Ashley P.
Black
ab,
Arturo
Torres
c,
Carlos
Frontera
ab,
M. Rosa
Palacín
ab and
M. Elena
Arroyo-de Dompablo
*c
aInstitut de Ciència de Materials de Barcelona (ICMAB-CSIC) Campus UAB, E-08193 Bellaterra, Catalonia, Spain
bALISTORE-ERI European Research Institute, CNRS FR 3104, Hub de l’Energie, 15 Rue Baudelocque, 80039 Amiens, France
cDepartamento de Química Inorgánica, Facultad de Ciencias Químicas, Universidad Complutense de Madrid, 28040 Madrid, Spain. E-mail: e.arroyo@quim.ucm.es
First published on 12th February 2020
Sustainability combined with high energy density prospects makes Fe-based oxides attractive as cathodes for calcium rechargeable batteries. This work presents a DFT evaluation of the CaFe2+nO4+n (0 < n < 3) family, for which both the average intercalation voltage and the theoretical specific capacity decrease with the increasing n value. The term n = 1/4, Ca4Fe9O17, meets the most appealing characteristics: a calculated average voltage of 4.16 V, a theoretical specific capacity of 230 mA h g−1 and the lowest energy barrier for Ca migration so far predicted for an existing oxide (0.72 eV). To overcome the previously reported synthesis difficulties, we employed a novel synthesis procedure in sealed quartz tubes followed by quenching in water. The XRD and SAED patterns of the prepared Ca4Fe9O17 powder reveal a certain degree of stacking defects along the c axis. Attempts to deinsert Ca ions from Ca4Fe9O17 by chemical means (NO2BF4 in ACN) and in electrochemical Ca cells were unsuccessful, although some hints of oxidation are observed in Li cells with the LP30 electrolyte. The suitability of Ca4Fe9O17 as a Ca cathode is pending further studies utilizing Ca-electrolytes with high anodic stability.
Sustainable cathode materials should include abundant and non-toxic elements, and at the same time, comply with the high energy density requirements. These criteria prompted the investigation of light and abundant oxidizing 3d-transition metal oxides as cathodes for Ca batteries. So far several Mn oxides and polyoxoanionic compounds have been reported, going from the virtual spinel CaMn2O422 to the CaMnO3 perovskite,16 its reduced derivative -Ca2Mn2O5,23 the known CaMn2O4 polymorphs (marokite and its high pressure form),15 a related oxidized phase, CaMn4O8,23 and the minerals K0.31MnO2 0.25H2O (birnessite),24 Mg0.19Na0.07MnO2 0.37H2O (todorokite),25 CaMn(CO3)2 (kutnahorite)26 and CaMnSi2O6 (johannsenite).26 In contrast, research on Fe oxides has been scarce; the perovskite CaFeO3 and the related brownmillerite Ca2Fe2O5 compounds have been only evaluated by computational means.16,23 It was found that Ca-perovskites are not competitive materials due to hindered Ca mobility (a calculated energy barrier of 2 V).16 In the brownmillerite structure the oxygen vacancies respective to the perovskite create a new pathway with much lower energy barrier (1 eV).23 The present investigation turns to other known oxides in the Ca–Fe–O system with general formulae CaFe2+nO4+n that might be worthy candidates (Fig. 1).
Fig. 1 Crystal structures of CaFe2+nO4+n oxides. (a) n = 0, CaFe2O4 (b) n = 2, CaFe4O6 (c) n = 1/4, CaFe9O17 showing the [CaO6]–[FeO4] plane in (d). Color code: Ca in blue, Fe in brown and O in red. |
The crystal structure of CaFe2O4 (S.G. Pmna, no. 53)27,28 consists of edge-sharing chains of MnO6 octahedra, forming double rutile chains (see Fig. 1a). Each double-rutile chain is connected to two adjacent chains through edge-sharing oxygen. As a result, tunnels appear along the b axis where Ca ions reside in eight-fold coordination. The calcium ferrite CaFe2+nO4+n family is formed by stacking along the c axis (S.G. Cmcm, no. 63) of blocks with the CaFe2O4 composition and blocks with the FeO composition (see Fig. 1b). Within this family, the compounds with n = 0, 1, 3/2, 2, 5/2, and 3 have been reported.29,30 Interestingly, for n = 1/4, the compound Ca4Fe9O17 crystallizes in a different structure (S.G. C2, no. 5), built up by layers of Fe atoms in edge-sharing FeO6 octahedra and vertex-sharing Fe in trigonal-based bipyramid coordination (see Fig. 1c and d).31 These hexagonal layers stack along the c-axis direction, being cross-linked by iron atoms in tetrahedral coordination. The [FeO4]-tetrahedra are connected in the ab plane of the structure by hexa-coordinated Ca–O polyhedra (Fig. 1d). The CaO6 distorted octahedra share edges forming rings around the FeO4-tetrahedra. Unsurprisingly, this structure with Fe in three different coordination environments, and in an average oxidation state of 2.888, is thermodynamically quite unstable.32–34 The oxygen partial pressure is critical in the stability of Ca4Fe9O17, and therefore the complex synthesis of the material usually results in a mixture of a low crystalline Ca4Fe9O17 and more stable oxides (CaFe2O4, FeO, and Ca2Fe2O5). Indeed, the crystal structure determination was accomplished using single crystals grown under very specific conditions.31
To evaluate the performance of CaFe2+nO4+n compounds as cathode materials for Ca batteries, we have selected the compounds with n = 0, 1/4, 1 and 2 for computational investigations. Density Functional Theory (DFT) calculations are used to anticipate the average intercalation voltage so as to estimate their theoretical energy density, assuming the full Ca extraction from the structures. Ca diffusion is a key parameter, for which the migration energy barriers provide an approximate indication.3,35 Following vast literature in the field,18,21,22,35 in this work the energy barriers for Ca migration are extracted from DFT calculations. The results found for Ca4Fe9O17 motivated an experimental investigation comprising the synthesis, characterization and electrochemical testing of the material. We present a novel synthesis method to produce highly crystalline powder.
Calcium mobility is investigated using the Nudged Elastic Band method (NEB) as implemented in VASP. To study diffusion, a vacancy is created in the structural model of the stoichiometric oxides and a simple jumping mechanism from an occupied position to the vacant site is considered. Supercells 1 × 2 × 1 (Ca8Fe18O34) and 1 × 4 × 1 (Ca4Fe8O16) for the n = 1/4 and n = 0 compounds, respectively, were used for the calculations. Constant volume calculations were performed within the GGA approximation for three or five intermediate images initialized by linear interpolation between the two fully relaxed end-points. To calculate the energy at the saddle point, cubic splines were fit through the relaxed images along each hop.
Scanning Electron Microscopy (SEM) and Energy Dispersive X-ray Spectroscopy (EDS) analyses were performed by using a FEI Quanta 200 FEG microscope equipped with an EDAX detector (resolution: 132 eV). Selected Area Electron Diffraction (SAED) studies were performed using a JEOL 1210 transmission electron microscope operating at 120 kV and equipped with a Gathan sample holder. The specimens for electron microscopy were prepared by dispersing the powders in dimethyl carbonate and depositing a droplet of this suspension on a carbon coated film supported on a copper grid.
The electrochemical tests were run in two and three electrode Swagelok cells assembled in an Ar-filled dry box, employing Al and stainless steel current collectors for the working and reference (RE)/counter (CE) electrodes, respectively, and a Whatman separator soaked in 0.6 ml of electrolyte solution. Mechanically polished Ca metal discs (obtained by pressing Ca pieces under an inert atmosphere) or Li metal were used as the counter/reference or counter electrodes, respectively.
Lithium hexafluorophosphate solutions in ethylene carbonate (EC) and dimethyl carbonate (DMC) (LP30, 1.0 M LiPF6 in EC/DMC = 50/50 (v/v), Solvionic) and the lithium bis(oxalato)borate salt (LiBoB, Chemetal, Battery grade) in a 1:1 v/v mixture of ethylene carbonate (EC) and propylene carbonate (PC) solvents (Solvionic, 99.9%) were used as electrolytes for the electrochemical tests in Li cells. The electrolyte used in Ca cells was a dry (<40 ppm H2O, Karl-Fischer titration) 0.45 M calcium tetrafluoroborate hydrate (Alfa Aesar) solution in EC/PC = 50/50 (v/v).
The electrochemical tests were performed on a Bio-Logic VMP3 potentiostat. Electrochemical deintercalation/intercalation tests were carried out in galvanostatic mode (i.e. Galvanostatic Cycling with Potential Limitation–GCPL). In order to avoid kinetic limitations the rate was set to C/100. A stabilization time of 5 hours at the OCV was applied before each measurement was carried out at 100 °C. Once tested, the cells were disassembled in an Ar-filled dry box and the electrodes were washed with dimethyl carbonate solvent (DMC, Sigma-Aldrich, ≥99%) and further collected in borosilicate glass capillaries of 0.5 mm diameter for ex situ XRD measurements.
CaFe2+nO4+n → Fe2+nO4+n + Ca | (1) |
The higher the n term, the lower the theoretical specific capacity is. These values are listed in Table 1, together with the average voltage predicted for n = 0, 1/4, 1 and 2. As n increases the Fe ions are in a lower oxidation state, and accordingly Ca deinsertion is predicted at lower voltages. Note that the average voltage depends primarily on the redox couple, but it is also influenced by the crystal structure.39 For instance, the voltages predicted for the Fe3+/Fe4+ couple are 4.16 V in brownmillerite-Ca2Fe2O5 (ref. 23) and 3.87 V in CaFe2O4. Taking the calculated voltage and the theoretical specific capacities into account, theoretical energy densities exceeding 600 mA h g−1 (the state of the art in the Li-ion technology) could be viable for n = 0, 1/4 and 1.
CaFe2+nO4+n composition | n | Fe redox couple | Theoretical specific capacity (mA h g−1) | Calculated average voltage (V) | Theoretical specific energy (mW h g−1) |
---|---|---|---|---|---|
CaFe2O4 | 0 | +3/+4 | 249 | 3.87 | 964 |
Ca4Fe9O17 | 1/4 | +2.88/3.77 | 230 | 4.16 | 957 |
CaFe3O5 | 1 | +2.66/+3.33 | 186 | 3.66 | 680 |
CaFe4O6 | 2 | +2.5/+3 | 149 | 2.37 | 353 |
One of the major concerns in cathode material design for the Ca battery technology is the mobility of Ca2+ ions in inorganic structures. Although other factors can certainly limit the success of Ca battery cathodes, a good cation diffusion, that is to say a low migration energy barrier, is a necessary prerequisite. It has been estimated that for a reasonable cell power rate (discharging time 2 hours) the energy barriers for cation diffusion should be below 0.525 eV in micrometer particles and 0.65 eV in nanosized-particles.3 The calculated energy barrier for CaFe2O4 (1.2 eV, see Fig. 1 in the ESI†) greatly surpasses this tolerable value, and therefore, the ferrites CaFe2+nO4+n built up from CaFe2O4 blocks will also suffer from poor Ca mobility. However, the n = 1/4 term might be a viable option. Fig. 2 shows the pathway for Ca diffusion in Ca4Fe9O17 and the calculated energy barrier. In the two-dimensional diffusion pathway in the [CaO6]–[FeO4] plane (shown in Fig. 1d) the diffusing Ca ions exit from their polyhedra across a triangular face to occupy the intermediate tetrahedral site (see Fig. 2a). The associated energy barrier for in-plane Ca diffusion is 0.72 eV (Fig. 2b), the lower energy ever reported for an existing transition metal oxide. The tetrahedral site represents a local energy minimum in the energy landscape. The similarity in Ca–Fe distances at this intermediate site and at the equilibrium sites (3.22 Å and 3.29 Å, respectively) suggests that the diffusing Ca ion does not experience large electrostatic repulsions with the neighbouring Fe+2.88 cations. Compared to other oxides, the lower repulsive forces between the diffusing Ca2+ and other cations in the structure reduce the migration energy.4 Although the barrier is relatively low for Ca diffusion, it is still too high for fast rate performance. Note that the barriers for Li diffusion in layered, olivine and spinel cathodes are below 0.3 eV.3,35,42,43
Interestingly, a related compound, K2Fe4O7, has been recently reported and suggested as a K-superionic conductor.44 The framework of K2Fe4O7 consists of planes of edge-sharing FeO6 octahedra sandwiched between planes of FeO4 tetrahedra and octahedrally coordinated K ions. The K octahedra form rings around FeO4 tetrahedra, thus providing a 2D conduction plane similar to that in Ca4Fe9O17 (Fig. 1d and 2a). When tested in the K/K2Fe4O7/Pt cell no significant current due to the solid electrolyte decomposition was observed up to 5.0 V. Despite the presence of Fe3+, the authors found that the material is also stable upon reduction down to 0 V.
In view of the discrepancies present in the literature, in this work different synthesis methods were attempted at temperatures ranging from 800 to 1200 °C, including conventional solid state synthesis using CaCO3 or CaO as Ca sources and Fe2O3, Fe3O4 or FeO as iron precursors. The Pechini method was also attempted.45 Most reaction products consisted of a mixture of Fe2O3 and CaFe2O4 in all cases, whatever the reaction times (from 6 to 64 hours), except for the samples treated at 1200 °C for which melting and thermal reduction of iron oxide was observed. Further synthesis attempts were carried out in a tube sealed under vacuum, and the phase Ca4Fe9O17 was observed in small amounts after synthesis at 1000° C for 48 h followed by quenching. Optimization of this reaction protocol, modifying precursors, temperature and time, enabled the synthesis of pure Ca4Fe9O17 using a stoichiometric mixture of Ca2Fe2O5, Fe2O3 and Fe3O4 as precursors, targeting to control the iron oxidation state by tuning the amount of oxygen available in the precursors. However, a pure phase was achieved only after several treatments at 1150 °C for 24 h each in a sealed quartz tube followed by quenching in water (a summary of synthesis results is provided in Fig. S2†).
Attempts to perform a Rietveld refinement of this X-ray diffraction pattern using the structural model reported by B. Malaman et al.31 resulted in unsatisfactory fitting of the model to the experimental pattern, despite all efforts to modify cell parameters, occupation factors, etc. The differences in the observed and calculated integrated intensities of several reflections suggest that the prepared material might present some structural differences with respect to the single crystal studied by B. Malaman (see Fig. S3†).
In order to further characterize the structure of the as prepared Ca4Fe9O17 phase, SAED was performed and micrometric particles were observed, as shown in Fig. 4. The cell parameters obtained by electron diffraction were a = 10.4 Å, b = 6.02 Å and c = 11.4 Å. The ED image along the c zone axis presents the characteristic scattered lines indicative of the presence of stacking defects along the c axis. The occurrence of stacking faults could explain the differences between observed and calculated intensities in Rietveld refinement. The stacking faults are not expected to perturb the in-plane Ca conductivity, unless they imply inter-layer cation mixing or phase transformations.42,46
Fig. 6 Synchrotron X-ray powder diffraction patterns of pristine (A) and electrochemically oxidized (in the LP30 electrolyte) (B) Ca4Fe9O17 samples. |
While the origin of these subtle changes has been unclear to date, one may speculate that they could be related to de-insertion of a small amount of calcium from Ca4Fe9O17 at the 4.0 V plateau (capacity 12 mA h g−1, curve B in Fig. 5b) concomitant with the oxidation of iron. In fact, hypothetical oxidation to yield Ca3.5Fe9O17 could be related to the Fe2+/Fe3+ redox pair, with an expected voltage considerably smaller than 4.16 V, and thus within the voltage window of the LP30 electrolyte stability. Further oxidation should involve the Fe3+/Fe4+ couple, with a higher voltage (ca. 4.2 V), which may be out of the voltage window of the electrolyte stability, especially if polarization is significant. Yet, in light of the low value of the capacity and the disorder present in the crystals, this phenomenon has not been further studied.
Regardless of the partial oxidation of Fe detected in Li cells (likely the Fe2+/Fe3+ couple), further oxidation has not been achieved either electrochemically or chemically. Noteworthily, the calculated average voltage for the full Ca deinsertion, the Fe+2.88/Fe+3.77 couple (4.16 V, see Table 1), is probably beyond the anodic stability of available Ca electrolytes5 and in the limit of Li electrolytes if cell polarization is significant. In this regard, we attempted the electrochemical and chemical oxidation of Ca2Fe2O5, for which DFT investigations found an average voltage of 4.16 V for the Fe3+/Fe4+ couple, although with a sluggish kinetics (calculated energy migration barrier 1 eV).23 The combination of electrochemical and XRD characterization (see Fig. 3 in the ESI†) reflects the inactivity of Ca2Fe2O5 in both Ca and Li electrolytes. A plausible explanation is that Ca deintercalation from these materials requires electrolytes operating in the 5 V range, so as to allow the oxidation of Fe to the +4 oxidation state.
To test the electrochemical stability window of Ca4Fe9O17, Li/LP30/Ca4Fe9O17 cells were discharged down to 1 V. Upon reduction, the potential vs. capacity profiles for Ca4Fe9O17 displayed in Fig. 5b(A) present a long plateau-like region centered at 1 V vs. Li+/Li and additional diffraction peaks are observed in the 2θ = 10° region of the corresponding ex situ X-ray powder diffraction pattern shown in Fig. 5c(A). This large capacity can be regarded as the combination of two competing processes: the low-voltage decomposition of the electrolyte and the reduction of Fe3+ to Fe2+ upon Li uptake. Noteworthily, a similar potential vs. capacity profile is observed for the reduction of Fe2O3 electrodes in Li cells;48 the electrochemical conversion reactions of Fe3+ compounds with lithium are known to evolve until the eventual formation of small Li2O and metallic iron domains.48,49 In the present case, the identification of the phases formed during the reduction remains undetermined due to the complex nature of the starting material. However, these results highlight the instability of Ca4Fe9O17 upon reduction, contrary to the reported stability of the related K-phase down to 0 V vs. K+/K.44
Footnote |
† Electronic supplementary information (ESI) available: Table S1: DFT and lattice parameters, Fig. S1: calculated energy migration barrier for CaFe2O4, Table S2 and Fig. S2: summary of synthesis methods, Fig. S3: Rietveld refinement, and Fig. S4: electrochemistry and XRD of Ca2Fe2O5. See DOI: 10.1039/C9DT04688G |
This journal is © The Royal Society of Chemistry 2020 |