis method allows us to reduce beam damage during data recording while recording a better signal-to-noise ratio 39,40 . Figure 3a shows the low-loss EEL spectra including Mg L 2,3 edges obtained from reference materials, i.e.
Mg metal sheet and MgF 2 powders, and MgF 2 anodes after/before de uorination. The low-loss EEL spectrum of Fe is obtained from a Fe particle in the pristine MgF 2 anode. We con rmed that the EEL spectra of Fe particles in MgF 2 anode before/after electrochemical de uorination showed no obvious change. This result means that Fe did not contribute to the charge capacity. These reference spectra can be used for ngerprinting methods to understand the structural change of MgF 2 anode caused by electrochemical reduction processes.
The shapes of plasmon peaks of MgF 2 powder, pristine MgF 2 anode, and de uorinated MgF 2 anode (region A) are almost the same. This result indicates that most of the active MgF 2 materials are not de uorinated. This is consistent with the results of electron diffraction analysis (Figs. 2c and 2f). On the other hand, a change in the plasmon peak, indicated by the red arrow in Fig. 3a, that was detected at region B or a speci c region, is deliberately extracted from low-loss EEL map data. A similar sharp peak located at an energy loss of 10.5 eV is also obtained from the Mg metal sheet as a reference. The metals with free electrons such as Li, Na, Mg, and Al are well known to show a sharp plasmon peak 41 . This is because there is no signi cant in uence of attenuation of plasma oscillations, which produce a broader plasmon peak caused by interband transitions of the valance electrons in semiconductors and insulators. The sharp peak denoted by the red arrow is assumed to have originated from Mg metal formation after de uorination.
Considering the feature of plasmon peaks, it is possible to visualise Mg metal location by extracting sharp plasmon peaks originating from Mg metal. First, the integration intensity I α with an energy window of 0.7 eV from 10.2 to 10.9 eV as a standard spectrum intensity was calculated from a region without Mg formation in de uorinated MgF 2 anode as shown in Fig. 3b. Next, the integration intensity I β with an energy window of 0.7 eV from 10.2 to 10.9 eV was calculated as the intensity of a target signal (Fig. 3c).
Finally, a relative plasmon intensity I rp was de ned as I rp = I β I α for the plasmon maps. The relative plasmon intensity I rp is larger than 1 in the region with Mg metal as shown in analysis results indicate that nanosized Mg metal below 10 nm size are formed by reduction processes even in poor electronic conductivity region, indicated by the white arrows in Fig. 5a and 5b.
The I rp map shows information regarding interphases on the surface of a de uorinated MgF 2 anode. The surface of the MgF 2 anode can be determined from the deep navy color (I rp ≈ 1) region, indicated by a dotted line in Fig. 5b. Further, in the HAADF STEM image (Fig. 5a), there exist contrasts, indicating the formation of different compounds on the surface of MgF 2 anode; these are identi ed by the dashed lines in Figs. 5a and 5b. Therefore, the surface of the de uorinated MgF 2 anode consists of some compounds that are different from MgF 2 and Mg metal. To identify the compounds by using the ngerprint method, low-loss EEL spectra are obtained from reference materials, i.e., Mg metal, Mg(OH) 2 , MgO, MgF 2 ,and LiF. Figure 5e shows low-loss EEL spectra of reference materials and spectrum extracted from the rectangle region α. The shape of the plasmon peak and peak top, indicated by the dotted-lines, obtained from the region α, are similar to that of MgF 2 . However, the energy loss positions where the spectra start to rise, which is indicated by the dashed lines, are different, although the shape of the spectra obtained from the region α is broader due to the existence of various unknown compounds. It should be noted that the band gap can be calculated from an energy position where a valence EEL spectrum rises 42 . The band gaps of MgF 2 and LiF calculated from each valence EEL spectrum were 11.2 eV and 12.4 eV, and are in relatively good agreement with previous reports 20, 43 . Figure 5f shows Mg L 2,3 and Li K edges obtained from reference materials. Compered to these EEL spectra of the reference materials and previous reports of lithium compounds 44 , one of the compounds in region α and on the surface of the de uorinated MgF 2 anode is considered to be LiF. STEM EDS analysis con rmed that a certain kind of oxide was also formed on the surface as shown in Supplementary Note 4 and Fig. S5. LiF and the oxide are considered to be formed by the decomposition reaction of the liquid electrolyte during the reduction process. Such production is well known as solid electrolyte interphase (SEI) in the eld of lithium-ion batteries 28 . SEI formed on the anode materials acts as a passive agent that prevents continuous degradation of a liquid electrolyte. On the other side, SEI may also cause deterioration of battery performance, i.e. fading (dis)charge capacities, rate, and cycling performances.
LiF is one of the typical SEI compounds in lithium-ion batteries 45 . LiF formation as a SEI on the anode materials may lead to a serious problem for uoride-shuttle batteries using a liquid electrolyte comprising of Li salts as LiF has poor electronic and uoride ion conductivity.

Discussion
To further understand the de uorination reaction in the MgF 2 anode, the ratio of the Mg metal formation region was estimated by counting the number of pixels with I rp > 1.5 as Mg metal regions from Since the liquid electrolyte is used for electrochemical de uorination in the present study, the de uorination is considered to proceed uniformly. Therefore, we assume that the dependence of the observation area should be low.
The main possible reasons for the formation of the lower amount of formed Mg metal are considered to be the following factors; the observed (dis)charge capacity may have increased owing to side reactions such as electrolyte decomposition. Additionally, SEI has been con rmed to form on surfaces of the MgF 2 anode. The detail in uence of side reductions on the (dis)charge capacity is not yet clari ed in the present results. Further, the Mg metal may have been uorinated by F ion in the electrolyte after de uorination, i.e. self-discharge of anode. This is because the formation enthalpy of MgF 2 is lower than that of Mg metal, indicating that MgF 2 may be easily formed in an F ion liquid electrolyte 25 . The other factor is formation of Mg metal particles with sizes below 1 nm and/or formation of non-stoichiometric MgF 2-δ with a uoride ion de ciency. Especially, F-de cient MgF 2-δ may existunder speci c environments such as particle surfaces or interfaces between Mg metal and MgF 2 although the formation enthalpy of MgF 2 suggests that the de uorination of MgF 2 basically occurs by a two-phase reaction 25 . The information of Mg metal with a size less than 1 nm and F de cient MgF 2-δ cannot be detected in the present method. Because the plasmon peak signal of Mg metal with smaller size is too weak to be buried in the large signal of MgF 2 . Then, the plasmon peak of MgF 2-δ is expected to be similar that of MgF 2 , indicating that it is di cult to extract the information from the present method. Here, we do not have any direct evidence showing nucleation of sub-nanometer sized Mg and/or MgF 2-δ . However, we do speculate that uniform generation of voids after de uorination of the anode may be originating from the volume shrinkage associated with the formation of sub-nanometer-sized Mg metals. To con rm our assumption, it is necessary to further develop the imaging and analysis techniques to avoid electron beam damages.
Although the de uorination state in MgF 2 anode has not yet been fully elucidated and still faces the remaining challenges, this study gives the first result for visualisation of Mg metal distributions and is appropriate for demonstrating the existence of SEI on the surface.

Conclusions
We demonstrated the Mg metal distributions and SEI formation on the surface in MgF 2 anode before and after electrochemical de uorination using an advanced STEM EELS technique. Although HAADF STEM imaging techniques do not visualise Mg metal formation, the analysis from low-loss EEL spectra succeeded in visualising that Mg metals in tens of nanometre sizes were partially formed near the additive Fe and surfaces of MgF 2 where electronic or uoride ion conductivities were su ciently maintained. These results indicate that creating electrodes capable of su cient electronic and uorideionic conductions is important for improving the utilisation rate. On the other hand, several nanometresized Mg particles were nucleated even in the regions where electronic conductivity was poor. This implies that e cient de uorination process can be achieved by designing the electrode con guration and/or elemental composition even without a large amount of conductive material.
Furthermore, the SEI composed of LiF and some oxides was formed on the interface between electrolyte and anode after de uorination. At present, the SEI on an anode may be related to degradation of rst (dis)charge capacity and cycle performance. The impact of SEI is considered to be small compared to the anode material utilisation issues. However, as research and development progress in the future, the control of the SEI composition and its properties must become an important challenge in uoride-shuttle batteries to improve battery performances. The ndings in this study do not only offer rst insights into the formation state of Mg metal and SEI in de uorinated MgF 2 anode, but also have implications for the development of anode materials.

Experimental Section
Sample preparation All sample preparations were carried out in an argon lled glovebox under low oxygen concentration below 0.3 ppm. To optimise the poor electron conductivity of MgF 2 , a composite electrode of MgF 2 and Fe was prepared. The oxidation and reduction of Fe can be ignored under present charge/discharge process of potential, and Fe can be considered to be as a conductive agent. MgF 2 and Fe powder were mixed at a weight ratio of 1:10 by mechanical milling at 600 rpm for 3 h. As a result of the mechanical milling, the particles of MgF 2 reduce to several nanometers in size. A working electrode or anode was prepared by cast lm process with a polyimide binder. A composite electrode of acetylene black (AB) and polytetra uoroethylene (PTFE) was used as a counter electrode for electrochemical measurements.
The charge/discharge measurement was performed by the constant current method using a multipotentiostat (Biologic VMP-300) at room temperature. The cut-off voltages were set to -2.85 V and -1.8 V. The current density of charge and discharge were -34.4 mAg -1 (0.04 C rate) and 17.2 mAg -1 (0.02 C rate), respectively. The speci c capacity was calculated by a weight of MgF 2 .

Structural analysis of MgF 2 anode
Cross-sectional STEM samples were prepared using a dual-beam focused ion beam scanning microscope (NB5000, Hitachi High-Technologies Co.) equipped with a Ga ion source. The thinning processes below 2 μm thickness by FIB were performed using a cold stage at -90 °C to reduce the Ga ion beam damages and to avoid the oxidation of the sample caused by ice (H 2 O). Ice adheres to the sample surface due to the degree of vacuum of our FIB instrument when the temperature is lower than -100 °C. Samples were transferred from the argon (Ar) lled glovebox to the FIB using a non-exposure transfer system. Reference samples were a cut sheet of Mg metal and powder samples of Mg(OH) 2 , MgO, MgF 2 , and LiF supported on holey carbon lms prepared in an Ar-lled glove box.
The structure of the MgF 2 anodes before and after electrochemical de uorination was investigated by   Low-loss EEL spectra and integration region for visualisation of Mg metal. a Low-loss EEL spectra of Mg metal sheet and original MgF2 powders as reference spectra, indicated by black color lines, Fe particle and MgF2 anode before/after de uorination. The EEL spectra of region A and B are obtained from de uorinated MgF2 anode. The red arrow in a indicates appearance of the plasmon peak originated from Mg metal. The integration energy region for the sharp plasmon peak of Mg metal with a width of 0.7 eV in the low-loss EEL spectra of b the standard spectrum (a region without Mg metal) and c the target spectrum (a region with Mg metal).