Spin‐Ladder Iron Oxide: Sr₃Fe₂O₅

Reduction with CaH2 of double-layered perovskite Sr3Fe2O7 yielded novel two-legged S=2 ladder compound Sr3Fe2O5. Together with the synthesis of SrFeO2, this opens up new avenues for solid-state chemistry and physics of a series of n-legged ladders Srn+1FenO2n+1. The picture illustrates the structural transformation from FeO6 octahedra in Sr3Fe2O7 to FeO4 squares in Sr3Fe2O5, which occurs via intermediate Sr3Fe2O6 (Sr blue, O red, Fe yellow). Spin ladders, which conceptually are unidirectional sections of an antiferromagnetic (AF) two-dimensional (2D) square lattice, as schematically illustrated in Figure 1, have attracted considerable attention in the last two decades. Theories predict that the ground state of an S=1/2 ladder is a gapped singlet spin liquid state when the leg number n is even, but is a gapless singlet spin liquid state when n is odd.1 Moreover, short-range AF spin correlation should lead to superconductivity when modest carriers are doped in even-legged ladders. Experimental counterparts are the observation of the gapped and gapless ground states of SrCu2O3 (n=2) and Sr2Cu3O5 (n=3), respectively,2 and the appearance of superconductivity in (Sr,Ca)14Cu24O41+x (n=2) under high pressure.3 Another S=1/2 two-legged Cu2+ ladder, namely, La1−xSrxCuO2.5, shows anomalous behavior associated with its nearly critical ground state.4 Spin ladders with n legs; solid and open circles denote magnetic ions and oxide ions, respectively. a) One-legged spin ladder (i.e., the 1D chain). b) Two-legged spin ladder realized in the present work. c) Three-legged spin ladder. d) ∞-legged spin ladder (i.e., the 2D square lattice) realized in SrFeO2.9 The motivation of theoretical investigations has been to test how the one-dimensional (1D) S=1/2 AF chain system (n=1), which is rigorously solved even when doped with carriers, can be related to the 2D square lattice (n=∞), which presents various theoretical difficulties and is far from being understood. Experimentally, Srn−1CunO2n−1 is the only example of a generalized spin-ladder system,5 though the composition and structure of a two-legged ladder has been extended to CaV2O5, Cu2(C5H12N2)Cl4, (C5H12N)2CuBr4, [Ph(NH3)([18]crown-6)][Ni(dmit)2] (dmit=dithiolene) and [(DT-TTF)2][Au(mnt)2] (DT-TTF=dithiophene tetrathiafulvalene, mnt=maleonitrile dithiolate).6 It is highly desirable from both theoretical and experimental viewpoints to extend these ladder systems with respect to n and S, and also with respect to possible skews such as mixing of ferromagnetic and AF interactions. An example of skew is the dramatic switching of the large-gapped (>400 K) singlet spin liquid of SrCu2O3 to an AF ordered state by the nonmagnetic substituent Zn2+, even at Zn/Cu≤1 %.7 The most pronounced structural characteristic of d9 copper(II) oxides is the favored formation of square-planar CuO4 units, which are stabilized due to the Jahn–Teller effect.8 This 2D coordination geometry provides strong Cu-O-Cu superexchange interactions within a ladder along the legs and rungs, while Cu⋅⋅⋅Cu interactions normal to the square plane (∥z) are negligibly small because the copper(II) ions have a magnetically inert, filled (d↑↓) configuration along this direction and also because an intervening oxygen atom is lacking. Unlike the case of cuprates, the coordination geometries in iron oxides have been almost exclusively restricted to 3D polyhedra such as octahedra and tetrahedra. However, this restriction was recently overcome9 by using calcium hydride at low temperatures as a reductant, as initiated and developed by Hayward, Rosseinsky and co-workers.10, 11 Low-temperature reaction of cubic perovskite SrFeIVO3 with CaH2 led to stable SrFeIIO2 with a square-planar oxygen coordination environment around the high-spin Fe2+ ion. The structure is isostructural with the “infinite”-layer cupric oxides.12 Herein we report the synthesis of novel spin-ladder iron oxide Sr3FeII2O5 through reaction of double-layered perovskite Sr3FeIV2O7 with CaH2. Together with the synthesis of SrFeO2, this opens up new avenues for the solid-state chemistry of iron(II) oxides with square-planar coordination, which potentially includes the serial ladder system Srn+1FenO2n+1, and for the solid-state physics of multiple-spin ladders. The slightly oxygen-deficient phase Sr3Fe2O7−y (y≈0.4), which can be prepared easily by a conventional, high-temperature solid-state reaction, was used as precursor. It is known that all oxygen vacancies are located at the apical O(1) sites shared by the double FeO2 sheets (Figure 2 a and b) and that the structure keeps the I4/mmm space group over the entire range from Sr3Fe2O7 (y=0) to Sr3Fe2O6 (y=1).13 The lattice parameters of the as-obtained precursor determined by powder X-ray diffraction (XRD) were a=3.872 and c=20.157 Å, corresponding to a phase with Sr3Fe2O6.6 stoichiometry. Note that Sr3FeIII2O6 (y=1) is thought to represent the lower end of the oxygen stoichiometry, in spite of intensive studies owing to their interesting properties such as metal–insulator transition, charge disproportionation, and possible applications as oxygen-separation membranes and solid-oxide fuel cells.14 Structural transformation from Sr3Fe2O7 to Sr3Fe2O5 via Sr3Fe2O6, where the blue, red, and yellow spheres represent Sr, O, and Fe atoms, respectively. a) Crystal structure of a stoichiometric (fully oxidized) phase Sr3Fe2O7 (y=0). b) Crystal structure of reduced phase Sr3Fe2O6 (y=1), once thought to be the lower limit of the oxygen content. c) Left: Crystal structure of the new phase Sr3Fe2O5 (y=2), described herein. Right: A spin ladder in Sr3Fe2O5 viewed along the a axis. d) Transformation from the octahedron (left) in Sr3Fe2O7 to the pyramid (middle) in Sr3Fe2O6 and to the square plane (right) in Sr3Fe2O5, where the white spheres represent oxygen vacancies. Reduction of Sr3Fe2O6.6 with CaH2 was performed in an evacuated glass tube at 623 K for 3 d. As shown in Figure 3, the room-temperature synchrotron powder XRD pattern of the final product was readily indexed with an I-centered orthorhombic unit cell with a=3.51485(2), b=3.95271(2), and c=20.91251(10) Å, except for a very minor unknown impurity (see Figure S1 in the Supporting Information). Compared to Sr3Fe2O6.6, the b axis remained nearly unchanged, while the a axis (c axis) substantially decreased (increased). The a axis is of similar length to the c axis of SrFeO2 (a=b=3.99 Å, c=3.47 Å).6 These observations strongly suggested that the FeO2 in-plane oxygen atoms O(2) along the a axis were all removed, while the apical oxygen sites O(1) were filled up, resulting in the formation of Sr3Fe2O5 crystallizing in space group Immm (Figure 2 c). The Rietveld refinement based on this model immediately converged to Rwp=0.0524 and χ2=5.28 along with reasonable individual isotropic displacement factors for all atoms. The bond valence sum calculation gave +1.88, +1.86, and +1.95 for Sr(1), Sr(2), and Fe, consistent with the expected valences of +2 for these elements. Neutron powder diffraction (NPD) analysis at 293 K confirmed the above structure model, with an excellent convergence of Rwp=0.0376 and χ2=2.78 (see Figure S2 in the Supporting Information). We tried to refine the oxygen occupancies and found that all O(1), O(2), and O(3) sites are fully occupied within 1 % error, and no extra oxygen atoms could be found. It also excluded possible hydrogen incorporation into the lattice, as found in LaSrCoO3H0.7.11 The three types of FeO bonds (dFeO(1)=2.013 (one bond), dFeO(2)=1.976 (two bonds), dFeO(3)=2.039 Å (one bond)) form a nearly square FeO4 plane. Structural characterization of Sr3Fe2O5 by Rietveld refinement of synchrotron XRD data at 293 K. The overlying crosses, solid lines, and ticks represent the observed and calculated intensities and the position of the calculated Bragg reflections, respectively. The difference between the calculated and observed profiles is plotted at the bottom. Sr3Fe2O5 adopts the Immm space group (No. 71), a=3.514825(2), b=3.95271(2), c=20.91252(10) Å, Z=2. Sr(1) on 2c (0.5,0.5,0), Sr(2) on 4i (0,0,0.31234(2)), Fe on 4i (0,0,0.09636(2)), O(1) on 2a (0,0,0), O(2) on 4j (0.5,0,0.59574), O(3) on 4i (0,0,0.19358(2)), with 100 % occupancy, Biso(Sr(1))=0.359(22) Å2, Biso(Sr(2))=0.213(13) Å2, Biso(Fe)=0.089(16) Å2, Biso(O(1))=0.12(11) Å2, Biso(O(2))=0.15(6) Å2, Biso(O(3))=0.59(8) Å2, Rp=0.0401, Rwp=0.0524,and χ2=5.28. Refining the occupation factors at the oxygen sites and the vacant site did not improve the result. Furthermore, Mössbauer measurements (Figure 4 a) indicated not only that all the Fe atoms are electronically and crystallographically equivalent, but also that they are most likely square-planar coordinated in high-spin state (S=2), since the obtained isomer shift (IS) of 0.46 mm s−1, quadrupole splitting ΔE of 1.28 mm s−1 at 297 K, and the magnetic hyperfine field of Hhf=43.7 T at 4 K are close to those of SrFeO2.9 These observations are consistent with the structural analysis. Thermogravimetric (TG) measurements showed the oxygen content of the product to be less than 6 (i.e., y>1) though its precise determination was hampered by partial reactions (oxidation, decomposition, hydration) of Sr3Fe2O5 while it was exposed to air for the experimental setup (see Figure S3 in the Supporting Information). Magnetic order in Sr3Fe2O5. a) Mössbauer spectra at 300 (top) and 4 K (bottom). The circles indicate the experimental data, while the lines denote the fits. b) Rietveld refinement of NPD data at 10 K measured at 1.91 Å. The solid lines and the overlying crosses and bars indicate the calculated intensities, the observed intensities, and the positions of the calculated chemical (top) and magnetic (bottom) Bragg reflections. The difference between the observed and calculated profiles is plotted at the bottom. See Tables S1 and S2 in the Supporting Information for details. When half of the O(2) sites in Sr3Fe2O7 are removed in the present ordered manner, the original double perovskite layers become decoupled into ladders with two legs running along [010] and rungs along [001] (Figure 2 c). Sr3Fe2O5 and SrFeO2 are thus two- and ∞-legged S=2 ladder compounds, respectively. The sectioning of the dense FeO2 sheet in SrFeO2 into the two-legged ladders loosely distributed in Sr3Fe2O5 should greatly change the magnetism, at least by drastically lowering the 3D magnetic ordering temperature. Indeed, the NPD pattern of Sr3Fe2O5 at 293 K does not contain any magnetic reflections, and the Mössbauer spectrum at 300 K just consists of a quadruple doublet, that is, Sr3Fe2O5 is still in the paramagnetic state, in contrast to the case of SrFeO2 with TN as high as 473 K. In the Mössbauer spectrum at 4 K (Figure 4 a), however, we found a sharp sextet, clearly indicative of long-range AF order. Furthermore, the NPD data at 10 K (Figure 4 b) revealed that the long-range AF order is characterized by a magnetic propagation vector q=(1/2, 1/2, 0) and that the iron moments of 2.76(5) μB are aligned parallel to the c axis (see Figure 4 S in the Supporting Information). This so-called G-type antiferromagnetic structure is frequently observed in double-layered Ruddlesden–Popper phases such as Sr3Mn2O7.15 The fact that the Fe moment is reduced by more than 20 % from that of 3.6 μB for SrFeO2 must be a consequence of a quantum magnetic fluctuation enhanced inherently in the ladder, though future theoretical support is needed. Considering that Sr3Fe2O7 and SrFeO3 are the n=2 and n=∞ members of a homologous series Srn+1FenO3n+1, where n represents the number of perovskite blocks,13 we can predict a new homologous n-legged S=2 ladder series Srn+1FenO2n+1 (Figure 1). In this context, the reaction of Sr2FeO4 (n=1) and Sr4Fe3O10 (n=3) with CaH2 should result in the one-legged ladder (i.e., 1D chain) compound Sr2FeO3 and the three-legged ladder compound Sr4Fe3O7, respectively (see Figure S5 in the Supporting Information). This work is in progress. Note that (LaSr3)(Fe1.5Co1.5)O7.5, obtained by hydride reduction of the n=3 La- and Co-substituted sample (LaSr3)(Fe1.5Co1.5)O10, has a different framework composed of three-coordinate metal ions flanked by square-based pyramidal coordination,10d instead of the three-legged ladder solely with square-planar coordination. It is also interesting to compare the copper(II) and iron(II) ladder systems Srn−1CunO2n−15 and Srn+1FenO2n+1. In the former, the ladders, each having the composition CunO2n−1, share their OO edges to form the dense CunO2n−1 sheet. The iron(II) ladders, on the other hand, are dispersed in the I-centered orthorhombic lattice with their legs and rungs oriented along the b and c axes, respectively. The linear Fe-O-Fe bonds are strong and AF, as found for SrFeO2. As a consequence, the phase shift along the leg direction by π/2 between the corner and center ladders should minimize their interactions by spin frustration, as in the dense CunO2n−1 sheet. However, if the electronic configuration of the FeO4 unit is (dyz,dzx)3(dxy)1(d)1(d)1 with S=2, as proposed for SrFeO2, the iron(II) ladders facing each other in phase along the a axis should interact more strongly in an AF Fe(d↑)⋅⋅⋅Fe(d↓) manner than the corresponding Cu⋅⋅⋅⋅Cu interactions, because these copper(II) ions have an inert, filled (d↑↓) configuration. Numerical calculations for a S=1/2 two-legged ladder system showed that the gapped spin liquid state can be replaced by an ordered state at a critical value of J′/J≈0.11, where J and J′ are the intra- and interladder exchange interactions, respectively.4b There are no calculations for S=2 systems, but there is no doubt that the magnetic order in Sr3Fe2O5 results mainly from interactions along the a axis. The mechanism of reduction from Sr3Fe2O7 to Sr3Fe2O5 merits detailed inspection. We tried to slow down the process by lowering the synthesis temperature to 553 K. Interruption of the reaction after one day resulted in Sr3Fe2O6, while a one-week reaction at the same temperature yielded Sr3Fe2O5, that is, the process is two-staged. The first step from Sr3Fe2O7 to Sr3Fe2O6 is nothing but continuous randomized removal of the O(1) ions leading to a coordination change from FeO6 octahedra to FeO5 square pyramids. This is quite common for the n=2 Ruddlesden–Popper phases and there is nothing special in the present case. However, the second step, the transformation of the FeO5 square pyramids in Sr3Fe2O6 into the FeO4 squares in Sr3Fe2O5 must be more complicated and cooperative. Supposing one basal oxygen atom of a given square pyramid escapes into the reducing atmosphere, another basal oxygen atom must move up to the vacant apical site, as illustrated in Figure 2 d, and these processes should take place cooperatively over a certain lattice volume for crystallization. Thus, the reduction mechanism changes at y=1. It is noteworthy that there are neither intermediate phases nor nonstoichiometry between Sr3Fe2O6 and Sr3Fe2O5, as well as between SrFeO2.5 and SrFeO2. This is remarkable given the presence of two intermediates, SrFeO2.875 and SrFeO2.75, between SrFeO3 and SrFeO2.5. It is also known that electrochemical formation of SrCoO3 from SrCoO2.5, which are isomorphous with SrFeO3 and SrFeO2.5, respectively, proceeds topotactically at room-temperature via the complex vacancy-ordered intermediate phase SrCoO2.82±0.07.16 The processes occurring in the formation of SrFeO2 from SrFeO2.5, whereby the FeO4 tetrahedra and the FeO6 octahedra linked to each other are both converted to FeO4 squares17 must be more complicated than the case discussed above. We now recognize that the underlying perovskite and related lattices are flexible enough to tolerate such drastic compositional and structural events. We thus believe that oxygen-transport materials working at low temperatures that feature formation of intermediate ordered structures should be promising. Finally, we must assume that the present synthetic strategy involving CaH2 reduction can be further generalized for a more rational design of new magnetic lattices comprising extended arrays of FeO4 square planes, for which formerly only copper(II) oxides were candidates. It would also be interesting to test other transition metal oxides. Precursor Sr3Fe2O7−y (y≈0.4) was prepared by a conventional high-temperature ceramic method from predried SrCO3 (99.99 %) and Fe2O3 (99.99 %) by heating a pelletized stoichiometric mixture of these at 1273 K in air for 24 h and again for 24 h at 1473 K after grinding and pelletization. The reduction of Sr3Fe2O7−y (y≈0.4) was performed with CaH2 as reducing agent. Sr3Fe2O7−y (0.43 g) and a four-molar excess of CaH2 (0.15 g) were finely ground in an Ar-filled glove box, sealed in an evacuated pyrex tube (V=15 cm3) with a residual pressure of less than 1.3×10−2 Pa, and heated at 623 K for 3 d. Residual CaH2 and the CaO byproduct were removed from the final reaction phase by washing with 0.1 M NH4Cl in dried methanol. Sr3Fe2O5 is air-sensitive, in contrast to SrFeO2. Thermogravimetric measurements were performed with a thermal analyzer (Rigaku Thermoplus TG8120). Measurements to analyze reoxidation behavior of Sr3Fe2O5 were performed on a sample of around 10 mg that was rapidly loaded into a platinum crucible and then heated at 10 K min−1 in air. Prior to the experiment, the sample was dried at 373 K for about 1 h. The identity of the reoxidized product was determined by powder X-ray diffraction. The synchrotron powder XRD experiment was performed on the large Debye–Scherrer camera installed at SPring-8 BL02B2 by using an imaging plate as detector. Incident beams from a bending magnet were monochromatized to 0.77709 Å. The sample was contained in a glass capillary tube with an inner diameter of 0.1 mm and was rotated during measurements. The diffraction data were collected at room temperature in a 2 θ range from 1 to 75° with a step interval of 0.01°. The NPD studies were carried out on the D1A diffractometer installed at the Institute Laue Langevin (Grenoble, France). A 250-mg sample sealed in an He-filled vanadium can was used, and a wavelength of λ=1.91 Å was employed. Mössbauer spectra were taken in transmission geometry by using a 57Co/Rh γ-ray source kept at room temperature on a powdered sample of Sr3Fe2O5. The source velocity was calibrated by using pure α-Fe as control material, and the isomer shift is relative to α-Fe. The spectra were fitted by using the Lorentzian function. The XRD and NPD patterns were analyzed by the Rietveld method with RIETAN 200018 and FULLPROF software,19 respectively. The agreement indices used were the weighted profile Rwp=[∑wi(yio−yic)2/∑wi(yio)2]1/2 and the goodness of fit (GOF), χ2=(Rwp/Rexp)2, where Rexp=[(N−P)/∑wiyio2]1/2, yio and yic are the observed and calculated intensities, wi is the weighting factor, N the number of yio data when the is and the number of The bond valence sum method was to the valence of by using Supporting for this is on the under from the The is not for the content of any by the than should be to the corresponding for the