Article pubs.acs.org/cm
Vanadium Clustering/Declustering in P2−Na1/2VO2 Layered Oxide Marie Guignard,*,† Dany Carlier,† Christophe Didier,†,¶ Matthew R. Suchomel,‡ Erik Elkaïm,§ Pierre Bordet,⊥ Rodolphe Decourt,† Jacques Darriet,† and Claude Delmas† †
CNRS, Université de Bordeaux, ICMCB, 87 avenue du Dr A. Schweitzer, 33608 Pessac Cedex, France Argonne National Laboratory, Advanced Photon Source, Lemont, Illinois 60439, United States § Synchrotron SOLEIL, l’Orme des Merisiers Saint-Aubin, 91192 Gif-sur-Yvette Cedex, France ⊥ CNRS, UJF, Institut NEEL, Cedex 9, 38042 Grenoble, France ‡
S Supporting Information *
ABSTRACT: The new layered phase P2−Na1/2VO2 has been synthesized by sodium electrochemical deintercalation. Its structure has been studied by high resolution powder diffraction, pair distribution function analysis, and nuclear magnetic resonance spectroscopy between 300 and 350 K. An increase of 2 orders of magnitude in its electronic conductivity has been observed at approximately 322 K, and a structural transition has been found to occur simultaneously. The arrangement of sodium ordering in P2−Na1/2VO2, which maximizes sodium−sodium distances to lower electrostatic repulsions between alkali ions, is found to be unchanged across this transition. At room temperature, high resolution powder diffraction and pair distribution function analysis reveal the triangular lattice formed by vanadium ions to be distorted by the formation of pseudotrimers clusters with vanadium− vanadium distances as short as 2.581 Å. Above the transition, the pseudotrimers disappear and the triangular vanadium lattice becomes more regular with a mean vanadium−vanadium distance of ∼2.88 Å. At 350 K, the increase in P2−Na1/2VO2 electronic conductivity is due to enhanced charge transport resulting from the declustering of vanadium ions. These results highlight how sodium ordering between the MO2 layers and the electronic transport within the MO2 layers are intimately correlated in NaxMO2-type sodium-layered oxides.
1. INTRODUCTION Sodium-layered oxides NaxMO2 (where M is a 3d cation) were initially studied thirty years ago for use as positive electrode materials in secondary batteries.1 However, competing lithiumbased battery technology soon showed more promise, and research on sodium compounds for battery applications was largely neglected. More recently, sodium batteries have once again generated interest, in part because sodium is much less expensive than lithium and it is widely available around the world. Numerous studies over the past few years have examined the structure and properties of sodium-layered oxide systems, including their performance as electrode materials in sodium battery technologies. (See, for example, ref 2 and references therein.) Among the sodium-layered oxide systems, the cobalt-based compounds have received the most attention due to the high thermoelectric response in Na0.70CoO23 and the discovery of superconductivity in Na0.35CoO2, 2H2O.4 During the last ten years alone, approximately six hundred papers were devoted to the NaxCoO2 system, with 0 ≤ x ≤ 1. The majority of these studies have considered the P2 type structure presented in Figure 1 (see ref 5 for details on the nomenclature of layered structures). In this structure type, CoO2 layers made of edgesharing CoO6 octahedra are packed following the ABBA stacking of the oxygen layers, leading to the formation of two © 2014 American Chemical Society
unique trigonal prismatic sites that can be occupied by sodium cations. The first prismatic site (Nae) is edge sharing to the CoO6 octahedra, whereas the second site (Naf) shares two faces with them. The distribution of sodium ion occupancy on these sites is determined by a competition between two interactions: first the Na+−Na+ electrostatic repulsions that tend to maximize distances between Na+ ions, and second the Na+− Co3/4+ repulsions that tend to minimize the occupancy of the face-sharing prismatic sites. This competition results in numerous Na+ ordering arrangements between the CoO6 layers.6 Consequently, the CoO2−NaCoO2 phase diagram is a complex mix of single-phase compositions, solid solutions, and biphasic domains with multiple superstructure or modulated structures.7 This Na+ patterning strongly influences the physical properties of these compounds, in particular affecting transport properties by controlling the electronic charge carried by the metallic ions.8 Recently, we have investigated and described an equally rich and complex phase diagram exhibiting compositionally dependent Na+ ordering in the P2−NaxVO2 system, and have demonstrated the existence of a new single phase for the Received: September 18, 2013 Revised: January 22, 2014 Published: January 27, 2014 1538
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials
Article
pair distribution function (PDF) analysis;20 the disappearance of these trimers at high temperature has been proposed to explain the large increase of the magnetic moment above the transition. In the current work reported here, we communicate the full details of a very similar transition of vanadium trimers ordering occurring in the P2−Na1/2VO2 system across the temperature range identified in our prior work. In this system, a lower symmetry parent structure and mixed valent V3+ and V4+ species results in many additional vanadium−vanadium distances, and corresponding adds complexity to the transition mechanism. In this paper, we precisely determine and report the structural modifications occurring during the transition using high resolution X-ray powder diffraction (HRXRPD), pair distribution functions (PDF) analysis, and 23Na nuclear magnetic resonance (NMR) spectroscopy. Although HRXRPD gives information about the modifications of the long-range order and of the cooperativeness of the transition, PDF analysis allows studying the changes that occur at the local scale. More specifically, PDF data obtained from X-ray scattering are particularly sensitive to the changes in vanadium−vanadium distances in P2−Na1/2VO2 because of the high atomic number contrast between vanadium and less electron-dense sodium and oxygen elements in the structure. This work affirms that elucidating the structural details of vanadium clustering is often the key to understanding the electronic and magnetic transitions observed in vanadium oxides. Moreover, we show that 23Na NMR spectroscopy is not only a powerful probe of the local sodium environment. In this paramagnetic material, it also gives insight into electronic structure of P2−Na1/2VO2.
Figure 1. Basic P2−NaxMO2 structure type with hexagonal P63/mmc symmetry.
specific composition P2−Na1/2VO2.9 The ambient temperature structure of this new phase was determined from high resolution X-ray powder diffraction, and it is distinguished not only by sodium ordering but also by the existence of vanadium pseudotrimers in the triangular lattice formed by the vanadium ions. Preliminary structural and electronic conductivity studies in that work have revealed concurrent transitions at ∼322 K, above which the electronic conductivity increases by 2 orders of magnitude. In vanadium oxides, structural transitions are generally intimately connected to their electronic properties. A metal−insulator transition (MIT) occurring at approximately 340 K in vanadium dioxide VO2 has been known for more than 50 years.10 Above 340 K, the VO2 structure is tetragonal (space group: P42/mnm)11 with equivalent vanadium−vanadium nearest-neighbor distances of 2.851 Å.12 Below 340 K, its structure is monoclinic (space group: P21/c)13 and is characterized by the existence of vanadium−vanadium dimers with V−V distances as short as 2.619 Å.14 However, although well documented over the past 50 years, the primary mechanism of the MIT in VO2 is still under debate. Questions remain concerning the relative importance of electron−lattice interactions and electron− electron correlations in the transition. The MIT not only has been observed in VO2 but also is seen in V2O310 and many Magnéli phases of the general formula VnO2n−115 containing vanadium ions with a formal valence from +3 to +4. The layered lithium vanadium oxide LiVO2, which is more closely related to P2−Na1/2VO2, also displays an electronic transition at approximately 460 K.16 LiVO2 does not achieve a metallic state above the transition, but the anomalies observed in both electronic conductivity and magnetic susceptibility on thermal evolution have been correlated to a structural transition. Below the transition, the magnetic moment is much lower than that expected for the isolated vanadium ions V3+. Goodenough suggested that the drastic change in the magnetic susceptibility, occurring in this material at the electronic transition temperature,17 could be explained by the clustering of vanadium ions to form trimers.18 The existence of these trimers has subsequently been confirmed by both extended X-ray absorption fine structure (EXAFS)19 and more recently by
2. EXPERIMENTAL SECTION 2.1. Synthesis. The P2−Na1/2VO2 phase has been obtained by sodium electrochemical deintercalation from P2−Na0.71VO2, which is the only composition that can be synthesized with the P2-type structure by classical high temperature solid state chemistry. The preparation of P2−Na0.71VO2 powder required several successive steps. First, NaVO3 was synthesized by solid-state reaction of a stoichiometric mixture of Na2CO3 (purity above 99.9%) and V2O5 (purity above 99.9%) by heating it under O2 atmosphere at 600 °C for 12 h. Then, NaVO2 was prepared in a gold crucible by chemical reduction of NaVO3 powder under H2 atmosphere at 700 °C for 8 h. Finally, P2−Na0.71VO2 was synthesized by solid-state reaction between NaVO2 and VO2 at 850 °C for 24 h in a gold tube sealed under argon. The electrochemical synthesis has been realized in a sodium battery with the following electrochemical chain: metal Na | NaClO4 (1M) in propylene carbonate | P2−Na0.71VO2. The P2−Na0.71VO2 compound has been used as the positive electrode either as a powder mixed with a graphite/carbon-black powder (90 wt % P2−Na0.71VO2, 10 wt % carbon) or as a sintered pellet of pure P2−Na0.71VO2 phase without any addition of carbon for the electronic characterizations of P2− Na1/2VO2. The batteries have been galvanostatically charged at the C/ 100 rate (1 e− per vanadium transferred in 100 h) up to 2.40 V that corresponds to the Fermi level (vs the Na+/Na couple) for P2− Na1/2VO2. This voltage was determined in our previous electrochemical study of the phase diagram of the P2−NaxVO2 system.9 Once this voltage has been reached, it has been continuously applied to the electrochemical cell. The current has been measured during the potentiostatic charge. When it drops down to zero, it means that the expected material is obtained. The voltage has been applied for at least 50 more hours in order to reach a very good equilibrium in the material. After the charge, the pellet or the powder has been recovered, washed with dimethyl carbonate, and dried under vacuum. Note that all the NaxVO2 phases (with x = 1, 0.71 and 1/2) are highly sensitive to humidity or to oxygen, and they have been always handled under argon in a glovebox. Furthermore, each characterization, either 1539
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials
Article
conditions, we ensure that the main signal observed is due to the −1/2 → + 1/2 central transition. The spectral width has been set to 1 MHz, and the recycle time D0 = 0.2 s is long enough to avoid T1 saturation effects. The baseline distortions resulting from the spectrometer dead time (8 μs) have been computationally removed using a polynomial baseline correction routine. The external reference was a 0.1 mol L−1 NaCl aqueous solution. We carried out the fit of signals using the DMfit software.24 In order to assign the 23Na NMR signals, density functional theory (DFT) calculations have been performed using the VASP code25 for the structural optimization and the WIEN2K code for the calculations of the electronic spin density at the different 23Na nuclei.26 The Fermi contact shifts were then further calculated considering the magnetic susceptibility at the experimental temperatures as described in reference.27 Note the use of the experimental structures obtained from the Rietveld refinements against HRXRPD data or the ones optimized by GGA calculations lead to similar EFG and Fermi contact shifts values for the different Na sites and, therefore, does not modify the signal assignment proposed in the paper. In the following, we chose to provide only the results obtained for the optimized cells.
physical or structural, has been realized under an argon atmosphere. This method allows synthesizing the P2−Na1/2VO2 powder in a very reproducible way. Note that a very small amount of V2O3, already present in the P2−Na0.71VO2 powder before the sodium deintercalation, is present as an impurity in the final product. Additionally, some synthesis attempts resulted in trace amounts of Na4V2O7 as detected by laboratory powder X-ray diffraction. 2.2. Physical Characterizations of P2−Na1/2VO2. Magnetic susceptibility has been measured using a superconducting quantum interference device (Quantum Design MPMS). Zero-field-cooled and field-cooled experiments have been carried out under a 5 kOe magnetic field and in the 2−350 K temperature range. Approximately 20 mg of powder have been used for the measurements, which have been realized in a gelatin capsule. The magnetic moment was measured every 2 K once the temperature was stabilized. The data have been corrected for the diamagnetism of the sample holder and of the material itself. For transport measurements, pellets have been used in an airtight apparatus, and the electrical resistivity has been measured using the four-probe technique in the 300−345 K temperature range. The resistivity of the pellet was measured while it was continuously heated up or cooled down at the 0.5 K/min rate. 2.3. Structural Characterizations of P2−Na1/2VO2. 2.3.1. High Resolution Powder Diffraction (HRXRPD). Synchrotron diffraction has been performed on P2−Na1/2VO2 at beamline 11-BM beamline of the Advanced Photon Source (Argonne National Laboratory, IL, U. S. A.). HRXRPD patterns have been recorded at 300 K and at 350 K over 0−55° 2θ-range with a 0.001° step size using a wavelength of λ = 0.41358 Å. Diffraction patterns have also been collected during a cooling sweep from 350 K down to room temperature. Synchrotron diffraction has been also performed on P2−Na1/2VO2 at 300 and 350 K at CRISTAL beamline (SOLEIL synchrotron, Gif-sur-Yvette, France). Diffraction patterns have been recorded over a large 2θrange (0−140°) with a 0.002° step size using a wavelength of λ = 0.45775 Å in order to extract pair distribution functions. The structure of P2−Na1/2VO2 has been refined against HRXRPD data collected in 0−40° 2θ-range at 300 and 350 K at 11-BM beamline using the Rietveld method as implemented in the Jana2006 software package.21 2.3.2. Pair Distribution Functions (PDF). PDFs have been calculated from powder diffraction data recorded at CRISTAL beamline. The diffraction data recorded on an empty glass capillary in the same conditions have been used to correct the diffracted intensity from the background. The measured diffracted intensity has been corrected from the incoherent scattering, the background, the polarization, and the absorption using the ad-hoc method provided by PDFgetX3 software.22 The corrected intensity I(Q) and the structure function S(Q) have been calculated using this program, with Q = 4πsinθ/ λ. Reduced PDFs G(r) have been obtained by Fourier transforming S(Q) functions over the 0.5−24 Å−1 Q-range. PDFs have been calculated from the structure obtained from the Rietveld refinement of the HRXRPD data and the structure has subsequently been refined to fit the experimental data using the PDFgui software.23 2.3.3. Solid State 23Na Nuclear Magnetic Resonance (NMR) Spectroscopy. Single pulse 23Na magic-angle spinning (MAS) NMR spectra have been recorded on a Bruker 300 Avance spectrometer at 79.403 MHz (B0 = 7.05T) and on a Bruker 500 Advance spectrometer at 132.302 MHz (B0 = 11.75T). Variable temperature NMR experiments have been carried out in the 300−350 K temperature range using a Bruker 2.5 mm DVT MAS probe. A spinning speed of 30 kHz has been used. Temperature calibration using lead nitrate indicates that while spinning at 30 kHz, about 80% of the sample in the rotor undergo a given temperature ±2 °C, and about 20% of the sample undergoes a temperature variation up to 10 °C versus the main temperature. The samples have been placed in zirconia rotors in a drybox and studied by NMR immediately after taking the rotors out of the glovebox. An X-ray diffraction pattern has been recorded after the NMR experiments to check that the samples did not evolve during the NMR data acquisitions, indicating satisfactory air-tightness of the rotors. As 23Na is a quadrupolar nucleus with I = 3/2, a short pulse length of 1 μs corresponding to a selective π/2 pulse determined using an aqueous 0.1 mol L−1 NaCl solution has been employed. In these
3. RESULTS 3.1. Physical Characterizations. Electronic conductivity and magnetic susceptibility measurements performed as a function of temperature for P2−Na1/2VO2 are shown in Figure 2a and b, respectively. Both panels show clear evidence of a
Figure 2. (a) Electronic conductivity, σ, of P2−Na1/2VO2 measured during heating from 306 to 345 K (black diamonds) and during cooling from 345 to 306 K (red diamonds). (b) Molar magnetic susceptibility, χm, of P2−Na1/2VO2 measured during heating from 300 to 350 K (black diamonds) and during cooling from 350 to 300 K (red diamonds).
response anomaly at approximately 322 K, with an increase of 2 orders of magnitude in the electronic conductivity and an unambiguous strengthening of magnetic susceptibility on heating. These anomalies are also observed during the cooling of P2−Na1/2VO2 with a small hysteresis, the hysteresis observed on the electronic measurements being larger because the data have been collected during the heating and cooling 1540
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials
Article
Figure 3. Contour plot representing a part of the synchrotron diffraction patterns of P2−Na1/2VO2 recorded during sweeping from 350 to 300 K.
Figure 4. High resolution X-ray powder diffraction (HRXRPD) patterns of P2−Na1/2VO2 recorded at 350 K (orange line) and at 300 K (blue line). An enlargement is shown in the inset (top right).
a true paramagnetic state cannot be reached before the sample starts to decompose at approximately 450 K. 3.2. Structural Characterizations of P2−Na1/2VO2. 3.2.1. High Resolution Powder Diffraction (HRXRPD). The room temperature structure of P2−Na1/2VO2 has been determined recently from the Rietveld refinement of HRXRPD data.9 Its unit cell is orthorhombic (space group: Pnma) and the cell parameters are a = 9.8879(3) Å, b = 5.7418(2) Å, and c = 11.4581(6) Å. The atomic positions of this structure are recalled in the crystallographic information file available in the Supporting Information of the current study. Figure 3 shows a contour plot representing a part of the diffraction patterns recorded during sweeping from 350 to 300 K at the 11-BM beamline. From 321 K, new diffraction lines appear, indicating that a new phase appears below this temperature. They correspond to those of the room temperature structure of P2− Na1/2VO2 and a biphasic domain exists down to 315 K. Below 315 K, the diffraction peaks characteristic for the “high temperature” phase (i.e., the phase existing at 350 K) are no longer visible. Therefore a reversible structural transition clearly
ramps, whereas the magnetic measurements have been realized using step scans once a stable temperature was reached. Both electronic and magnetic measurements show that the transition occurring in P2−Na1/2VO2 at approximately 322 K is reversible when the material is heated up to 350 K and cooled down back to room temperature. Both below and above the transition temperature the electronic behavior is semiconducting, with an increase of conductivity when the temperature increases. In our previous paper, we have shown that the inverse of the magnetic susceptibility can be fit between 30 and 300 K using Curie− Weiss’ law corrected by temperature-independent paramagnetism (χm = C/(T − θ) + χ0, with C = 0.084 emu K mol−1, θ = −7.4 K, and χ0 = 140 × 10−6 emu mol−1).9 The Curie constant calculated in this case is very low compared with the one calculated for an equimolar mixture of isolated V3+ (S = 1) and V4+ (S = 1/2) ions (C = 0.69 emu K mol−1). This has been explained by the presence of vanadium pseudotrimers within the VO2 slabs. Magnetic measurements have also been performed above the transition from 325 to 450 K. However, 1541
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials
Article
Figure 5. Experimental high resolution X-ray powder diffraction (HRXRPD) pattern of P2−Na1/2VO2 recorded at 350 K (red diamonds), and the calculated pattern obtained from the Rietveld refinement (black line) and difference line (blue line). An enlargement is also shown in the inset (top right).
have not been added as additional phases in the Rietveld refinement because their addition leads to negligible improvement of the agreement between the experimental and the calculated diagrams for the P2−Na1/2VO2 phase. After this analysis, we find that the structure of P2− Na1/2VO2 at 350 K possesses an orthorhombic unit cell as the room temperature structure but with approximately one-half of the unit size. The atomic positions determined from the Rietveld refinement of the diffraction data at 350 K are given in Table 1 (the structural data are available in a crystallographic
occurs at a temperature range close to that of the electronic and magnetic transition. In order to study the structural modifications occurring during the phase transition, HRXRPD data have been recorded at 350 K in this study to determine the high temperature structure. Figure 4 shows the diffraction patterns of P2− Na1/2VO2 recorded at beamline 11-BM. To better highlight subtle changes in the diffraction pattern below and above the transition, data have also been recorded as part of the same experiment after in situ cooling down to 300 K (Figure 4). It is clear from the disappearance of many weak diffraction peaks that the high temperature structure observed above the transition is more symmetric than the room temperature structure below the transition. Moreover, we have previously demonstrated that orthorhombic distortions of the triangular lattice formed by the vanadium ions leads to the splitting of some diffraction peaks of P2−Na1/2VO2 at room temperature,9 in particular those centered around 2θ = 9.6, 10.4, 11.4, and 12.7° in Figure 4. This splitting is less visible at 350 K, indicating that the triangular lattice is less distorted above the transition (see, for example, the splitting of two diffraction peaks with the indexation (223) and (403) at 300 K that give rise to a single peak, the sum of two peaks with the indexation (231) and (032) at 350 K). The high temperatures structure of P2−Na1/2VO2 has been determined in this study from HRXRPD data measured at 350 K. The diffraction pattern has first been indexed with the following orthorhombic cell parameters: a = 5.7447(1) Å, b = 11.4765(3) Å, c = 4.9813(1) Å in the space-group Pmmn, which was found from the extinction of certain diffraction peaks. The vanadium atoms have been placed in the unit cell to form the triangular lattice. The atomic positions for sodium and oxygen ions were found using the Fourier difference maps and are in agreement with other similar, known layered structures. Finally, all the atomic positions have been refined when it was allowed by the symmetry using the Rietveld method leading to good reliability factors (Rwp = 12.6%, RBragg = 5.9%). The good agreement between the calculated diffraction diagram and the experimental one is shown is Figure 5. For clarity and simplicity, a small V2O3 impurity and the graphite additive
Table 1. Refined Atomic Positions and Atomic Displacement Parameters (Biso) in P2−Na1/2VO2 at 350 K in the Orthorhombic Unit Cell with a = 5.7447(1) Å, b = 11.4765(3) Å, c = 4.9813 (1) Å and the Space-Group Pmmn (Origin Choice 2) V(1) V(2) Nae(1) Naf(2) O(1) O(2) O(3)
x
y
z
Biso (Å2)
0 0.25 −0.25 −0.25 0.0017(4) −0.25 0.25
0 −0.00795(6) 0.25 −0.25 0.0945(2) −0.0876(2) −0.0951(2)
0.5 0.01695(12) 0.3489(5) 0.0498(5) 0.1624(4) 0.3270(5) 0.3368(5)
0.74(2) 0.53(2) 1.91(4) 3.29(6) 0.46(5) 0.36(5) 0.62(5)
information file in the Supporting Information). The structure of P2−Na1/2VO2 after cooling down to 300 K has also been refined from the HRXRPD data recorded at 11-BM beamline, and it is in good agreement with that determined previously at the same temperature in our earlier study,9 confirming that the transition is completely reversible. The result of this refinement is given in Supporting Information. Figure 6 shows a projection of the P2−Na1/2VO2 structure at 300 and 350 K determined from Rietveld refinement of the HRXRPD data. Above the transition, the triangular lattice becomes more symmetric at 350 K with only two crystallographic sites occupied by vanadium ions instead of three at 300 K. As a consequence, the vanadium pseudotrimers disappear above the transition and V−V distances are more homogeneous between 2.804 Å and 2.949 Å with a V−V mean distance of 1542
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials
Article
Figure 6. Projection of the structure of P2−Na1/2VO2 along the direction perpendicular to the VO2 layers. The structures have been obtained from the Rietveld refinement of the X-ray high resolution powder diffraction (HRXRPD) data recorded (a and c) at 300 K and (b and d) at 350 K. All the possible vanadium−vanadium distances are reported, and the unit cells are drawn with a black line.
scattering data for PDF analyses were collected at the CRISTAL beamline on P2−Na1/2VO2 at 300 and 350 K. The reduced PDFs, G(r), for P2−Na1/2VO2 at 300 and 350 K are shown in Figure 7. Clear differences in G(r) between 300 and 350 K are observed over the r ranges of 2.4−3.8 Å, and 7.9−9.0 Å, as highlighted on the Figure 7a. Subtle differences are also visible at r distances of ∼4.53, ∼ 4.9, and ∼6.0 Å. The expression of G(r) can be written as a function of the total PDF, g(r), as
2.877 Å (compare to a V−V range of 2.581−3.055 Å at 300 K). Above the transition, the sodium ordering observed at room temperature is maintained and each of the two crystallographic sites available for sodium ions is still fully occupied. This ordering requires using an orthorhombic cell to describe the structure of P2−Na1/2VO2 above the transition temperature, although the splitting of some diffraction lines mentioned above is now negligible. The sodium/vacancy ordering found at 350 K permits maximal Na+−Na+ distances between the VO2 layers in order to minimize the electrostatic repulsions as it was observed at 300 K. However, an increase in Na mobility above the transition is suggested by the increase of the atomic displacement parameters found for Na+ ions (Biso = 1.91 and 3.29 Å2 for Nae(1)a and Naf(2) at 350 K, respectively, instead of 1.15 Å2 for the two ions at 300 K). Therefore, we conclude that the main structural changes occurring during the transition are related to the vanadium triangular lattice. 3.2.2. Pair Distribution Functions (PDF). Pair distribution function (PDF) analyses have also been performed in order to more accurately track short- and medium-range ordering changes through the structural transition, because distances can be dramatically affected by the error made on the atomic positions determined from the Rietveld refinement. High Q
G(r ) = 4πrρ0 [g (r ) − 1]
where ρ0 is the atomic number density of the sample. Moreover, the total PDF, g(r), can be written as a function of the partial PDFs, gij(r), which represents the probability of finding an atom j at a distance r from atom i. (See ref 28 for the definition of the different PDFs.) Given both the scattering factor of vanadium and its relative concentration in P2− Na1/2VO2, it follows that the partial PDFs involving vanadium dominate G(r). The reduced PDF, G(r), has also been calculated from the crystal structures determined by Rietveld analysis at 300 and 350 K. These structures have been relaxed in order to better fit the calculated G(r) against the experimental data, but only small changes in the atomic positions (less than 2%) have been found, indicating that the 1543
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials
Article
Rietveld refinement of HRXRPD data.9 At 350 K, only one broader peak is observed between 2.4 and 3.2 Å in the experimental G(r) indicating that the vanadium−vanadium distances are now more homogeneous with a V−V mean distance of 2.91 Å, again in good agreement with the mean V− V distance of 2.877 Å obtained from HRXRPD. This rearrangement of vanadium ions within the triangular lattice also explains the changes in G(r) visible at approximately 4.9 and 6.0 Å, and between 7.9−9.0 Å. Only very weak changes are observed in the partial PDFs that are not involving vanadium, indicating that sodium and oxygen positions are nearly invariant through the transition. These species build the fixed part of the lattice in P2−Na1/2VO2 within which the vanadium ions can shift. Therefore, as the vanadium ions displace within the structure, the vanadium−oxygen distances are altered, which explains changes in experimental G(r) observed at approximately 3.5 and 4.5 Å. 3.2.3. Solid State 23Na Nuclear Magnetic Resonance (NMR). Figure 8 shows the 23Na MAS NMR spectra of the P2−Na1/2VO2 phase recorded under different magnetic fields and at different temperatures. The 23Na nucleus exhibits a strong quadrupolar constant. Therefore, if 23Na is located on a site with a noncubic symmetry, a broad signal remains under magic-angle spinning conditions due to the second order quadrupolar interaction that is affected, but not suppressed, by MAS. Using a higher magnetic field permits an increase in spectral resolution because the second order quadrupolar interaction line width, expressed in parts per million, decreases when the magnetic field increases. At 300 and 350 K, 23Na signals of the P2−Na1/2VO2 phase are observed in the 50−250 ppm range, corresponding to the central transitions of Na sites. Sets of spinning side bands are also observed on both sides of the central transitions. Another signal located just below 0 ppm, with slightly longer T1 relaxation time, is also observed and assigned to diamagnetic sodium impurities (for example, the Na4V2O7 impurity suggested by X-ray diffraction or a small amount of remaining NaClO4 salt used in the electrolyte). The signals of P2−Na1/2VO2 exhibit a clear second order quadupolar line shape for B0 = 7.05 T and a less pronounced one, as expected, for B0 = 11.75 T. The spectra can be fitted in order to extract the electric field gradient parameters and the isotropic positions. The fitted spectra are given in the Supporting Information, and the resulting parameters are given in Table 2. Because the isotropic positions are located far out of the range of chemical shifts of diamagnetic compounds [−20; 55 ppm],29 we conclude that the Fermi contact interaction mostly governs the signals position, whereas the second order quadrulopar interaction governs the signal line shape. Using DFT and DFT+U calculations, we undertook the modeling of the electronic structure of P2−Na1/2VO2 with and without vanadium pseudotrimers and the resulting hyperfine interactions and EFG parameters on the different Na sites. These calculations will be published in a forthcoming paper; only the resulting signal assignments are reported here based on the calculated Fermi contact shifts as described in ref 27 using the wien2k code (Table 2). Below the transition temperature, the P2−Na1/2VO2 phase exhibits two signals, in agreement with the structural model. Spinning side bands are observed on both sides of these signals and are extended over a very large range (>1 MHz). The electric field gradient parameters and isotropic positions of the two signals are reported in Table 2, together with the resulting Fermi contact
Figure 7. (a) Experimental reduced pair distribution functions (PDFs), G(r), of P2−Na1/2VO2 at 300 K (blue line) and 350 K (orange line). (b and c) Experimental G(r) (red diamonds), calculated G(r) (black line), and difference line (blue line) of P2−Na1/2VO2 at 300 and 350 K, respectively.
short-range order obtained from the structural model is in good agreement the PDF data. Therefore, relaxed structures obtained from the refinement of the experimental G(r) are not considered in this paper. The calculated PDF shown in Figure 7b and 7c are in good agreement with the experimental data. Partial PDFs, gij(r), have also been calculated (see Figure S1 in the Supporting Information), and they clearly show that the main changes observed in G(r) between 300 and 350 K are related to the vanadium−vanadium correlations. One of the most significant changes in atomic correlations across the transition is over the r range from 2.4 to 3.2 Å that corresponds to the nearest vanadium−vanadium distances (Figure 7a). At 300 K, a partial PDF gVV shows the existence of two vanadium−vanadium nearest neighbor distances, a short correlation at 2.59 Å and a longer one at 2.94 Å that are also visible in the experimental G(r). These distances are in good agreement with the average short and long vanadium−vanadium distances (2.616 and 2.965 Å, respectively) found in the structure obtained from the 1544
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials
Article
Figure 8. Variable temperature 23Na MAS NMR spectra (νR = 30 kHz) observed for the P2−Na1/2VO2 phase at 300 K (blue line) and 350 K (orange line) at 7.05 T (a) and at 11.75 T (b). An enlargement around the central transition is also shown (c and d). The spinning side bands are topped by an asterisk (*).
a chemical exchange of the two Na sites observed at 300 K because its position is not the average of the two other ones. Moreover, according to the increase of the magnetic susceptibility observed above the transition temperature (Figure 2), a larger mean contact shift would be expected since it is proportional to the molar magnetic susceptibility following27
Table 2. Fitted EFG Parameters and Isotropic Positions of the 23Na MAS NMR signals (νR = 30 kHz) Observed for the P2−Na1/2VO2 Phase at 7.05 T and for T ∼ 300K and T ∼ 350 Ka experimental 300 K 350 K
Nae(1) Naf(2) Nae(1) Naf(2) average
calculated
Qcc (MHz)
η
shift (ppm)
shift (ppm)
1.7 1.5
0.57 0.23
173 232
1.5
0.5
109
163 266 −66 204 69
i δiso =
1 i ρ (0)χm 3S
where ρ(0) is the electronic spin density at the probed nucleus and S is the spin of the transition metal ion. The position of the average signal at 350 K clearly indicates that a real change in the electronic structure of P2−Na1/2VO2 occurs during the transition. In our modeling, we computed a completely different electronic structure when the vanadium pseudotrimers are not present anymore. The resulting shifts on the two Na sites are given in Table 2: Naf(2) is expected around 204 ppm and Nae(1) is expected around −66 ppm. Because the sign of the Fermi contact shift of the two Na ions is opposite, this means that the spin transfer mechanisms from the V ions to the Na nuclei is different: a delocalization mechanism is involved for Na e(1) and a polarization mechanism is involved for Naf(2).30 If we calculate the expected average position resulting from the Na chemical exchange at the NMR time scale, we predict the signal to be located around 69 ppm, which is close to the 109 ppm experimental value, indicating the good modeling of the electronic structure. After the heat treatment up to 350 K in the rotor, two identical 23Na MAS NMR signals were recorded again at 300 K, showing the good reversibility of the transition seen from a local probe. Note that variable temperature NMR data were acquired at intermediate temperatures between 300 and 350 K, but they did not show the presence of another
The corresponding fitted spectra are given in the Supporting Information. Calculated Fermi contact shifts using DFT and DFT +U are also provided.
a
shift calculations. On the basis of our modeling, the mostshifted signal located at 232 pm is assigned to Naf(2), and the least shifted signal located at 173 ppm is assigned to Nae(1). Note that the chemical shift contribution was neglected for our assignment. At 350 K, above the structural transition, a single signal is observed. This could correspond to the central transition of one Na environment; however, two Na distinct environments would be expected from the structural model determined from HRXRPD. This possible contradiction can be explained by chemical exchange between the two sites occurs at the NMR time scale, leading to an averaged single signal at 350 K. Indeed, this signal contains approximately the same intensity as that of the sum of the central lines of the two Na sites spectra recorded at 300 K and a weaker extension of the spinning side bands manifolds around the central lines, in agreement with a motional narrowing of the entire spectrum. Note, however, that the average position of the signal at 350 K does not result from 1545
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials
Article
P2−Na2/3VO2 and P2−Na5/8VO2.9 Even if the structure of these two new phases has not been fully solved yet, HRXRPD indicates that the basic structure remains hexagonal and, therefore, the triangular lattice formed by vanadium ions is unlikely to be distorted due to trimerization. Moreover no electronic or magnetic transitions are found in P2−Na2/3VO2 and P2−Na5/8VO2 over the temperature range (2−350 K) where these transitions related to metal clustering are usually observed in vanadium oxides. These two aforementioned phases differ from P2−Na1/2VO2 in their sodium content and, consequently, in the formal charge of vanadium ions, which decreases as the sodium content increases. There are more electrons filling the t2g orbitals in P2−Na2/3VO2 and P2− Na5/8VO2. We argue that the V clustering in P2−Na1/2VO2 is due to its particular electronic structure, in which its energy can be lowered by a trimerization of vanadium ions similar to that seen in LiVO2. Above the 350 K transition in P2−Na1/2VO2, vanadium pseudotrimers disappear. Electronic and magnetic measurements provide corroborating evidence that the electronic charge is less localized in P2−Na1/2VO2 above the transition. It is interesting to consider the critical cation−cation separation Rc, proposed by Goodenough, which allows calculating the critical distance below which a cation−cation overlap of the t2g orbitals occurs in transition-metal oxides leading to itinerant electrons35
signal that, for example, would result from the Na exchange in P2−Na1/2VO2 structure below the temperature transition with vanadium pseudotrimers.
4. DISCUSSION In vanadium oxides, structural transitions generally involve variations in vanadium clustering that alter electron localization and, consequently, are typically commensurate with changes in the materials’ electronic properties. We present here a new example of a reversible structural transition correlated to an electronic transition in a vanadium oxide. The structural characterization of the P2−Na1/2VO2 phase using Rietveld refinement of HRXRPD data, PDF analysis, and 23Na MAS NMR show that the changes in electronic properties observed at 322 K are associated with a disappearance of the vanadium pseudotrimers without any significant change in the atomic positions of the other ions. To understand the involved mechanisms, it is necessary to consider the structure of P2− Na1/2VO2 at room temperature. It presents three types of vanadium ions, which form three different rows made of VO6 octahedra along the b axis (Figure 6a). Each V(2)O6 and V(3)O6 octahedron shares one of its faces with one Naf(2)O6 prism, whereas each V(1)O6 octahedron shares only edges with the Nae(1)O6 prisms. However, only V(1) and V(2) atoms are involved in the pseudotrimers. The formation of these pseudotrimers tends to increase some of V3/4+−Na+ distances and, therefore, tends to stabilize the system at room temperature. When the temperature increases from 300 to 350 K, the thermal expansion induces an increase in the mean vanadium− vanadium distance in the triangular lattice. Simultaneously, the higher-entropy state becomes thermodynamically favored, the phases below and above the transition having different electronic and vibrational entropies arising from electron localization and the concomitant change in bond lengths. There is a displacement of vanadium ions in the triangular lattice that gives rise to two nonequivalent vanadium positions at 350 K instead of three nonequivalent positions in the structure below the transition temperature. The structures of P2−Na1/2VO2 at 300 and 350 K are related, and one parameter of the 350 K unit cell is doubled to describe the 300 K strucutre. The relationships between the cell parameters of the structure at 300 K and that at 350 K are
R c = 3.20 − 0.05m − 0.03(Z − ZTi) − 0.04Si(Si + 1)
where m is the formal cationic charge of the transition-metal, Z is its atomic number, and Si is its spin number. The critical cation−cation separation Rc is 2.94 Å for either V3+ or V4+ ions. At 350 K, above the transition temperature, the mean V−V distance is approximately 2.88 Å in the P2−Na1/2VO2 phase. Therefore, one would expect the electrons to be delocalized within the triangular lattice because the critical V−V separation is larger than the experimental average value. Moreover, the short V−O bond of approximately 1.75 Å that we referred to as a pseudovanadyl bond in our previous study9 disappears above the transition temperature (this bond is equal to 1.89 Å above the transition temperature). This indicates that the distribution of the charge carried by vanadium atoms seems even more delocalized above the transition temperature. However, the electronic conductivity remains thermally activated above the transition. Nevertheless, the sodium/vacancy order may prevent from the occurrence of a metallic state by partially localizing electronic charges on vanadium ions. Na+ ion positions in P2−Na1/2VO2 are relatively fixed across the transition, indicating that the driving force of electronic changes in P2−Na1/2VO2 is not fully related to sodium ordering. Moreover, this metal clustering is not observed in P2−Na1/2CoO2, which is isostructural to P2−Na1/2VO2 at 350 K and, therefore, has the same sodium ordering.36 The MIT occurring at approximately 53 K in this cobaltate has not been associated to any phase transition.36,37 Although this sodium/ vacancy order is not the driving force of the transition in P2− Na1/2VO2, the influence of Na cannot be completely excluded. We have recently synthesized another sodium-layered oxide structure, O’3−Na1/2VO2, which has exactly the same chemical composition and, therefore, the same number of electrons.38 This material is also made of edge-sharing VO6 layers, but the Na+ ions in this phase are located in octahedral sites instead of prismatic sites. Its structure, also determined using X-ray diffraction, showed the existence of vanadium−vanadium pairs
⎧ a300K ≈ 2c350K ⎪ ⎨ b300K ≈ a350K ⎪ ⎩ c300K ≈ b350K
The arrangement of vanadium ions below the transition temperature results in the formation of vanadium pseudotrimers with V−V distances as short as 2.581 Å, as confirmed by real space evidence from PDF analysis. The existence of such trimers is not unusual for vanadium compounds and has been reported both for LiVO2 and in more complex layered structures.31−33 In NaV6O11 and BaV10O15, for example, vanadium trimerization is also correlated to anomalies in the electronic and magnetic properties.32,33 Electronic band structure calculations show that the vanadium trimerization in LiVO2 (where vanadium ions are all d2) can lower the total energy and open a band gap at the Fermi level for a very small displacement of the vanadium ions.34 Recently, we have shown the existence of other single phases in the P2−NaxVO2 system: 1546
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials in this polymorph instead of vanadium pseudotrimers. Moreover, no structural transition is observed in O’3− Na1/2VO2 from 2 to 350 K. This indicates that its electronic structure must be different than that of P2−Na1/2VO2.
■
REFERENCES
(1) Delmas, C.; Braconnier, J.; Maazaz, A.; Hagenmuller, P. Rev. Chim. Miner. 1982, 19, 343−351. (2) (a) Yabuuchi, N.; Kajiyama, M.; Iwatate, J.; Nishikawa, H.; Hitomi, S.; Okuyama, R.; Usui, R.; Yamada, Y.; Komaba, S. Nat. Mater. 2012, 11, 512−517. (b) Buchholz, D.; Moretti, A.; Kloepsch, R.; Nowak, S.; Siozios, V.; Winter, M.; Passerini, S. Chem. Mater. 2013, 25, 142−148. (c) Mortemard de Boisse, B.; Carlier, D.; Guignard, M.; Delmas, C. J. Electrochem. Soc. 2013, 160, A569−A574. (3) Terasaki, I.; Sasago, Y.; Uchinokura, K. Phys. Rev. B 1997, 56, R12685−R12687. (4) Takada, K.; Sakurai, H.; Takayama-Muromachi, E.; Izumi, F.; Dilanian, R. A.; Sasaki, T. Nature 2003, 422, 53−55. (5) Delmas, C.; Fouassier, C.; Hagenmuller, P. Physica B+C (Amsterdam) 1980, 99, 81−85. (6) Zandbergen, H. W.; Foo, M.; Xu, Q.; Kumar, V.; Cava, R. J. Phys. Rev. B 2004, 70, 024101. (7) Berthelot, R.; Carlier, C.; Delmas, C. Nat. Mater. 2011, 10, 74− 80. (8) Roger, M.; Morris, D. J. P.; Tennant, D. A.; Gutmann, M. J.; Goff, J. P.; Hoffmann, J. -U.; Feyerherm, R.; Dudzik, E.; Prabhakaran, D.; Boothroyd, A. T.; Shannon, N.; Lake, B.; Deen, P. P. Nature 2007, 445, 631−634. (9) Guignard, M.; Dider, C.; Darriet, J.; Bordet, P.; Elkaïm, E.; Delmas, C. Nat. Mater. 2013, 12, 74−80. (10) Morin, F. J. Phys. Rev. Lett. 1959, 3, 34−36. (11) Westman, S. Acta Chem. Scand. 1961, 15, 217. (12) McWhan, D. B.; Marezio, M.; Remeika, J. P.; Dernier, P. D. Phys. Rev. B 1974, 10, 490−495. (13) Andersson, G. Acta Chem. Scand. 1956, 10, 623−628. (14) Longo, J. M.; Kierkegaard, P. Acta Chem. Scand. 1970, 24, 420− 426. (15) Kachi, S.; Kosuge, K.; Okinaka, H. J. Solid State Chem. 1973, 6, 258−270. (16) Kobayashi, K.; Kosuge, K.; Kachi, S. Mater. Res. Bull. 1969, 4, 95−106. (17) Bongers, P. F. Ph.D. Thesis, University of Leiden, 1957. (18) Goodenough, J. B. Magnetism And The Chemical Bond; Wiley Interscience: Hoboken, NJ, 1963). (19) Imai, K.; Sawa, H.; Koike, M.; Hasegawa, M.; Takei, H. J. Solid State Chem. 1995, 114, 184−189. (20) Pourpoint, F.; Hua, X.; Middlemiss, D. S.; Adamson, P.; Wang, D.; Bruce, P. G.; Grey, C. P. Chem. Mater. 2012, 24, 2880−2893. (21) Petricek, V.; Dusek, M.; Palatinus, L. Jana2006. The crystallographic computing system. Institute of Physics: Praha, Czech Republic, 2006. (22) Juhas, P.; Davis, T.; Farrow, C. L.; Billinge, S. J. L. J. Appl. Crystallogr. 2013, 46, 560−566. (23) Farrow, C. L.; Juhas, P.; Liu, J. W.; Bryndin, D.; Bozin, E. S.; Bloch, J.; Proffen, T.; Billinge, S. J. L. J. Phys.: Condens. Matter 2007, 19, 335219. (24) Massiot, D.; Fayon, F.; Capron, M.; King, I.; Le Calvé, S.; Alonso, B.; Durand, J.-O.; Bujoli, B.; Gan, Z.; Hoatson, G. Magn. Reson. Chem. 2002, 40, 70−76. (25) Kresse, G.; Furthmüller. J. Comput. Mater. Sci. 1996, 6, 15−50. (26) Schwarz, K.; Blaha, P.; Madsen, G. K. H. Comput. Phys. Commun. 2002, 147, 71−76. (27) Castets, A.; Carlier, D.; Zhang, Y.; Boucher, F.; Marx, N.; Croguennec, L.; Ménétrier, M. J. Phys. Chem. C 2011, 115, 16234− 16241. (28) Emagi, T.; Billinge, S. J. L. Underneath the Bragg peaks; Elsevier: Amsterdam, 2012.
ASSOCIATED CONTENT
S Supporting Information *
The crystallographic information files for P2−Na1/2VO2 at 300 and 350 K can be found in the Supporting Information. The fit of the HRXRPD data recorded at 11-BM beamline after cooling down to 300 K of P2−Na1/2VO2 is given in Supporting Information in a PDF file, as well as an enlargement of the fit of the HRXRPD data recorded at 350 K 11-BM beamline. The fits of the 23Na MAS NMR spectra of the P2−Na1/2VO2 phase recorded at 7.05 T using a νR = 30 kHz spinning frequency at T ∼ 300 K and T ∼ 350 K are also available in this file. Finally, partial pair distribution functions, gij(r), are given in this file, as well as tables with the main interatomic distances in P2− Na1/2VO2 at 300 and 350 K. This material is available free of charge via the Internet at http://pubs.acs.org.
■
ABBREVIATIONS
MIT, metal−insulator transition; HRXRPD, high Resolution Powder Diffraction; PDF, pair distribution function; NMR, nuclear magnetic resonance
5. CONCLUSION The structural transition in the layered P2−Na1/2VO2 phase at 322 K is associated with the disappearance of vanadium pseudotrimers in the triangular lattice formed by the vanadium ions. It induces a change in symmetry compatible with an essentially fixed sodium ordering framework. The elucidation of this new vanadium-ordering-based transition presented in this study can have implications for other P2−NaxVO2 phases and compositions with specific sodium ordering or solid solution domains with modulated structures. In particular, these findings are significant for other NaxVO2 polytypes with O’3 and P’3 oxygen packing, which can be obtained by electrochemical deintercalation from NaVO2. The possibility to control the vanadium oxidation state (and the sodium amount in the sodium layer) and the topology of the electrostatic interactions between Na+ and vanadium ions opens new avenues of investigation for solid state physics and condensed matter studies of conductivity mechanisms at play in these compound types.
■
■
Article
AUTHOR INFORMATION
Corresponding Author
*M. Guignard. E-mail:
[email protected]. Present Address ¶
Department of Chemistry, University of Liverpool, Liverpool, Merseyside L69 7ZD, United Kingdom.
Funding
Financial support was provided by the CNRS, Région Aquitaine and a grant from Agence Nationale de la Recherche (Blanc Inter II, SIMI 8) no. 2011-IS08-001-01. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS Use of the Advanced Photon Source at Argonne National Laboratory was supported by the U.S. Department of Energy under Contract No. DE-AC02-06CH11357. 1547
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548
Chemistry of Materials
Article
(29) Mackenzie, K. J. D.; Smith, M. E. Multinuclear Solid-State Nuclear Magnetic Resonance of Inorganic Materials; Pergamon: Oxford, U. K., 2002). (30) Carlier, D.; Ménétrier, M.; Grey, C. P.; Delmas, C.; Ceder, G. Phys. Rev. B 2003, 67, 174103. (31) Gourdon, O.; Evain, M.; Jobic, S.; Brec, R.; Koo, H.-J.; Whangbo, M.-H.; Corraze, B.; Chauvet, O. Inorg. Chem. 2001, 40, 2898−2904. (32) Kanke, Y.; Izumi, F.; Morii, Y.; Akiba, E.; Funahashi, S.; Kato, K.; Isobe, M.; Takayama-Muromachi, E.; Uchida, Y. J. Solid State Chem. 1994, 112, 429−437. (33) Kajita, T.; Kanzaki, T.; Suzuki, T.; Kim, J. E.; Kato, K.; Takata, M.; Katsufuji, T. Phys. Rev. B 2010, 81, 060405. (34) Rovira, C.; Whangbo, M.-H. Inorg. Chem. 1993, 32, 4094−4097. (35) Goodenough, J. B. Czech J. Phys. B 1967, 17, 304−336. (36) Huang, Q.; Foo, M. L.; Lynn, J. W.; Zandbergen, H. W.; Lawes, G.; Wang, Y.; Toby, B. H.; Ramirez, A. P.; Ong, N. P.; Cava, R. J. J. Phys.: Condens. Matter 2004, 16, 5803−5814. (37) Williams, A. J.; Attfield, J. P.; Foo, M. L.; Viciu, L.; Cava, R. J. Phys. Rev. B 2006, 73, 134401. (38) Didier, C.; Guignard, M.; Darriet, J.; Delmas, C. Inorg. Chem. 2012, 51, 11007−11016.
1548
dx.doi.org/10.1021/cm403114k | Chem. Mater. 2014, 26, 1538−1548