Reversible Electrochemical Intercalation and Deintercalation of Fluoride Ions into Host Lattices with Schafarzikite‐Type Structure

Abstract Herein, we report the successful electrochemical fluorination and defluorination of schafarzikite‐type compounds with the composition Fe0.5 m 0.5Sb2O4 (M=Mg or Co). We show that electrochemical methods can present a more controllable and less environmentally damaging route for fluorinating compounds in contrast to traditional methods that involve heating samples in F2‐rich atmospheres. The reactivity of the host lattices with fluoride during electrochemical fluorination makes this material an interesting candidate for fluoride‐ion battery applications. However, deleterious side reactions with the conductive carbon matrix during fluorination suggests to the contrary. Regardless of the side reactions, the schafarzikite structure was found to be an alternative reversible host lattice for fluoride incorporation and removal in addition to the previously reported Ruddlesden–Popper‐type compounds.


Introduction
Fluorinei nsertion into metal oxides has become an interesting topic over the past years, owing to the potentialf or modifying the electronic, [1] magnetic, [2] ands uperconducting behavior [3] of host latticest hrough structurala nd compositional changes. Furthermore, such oxide materials are considered reversible electrode materials for fluoride-ion batteries (whichw ere previously based on conversion-type compounds), [4] for which currently only compounds with Ruddlesden-Popper-type structure are known to show principle structural reversibility. [5] Chemical fluorination of the oxides has predominantly been performed via chemical reactions, for example, by heating samples under flowing F 2 gas [6] or with milder fluorination agents such as PVDF. [7] The use of oxidative agents (F 2 ,C uF 2 , AgF 2 )i sc hallenging, and can often lead to the decomposition of the target compounds. [8] The reason fort his originates from the fact that such reagents alwaysw ork at ac ertain chemical fluorination potential, which can only be altered by the choice of the metal fluoride. Especially for fluorine gas, the reactioni s then mainly controlled by experimental parameters such as temperature, time, and fluorine concentration. [6a] In recent reports,o ur group has shown that the electrochemical fluorination of compounds (e.g. LaSrMnO 4 ,L a 2 CoO 4 ,a nd BaFeO 2.5 ) within an all-solid-state fluoride-ion battery can serve as an alternative method [8,9] for the preparation of oxyfluorides, where the degree of fluorination can be adjusted through tuning by choosing suitable electrochemical potentials and charging times.
The schafarzikite-types tructure (see Figure 1) of compounds with the compositionM Sb 2 O 4 [10] (known for their antiferromagnetic properties with variousd ifferent magnetic structures) [11] possesses at etragonal symmetry (space group P4 2 /mbc). The structure can be understood as being built up of chains of edge-linked MO 6 octahedra runninga long the [0 01]d irection; Herein, we report the successful electrochemical fluorination and defluorination of schafarzikite-type compounds with the composition Fe 0.5 M 0.5 Sb 2 O 4 (M = Mg or Co). We show that electrochemical methods can presentamore controllable and less environmentally damaging route for fluorinating compounds in contrast to traditional methods that involve heating samples in F 2 -rich atmospheres. The reactivity of the host latticesw ith fluorideduring electrochemical fluorination makest his material an interesting candidatef or fluoride-ion battery applications. However,d eleterious side reactions with the conductive carbon matrix during fluorination suggestst ot he contrary.R egardless of the sider eactions, the schafarzikite structure was found to be an alternative reversible host lattice for fluoride incorporation and removal in addition to the previously reported Ruddlesden-Popper-typec ompounds.
[a] M. A. Nowroozi KGaA. This is an openaccessarticleunder the termsoft he Creative Commons AttributionL icense, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. the chains are connected through trigonal pyramidal SbO 3 units. Recent studies have shown that it is possible to fluorinate variantso ft his material (see Figure 1), which contain Fe 2 + on the Ms ite, by using topochemical reactions. [6b, 10,12] The proposed mechanism forf luorination is based on two key principles. [6b] Firstly,t he phase must possess Fe 2 + to act as the redox active centerw hilst the degree of oxidation is limitedt ot he amount of Fe 2 + to be oxidized to Fe 3 + .F urthermore, it has been shownt hat there is ap ropensity for the Sb 3 + ,w hich line the walls of the channel, to also play ap art in the oxidation process depending on the atmosphere and conditions that the materialish eatedi n. [13] This material is of interest because of the mechanism for the inclusiono fe xcessf luoride ions within the channel of the structure( see Figure1). Therefore, it can be consideredt ob ea 1D intercalation material, like olivine-type materials for lithiumion batteries. [14] This is in contrast to the layered ordering of intercalatedf luoridei ons within Ruddlesden-Popper-type compounds, whicha re 2D intercalation materials (like layered materials,e .g.,L iCoO 2 for lithium-ion batteries). [15] For both structure types, intercalated fluoridei ons were found to be located on ad ifferent crystallographic site than the oxide ions, and such ordering of the intercalated ions is ak ey feature of intercalation based battery materials.
Here, we build upon the previouss tudy,w hich explored the chemicalf luorination behavior of schafarzikite-type Mg 0.5 Fe 0.5 Sb 2 O 4 and Co 0.5 Fe 0.5 Sb 2 O 4 using gaseous fluorine to form Mg 0.5 Fe 0.5 Sb 2 O 4 Fa nd Co 0.5 Fe 0.5 Sb 2 O 4 F x (where x % 0.5). In this article, we investigate their suitability for electrochemical applicationsw ithin all-solid-state fluoride-ionb atteries. The inclusion of 0.5 Fp er formula unit corresponds to the specific chargingc apacityo fr oughly3 6-39mAhg À1 . [6b] We show that this class of material is found to be the second suitable host materialf or the fully reversible intercalation/deintercalation of fluoridei ons. However,h igh chargingp otentials weref ound to currently impede their use as intercalation-based cathodes for fluoride-ion batteries when carboni su sed as the conductive additive.

Results and Discussion
The lattice parameters of the fluoridated samples were obtained from Rietveld analysiso ft he XRD data [9b] and can be compared to the parent materiala nd the fluoridatedc ompounds reported previously. [10] It is necessary to confirm that changes of lattice parameters, which were observed after the charging/discharging of the samples,r eally resulted from an electrochemical reaction. To verify this, we also investigated fully assembled cells, which were only heated to the battery operation temperature withouta pplying any current. From this, one can rule out unwanteds ider eactions, which would also result in changes of lattice parameters, for example, an oxide for fluoridesubstitution reaction with the La 0.9 Ba 0.1 F 2.9 admixture according to [ Indeed, no significant changes of lattice parameters were found after heating at 170 8Cf or 24 h, confirming the stability of the La 0.9 Ba 0.1 F 2.9 towards the schafarzikite compounds ( Figure 2a,b and Ta ble 1). This ruled out the possibility for the potential degradation of the parent phase through temperatureinduced non-oxidative substitution reactions. Hence, all structural changesfound on electrochemical treatment can be associated witht he electrochemical charging and discharging reactions of the compounds.
The electrochemical charging curves of Mg 0.5 Fe 0.5 Sb 2 O 4 and Co 0.5 Fe 0.5 Sb 2 O 4 against Pb + PbF 2 are showni nF igure 3. The chargingc urves show three distinct regions:f irst as harp increase up to 1.5 V, followed by ap lateau betweenr oughly1 .6-1.7 V. In the third region,asharp increase in voltage is observed, indicatingt he end of the electrochemical reaction. Within the first region,n or eactiono ft he schafarzikite-type compounds could be identified (also see later in this article), and structural changes were mainly found to occur in region 2.
Greaves and co-workers [6b] reported ac apacity of approximately 0.5 fluoride ions per formula unit Mg 0.5 Fe 0.5 Sb 2 O 4 and Co 0.5 Fe 0.5 Sb 2 O 4 via chemical fluorination, which would correspond to capacities of 36-39 mAh g À1 for the charging/electrochemicaloxidation reaction corresponding to [Eq. (2)]: The lengths of the observed charging plateaus in region 2 exceed this capacity significantly,w hich can be explained from an overlap of the charging plateau with the electrochemical fluorination of the conductive additive of carbon to CÀFs pecies [16] (the amount of carbonadded can contribute to ac harging capacity which is at least 10 times higherthan the absolute capacityo riginating from the amount of the schafarzikite compounds [9a] ). As found previously,t his can impede the discharging (defluorination) of the target compounds, owing to the destruction of the electronic conductivem atrix under formation of CÀFs pecies. [16] This would prohibit their use as cathode materials for reversible fluoride-ion batteries [9a] when carbon is used as an additive for achieving electronic conductivity within the active composite. For Co 0.5 Fe 0.5 Sb 2 O 4 ,t he plateau region is significantly longer, which might be explained by ah igherc atalytic activity for the fluorination of carbon.I nc ontrast, the decomposition of the electrolyte at the cathode side can be basically ruled out, as La 0.9 Ba 0.1 F 2.9 is not sensitive towards oxidation, as verifiedb yn os ignificant changes of the cell parameters of the solid electrolyte after charging (see Table S1).
Furthermore, we note that the potentialf or the fluorination of the schafarzikite compounds, which only involves the Fe 2 + /Fe 3 + redox couple, [6b] is highert han Mn 3 + /Mn 4 + and Co 2 + /Co 3 + in LaSrMnO 4 and La 2 CoO 4 ,r espectively (ca. 1.2 Vf or LaSrMnO 4 and 0.9 Vf or La 2 CoO 4 against ac omposite of Pb + PbF 2 at the same condition), [8,9] and this would not be expected intuitively from the electrochemical series. [17] This could either result from an unusually high electrochemical potential of Fe 2 + within this structure type, or from higher overpotentials fort he schafarzikite-type structurea sc ompared to the Ruddlesden-Popper-typec ompounds.T his might be related to  Figure S1.
The change in lattice parameters after electrochemical fluorination can be followed visually from the changes of the reflection positions in Figures 2c and 2d,w here the refined values given in Ta ble 1a re consistent with the changes found by  Greaves and co-workers. [6b] The small difference in the lattice parameters between the chemically fluorinated and electrochemically fluorinated samples (please see Ta ble 1) could arise from slightly different amountso fi ntercalated fluorine within each sample. In our previous article, [8] different cut-off capacities in combination with aq uantitative phase analysiso ft he fluorinated and non-fluorinated phase were used to determine the detailed amount of intercalated fluoridei ons. However, such attempts to investigate the fluorination process intercalation process in more detail by choosing different cut-off capacities didn ot prove to be successful in this study;h ere, we either observed the lattice parameters of the unreacted starting product, or the lattice parameter had changed within errors to the ones obtained after charging to 3V .N on on-fluorinated parentp hase was found in addition to the fluorinated phase, and this is differentt ot he electrochemical charging of La 2 CoO 4 to La 2 CoO 4 F 1.2 ,f or which ac oexistence of both phases can be found in the plateau region. Therefore, the fluorination of schafarzikite-type compounds to compounds with the com-positionM Sb 2 O 4 F x appears to result in single-phase compounds for ab road region of x,w hereas compositions La 2 CoO 4 F x (0 < x < 1.2) appear to result in two-phase mixtures of ( x / 1.2 La 2 CoO 4 F 1.2 + (1.2Àx) / 1.2 La 2 CoO 4 ). We also note that the cell parameters after the fluorination possess ap seudocubic metric [a/(c* p 2)] % 1, see Ta ble1). This ratio is well reproduced regarding the chemical and electrochemicalf luorination processes. We investigated the possibility that ah igherc ubic symmetry could exist for fluorinated compounds by testingp ossible supergroups of the tetragonal schafarzikite-type structure. However,arearrangemento fp olyhedra within the schafarzikite structuret or esult in at hree-fold rotational axis (required for cubic symmetry) does not appear possible. Therefore, no simple group-subgroupr elationships could be identified, which would explain ac hange to cubic symmetry,a nd this is in agreement with previous symmetry analyses. [8] The volumec hanges of the active cathode materials tructures also calculated to be approximately 1. 8  .W e would like to point out that those changes are very low as compared to Ruddlesden-Popper-type compounds,w hich are on the order of 10-20 %. [5,[8][9] Once fully charged, the discharge profiles of the materials were investigated, as shown in Figure4a. The discharge capacities were found to be very low,o nt he ordero f 6.0 mAh g À1 (corresponding to ca. 0.08 F À )a nd 3.0 mAh g À1 (corresponding to ca. 0.04 F À )f or Co 0.5 Fe 0.5 Sb 2 O 4 and Mg 0.5 Fe 0.5 Sb 2 O 4 ,r espectively (see Figure 4a). This observation is similart oo ur previous findings for LaSrMnO 4, [9a] for which the chargingplateau also wasfound to overlap with the decomposition of the carbon matrix. However, on discharging to negative potentials against Pb/PbF 2 ("forcedd ischarging" due to accessingp otentials that would correspond to an endergonic or hindered process, as would be the case for the destruction of carbon on charging), lattice parameters weref ound to change back to close to the valueso bserved for the startingm aterials (see Table 1a nd Figure 4b,c). For Mg 0.5 Fe 0.5 Sb 2 O 4 ,t he difference in lattice parameters compared to the unreacted compound (Da = 0.032 and Dc MFSO = 0.011 )i sb igger (although overall small) than that for Co 0.5 Fe 0.5 Sb 2 O 4 (Da CFSO = 0.002 and Dc CFSO = 0.010 ), which might indicate the presence of residual fluoridei ons within the compound (see Ta ble 1a nd Figure 4d,e). From the shape of the dischargingc urve,o ne can also derive ap rinciple fluorinec ontent in the order of 40-50 mAh/ g, which corresponds to 0.5-0.6 fluoride ions and is in well agreementw ith the fluorinec ontentsf ound for the chemicalfluorination reactions. [6b] These findings showt hat the schafarzikite-type structure allowsf or reversible intercalation/deintercalation of fluoride ions throughe lectrochemical fluorination, making it the second structure type known fort he structurally reversiblei ncorporation of fluoride ions so far.A sw ith the fluorinated Ruddlesden-Popper-typecompounds, [8, 9a] the oxide and fluoride in the fluorinated schafarzikite structure have been shown [6b] to  idity of the proposed models. Furthermore, it wass uggested that the fluoride ions only form bonds to the soft antimony cations (andn ot to the transition metal M) without primarily oxidizingt he Sb 3 + to Sb 5 + .S uch bonding behavior and associated localized structural distortions could lower the activation energy of fluoridei ons for migration through the structure. Compounds with ns 2 cations (such as SnF 2 ,P bF 2 ,a nd SbF 3 ) [18] are known to be good fluoride-ion conductors, owing to the high polarizability of the cations. Therefore,t he local chemical environment of the fluoridei ons in the schafarzikite structure closely resembles the situation found in the binary fluorides of ns 2 metals.Again, this also resembles the scenario found in the Ruddlesden-Popper-types tructure, where fluoride ions only form bonds to the alkaline-earth/rare-earth cations (for which the binary metal fluorides are also good fluoride-ion conductors). [18] Both structuralf eatures (anion ordering and type of MÀFb onds formed) might, therefore, determine ap rerequisite for the selective deintercalation of fluoride ions and full structural reversibility.

Conclusions
In this article, we have shown that the electrochemical fluorination process is applicable to schafarzikite-type compounds Mg 0.5 Fe 0.5 Sb 2 O 4 and Co 0.5 Fe 0.5 Sb 2 O 4 .A nalysis of lattice parameters before electrochemical fluorination and after charging/dischargingr evealed ac lose similarity of products for both reaction routes. This shows that fluorinated schafarzikite compounds can be prepared by using significantly milder,l ess dangerousr eactionc onditions through electrochemical routes. However,i ts hould be taken into considerationt hat the final fluorinated product is mixed with the electrolyte material and carbon additive and, so far,n os eparation strategies were examined to obtain the electrochemical products isolated from the additives;f urthermore, the material is obtained in low quantity compared to what can be obtained by using chemical methods. The voltage plateau of the intercalation process coincides with the decomposition of the conductive additive carbon,w hich currentlym akes the materialabad candidate for battery applicationsu nless otherm ore stable conductive additives can be found (the authors would like to pointo ut that such attempts were made, for example, by using silver, but did not prove to be successful). Regardless of this, the compounds show excellent structuralr eversibility for the fluorine intercalation/deintercalation process, which is most likely facilitated by the ordering of oxide and fluoridei ons in addition to local bondings cenarios aroundt he fluoride ions and their arrangementw ithin 1D channels. In the future, we aim to extend our investigation to other schafarzikite compounds or compounds within the Mullite family. [19] Experimental Section Schafarzikite-type compounds with the composition Mg 0.5 Fe 0.5 Sb 2 O 4 and Co 0.5 Fe 0.5 Sb 2 O 4 have been prepared by using the method described by de Laune et al. [6b] Stoichiometric amounts of ad ried mixture of the metal oxides and antimony metal (CoO, 325 mesh Sigma-Aldrich;F e 2 O 3 , ! 99.9 %Sigma-Aldrich;Sb 2 O 3 ,Reagent Plus, Sigma-Aldrich;S b, BDH;M gO, ! 99 %3 25 mesh Sigma-Aldrich) were heated in evacuated sealed quartz tubes for between 6a nd 36 hat700 8C, with intermittent grinding.
An electrochemically active composite (EAC) was prepared by mixing the Co 0.5 Fe 0.5 Sb 2 O 4 and Mg 0.5 Fe 0.5 Sb 2 O 4 compounds with La 0.9 Ba 0.1 F 2.9 (a fluoride-conducting electrolyte, [20] in accordance with previous studies) [9a] and dried black carbon in aw eight ratio of 30:60:10, respectively.T he mixture was milled for 3h at ar otational speed of 250 rpm (Retsch PM100-CM, for 10 min intervals with 20 min of resting between the intervals). The volume of the milling vial and the diameter of each ball were approximately 244 cm À2 (0.24 L) and 10 mm, respectively.T he ball-to-powder ratio was 30:1 using 10 balls with at otal mass of almost 30 g. All milling processes were performed in ZrO 2 vials, which were filled and sealed inside ah igh-purity Ar-filled (99.999 %) glovebox. Ac omposite of Pb + PbF 2 ,a sp reviously described in Ref. [9a] was used as the counter electrode and the source of fluoride ions. The use of the EAC instead of pure schafarzikite compounds is required, owing to the insufficient fluoride-ion and electronic conductivity of pure schafarzikite at low temperatures, and this is ac ommon procedure for the investigation of electrode compounds. [4a] For electrochemical fluorination/defluorination, af luoride-ion battery setup was used. [9a] Three layers (EAC, La 0.9 Ba 0.1 F 2.9 ,a nd Pb + PbF 2 )w ere compacted to ab attery cell at al oad of 2tons for 90 s over an area of 0.42 cm 2 ,u sing ad esktop press (Specac) and steel die set inside an Ar-filled glovebox. The dimensions of the overall cell were measured to be 1.6 mm thick and 7.3 mm in diameter. Battery cells were spring-loaded (as described in Ref. [9a]) into a modified Swagelok-type cell with current collectors made of stainless steel. The applied charging and discharging currents were chosen to be AE 10 mA( 24 mAcm À2 ). The values of the charging/discharging current are based on our previous experience on the magnitude of overpotentials during the charging/discharging reactions. [9a] To ensure sufficientm obility of the fluoride ions, the electrochemical cells were heated by band heaters and measurements were taken at 170 8C. The temperature of 170 8Cw as chosen, as it facilitates sufficient conductivity of the solid electrolyte, which is required to limit overpotentials arising from the so-called IR drop to below 0.1 Vf or the current densities used in this study (below 24 mAcm À2 ). [20,21] Ap otentiostat (BioLogic SP-150 &V SP300) was used for all of the galvanostatic charging measurements.
Ex situ X-ray diffraction was used to monitor structural changes of the target compounds. The measurements were performed by using aB ruker D8 Advance in Bragg-Brentano geometry and Cu K a radiation (VANTEC detector). To avoid potential side reactions with the atmosphere, all samples were loaded into al ow-background specimen holder (Bruker A100B36/B37) and sealed inside an Arfilled glovebox before every measurement. Data were recorded between 20 and 708 (2q)f or at otal measurement time of 4husing a step size of approximately 0.0078 and af ixed divergence slit of 0.38.A ll analyses of diffraction data were performed by using the Rietveld method in TOPAS V5. [22] The instrumental intensity distribution, that is, the apparative broadening of reflections, was determined empirically from af undamental parameter set by using a reference scan of LaB 6 (NIST 660a). The microstructural parameters (crystallite size and strain broadening) were refined to adjust the peak shapes. Thermal displacement parameters were refined and constrained to be the same for all of atoms of all phases to minimize quantification errors and to account for angular dependent intensity changes induced by absorption and surface roughness.