2019, Vol.92, No.8

The Chemical Society of Japan Award for Creative Work for 2016

The discovery of new extended structures has often led to the development of new fields in chemistry and physics. However, the numerous combinations of metals (or cations) to yield new materials have been largely exhausted. Materials development based on the anion-centered strategy allows us to access several new classes of materials, such as iron oxides with square-planar coordination and mixed-anion oxides including oxyhydrides.

Inorganic (ionic) materials such as oxides are widely used in various applications, such as catalysts, magnets, electronic devices, superconductors, and batteries, and thus form the basis upon which our technological society rests.2 The discovery of new compounds and materials such as copper oxide superconductors always served as breakthroughs for chemistry and physics. However, conventional solid-state techniques to develop new inorganic materials (typically oxides) is reaching a point where all experimental conditions have been largely exhausted, making further development increasingly difficult. In oxides, the numerous combinations of various elements (i.e. metals, or cations) to yield new materials have been largely exhausted. A considerable amount of inorganic material discovery is based on solid state reactions at high temperature (>1000 °C). However, this approach lacks the step-by-step designability encountered in organic synthesis, making the rational design of complex materials difficult.

Focusing on the anionic side, though, offers a solution to this problem. As most anionic species are light elements which are volatile/gaseous in elemental or molecular form, many new reactions can be envisioned. Furthermore, just as physical properties have been controlled by cation choice, anions may also be selected to control various properties, adding a new dimension of flexibility. In this Highlight Review, we provide a brief survey of research activities in our group at Kyoto over the past decade, carried out with anion-centered strategy, in particular, focusing on oxypnictide superconductors, oxyhalide two-dimensional magnets, square-planar coordinate iron oxides, and oxyhydrides with unique properties derived from H anions. Several review articles related to each topic could be found elsewhere.38

Titanium pnictide oxides BaTi2Pn2O (Pn = As, Sb, Bi) have a layered structure comprising Ti2Pn2O layers separated by Ba cations, as shown in Figure 1.9 The Ti3+ cations with heteroleptic octahedral coordination of TiO2Pn4 forms a Ti2O square lattice, which is an anticonfiguration to the CuO2 square lattice in high-Tc cuprates. The electronic configuration of 3d1 is complementary to Cu2+ (3d9).

BaTi2As2O shows metallic behavior without any trace of superconductivity, but with anomalies in magnetic susceptibility and electrical resistivity at 200 K, which was ascribed to a charge-density-wave (CDW) transition.10 We synthesized new antimonide and bismuthide, BaTi2Sb2O and BaTi2Bi2O,11,12 and observed a significant reduction of CDW transition to TCDW = 50 K for BaTi2Sb2O, while a CDW instability completely disappears in BaTi2Bi2O. The CDW suppression/elimination led to superconductivity with Tc = 1.2 K (BaTi2Sb2O) and 4.6 K (BaTi2Bi2O). Specific heat, 121/123Sb-nuclear quadrupole resonance (NQR), and muon spin relaxation measurements on BaTi2Sb2O revealed a full-gap s-wave state, with a novel microscopic coexistence of superconductivity and CDW instability.1315

The isovalent anionic solid solution of BaTi2(AsxSb1−x)2O and BaTi2(Sb1−yBiy)2O displays an unusual electronic phase diagram,16,17 as shown in Figure 2. The Sb-for-As substitution (x = 1 → 0) continuously decreases TCDW and the superconducting phase appears with a Tc maximum at around y = 0.2. With a further increase of y, the Tc exhibits a sudden jump to 4.5 K at y = 0.60 and remains nearly constant to the terminal composition of BaTi2Bi2O (y = 1). It is thus clear that there are two distinct superconducting phases, SC1 for 0 ≤ x ≤ 0.1 and 0 ≤ y ≤ 0.55, and SC2 for 0.6 ≤ y ≤ 1. A similar two-dome structure in Tc has also been reported in the iron pnictide superconductor LaFeAs(O1−xHx), which was ascribed to the multiband structures by Fe-3d orbitals.18 Likewise, the two-dome structure of our titanium pnictide oxides appears to link with a multiband character at the Fermi level, which is indeed supported by theoretical studies, where Ti 3dxy, 3dx2−y2, and 3dz2 orbitals are occupied at the Fermi level EF.1922

The overall electronic phase diagram (Figure 2) strongly suggests that the superconductivity in the SC1 phase is driven by the suppression of CDW phase, as often seen in other superconductors. Theoretically, a lattice instability associated with the formation of \(\sqrt{2} \) × \(\sqrt{2} \) × 1 superstructure is predicted in BaTi2Sb2O.22,23 Experimentally, however, no superlattice has been observed below TCDW, implying that the CDW state in BaTi2Sb2O is unconventional. Instead, the NQR/NMR spectra of BaTi2Sb2O provided clear evidence of broken in-plane four-fold symmetry,14 whereas high-resolution neutron diffraction study detected a small lattice distortion to the orthorhombic symmetry with the orthorhombicity, η = 2 × (ab)/(a + b), of 0.22% and 0.05% (at 20 K) for BaTi2As2O and BaTi2Sb2O, respectively.24 The observed distortions are much smaller than typically observed in Jahn-Teller compounds and iron pnictides. It means that the electronic instability drives the lattice distortion. One possible scenario to account for the experimental results is an intra-unit-cell nematic charge order or an orbital order.21,24

Low-temperature topochemical reactions such as intercalation and ion-exchange reactions enable a rational design of new types of structures which cannot be obtained by conventional high-temperature solid state reactions.25 A wide range of oxide materials have been investigated as precursors for such reactions.26 In particular, the Dion-Jacobson (DJ) type layered perovskite has manifested as a platform of various topochemical reactions.25,27 The DJ phase has a general chemical formula given by A′An−1BnO3n+1, where A′ is an alkali metal, A is an alkaline earth or rare earth metal, B is a d0 transition metal, and n is the number of perovskite layers. The interlayer A′ cations are highly reactive and subject to ion-exchange reactions, while the perovskite blocks are chemically inert. Most notably, Wiley et al. demonstrated a co-exchange of cations and anions within the interlayer space;28 a reaction of the non-magnetic oxide RbLaNb2O7 (n = 2) with CuCl2 results in a magnetic compound (CuCl)LaNb2O7 (Figure 3). This reaction can be extended to n = 3 and 4 phases.2934 In (CuX)An−1BnO3n+1 (X = Cl, Br), the Cu2+ cations are trans-coordinated with two apical oxide ions from the perovskite blocks and four halide ions within the ab plane. As a result, the CuO2X4 octahedra share edges to form a square-lattice network, as in the high-Tc cuprates. The co-insertion of halide anions allows super-exchange pathways within the CuX layer. These CuX layers are spatially separated by the non-magnetic perovskite blocks, with the interlayer distance of ∼12 Å for n = 2, ∼16 Å for n = 3, and ∼20 Å for n = 4. These features make (CuX)An−1BnO3n+1 a promising candidate for 2D quantum magnetic phenomena, or even superconductivity if holes/electron can be donated.

We have found novel quantum magnetic states in (CuX)LaNb2O7.28 The magnetic susceptibilities (Figure 4) exhibit a broad maximum centred at 16.5 K (X = Cl) and 36 K (X = Br),35,36 a characteristic feature of low-dimensional antiferromagnets. While the bromide undergoes a magnetic transition at TN = 30 K with a stripe spin texture,36 the chloride does not show any anomaly associated with magnetic transition. Inelastic neutron scattering experiments on (CuCl)LaNb2O7 revealed a spin-singlet ground state and a triplet excitation gap of 2.3 meV.35 The triplet state is destabilized by applying a magnetic field and when it reaches the spin-singlet ground state at 11 T, a field-induced transition involving Bose-Einstein condensation of magnons occurs.37,38 Cationic and anionic substitution lead to the observation of novel phenomena including a quantum phase separation.3941 The observed properties were initially interpreted in terms of a frustrated square-lattice, or the J1-J2 model with an antiferromagnetic nearest-neighbour interaction J1 and a ferromagnetic next-nearest neighbour interaction J2 (see Figure 3).42 However, the in-depth structural analysis revealed the presence of cooperative tilting of the NbO6 octahedra in the perovskite block, which lead to a large displacement of Cu and X atoms,4345 giving a 2ap × 2ap × cp superlattice (space group Pbam). To our surprise, first principles calculations show that the fourth-nearest Cu–Cl–Cl–Cu pathway, J4, is responsible for the spin dimer formation (Figure 5), and together with a ferromagnetic component, the magnetism of (CuCl)LaNb2O7 can be best described by a ferromagnetic version of the Shastry-Sutherland model.4648

Triple-layered (CuCl)Ca2Nb3O10 and (CuBr)Sr2Nb3O10 have a non-magnetic and magnetic (helical) ground state, respectively.4951 (CuBr)Sr2Nb3O10 has a 1/3 magnetization plateau for 2.0 < H < 8.1 T (Figure 6),50 which is unprecedented since a 1/2 plateau is expected for the J1-J2 model.52 Neutron diffraction experiments up to 10 T show that a competition between the two magnetic phases with propagation vectors of k = (0, 0, 0) and (0, 1/3, 0.46) incidentally induces the 1/3 plateau.53 Compounds with other 3d transition ions (thus having different spin degrees of freedom) were also studied. For example, (MCl)LaNb2O7 (M = V, Cr, Mn, Fe, Co) exhibit various spin ordered states including stripe order, G-type order, and partially antiferromagnetic state.5458

In this way, topochemical reactions offer extensive, rational and flexible design of 2D magnetic system, leading to discoveries of novel physical properties. This approach contrasts with the conventional search for materials via solid state reaction, in the sense that the systematic understanding and controlling is possible by varying constituent, M, X, A, B and n.

Square planar coordination is often found in oxides containing d4 or d9 electronic configurations, most notably copper oxide high-Tc superconductors with the CuO2 square-lattice layer. Iron oxides prepared by conventional synthetic techniques are usually found with iron in octahedra, pyramids and tetrahedra. In 2007, SrFeO2 with iron in square planar coordination with the “infinite-layer” structure was prepared from a low-temperature (280 °C) topochemical reduction of SrFeO3 perovskite,59 using the CaH2 reduction method developed by Hayward and Rosseinsky.60 FeO4 square planes share corners to construct the square lattice, as in cuprate high-Tc superconductors. The FeO2 layers are separated by Sr2+ cations (Figure 1). The iron is in a divalent state with a high-spin configuration (S = 2) and has the electronic configuration of \((d_{z^{2}})^{2}(d_{xz},d_{yz})^{2}(d_{xy})^{1}(d_{x^{2} - y^{2}})^{1}\).61,62 Only a handful (distorted) square planar coordinate iron oxides (e.g., gillespite BaFeSi4O10) were known previously, but with the FeO4 units being isolated with respect to each other.63,64 The applied topochemical hydride reduction has opened a pathway to square coordinate iron oxides. For example, hydride reduction of Ruddlesden-Popper perovskite oxides of Sr2FeO4 and Sr3Fe2O7 resulted in Sr2FeO3 and Sr3Fe2O5, which is, together with SrFeO2, given by Srn+1FenO2n+1 (n = 1, 2, ∞; Figure 7).32,65

Despite the apparent 2D structure, SrFeO2 undergoes an antiferromagnetic order at a high temperature of TN = 473 K. However, the subsequent experimental and theoretical studies have revealed that Fe moments are indeed strongly coupled three dimensionally, via in-plane superexchange interactions and out-of-plane direct exchange interactions.61,62 This contrasts with the isostructural SrCuO2 with the quasi-2D magnetic correlations. The effect of dimensional reduction is observed in Sr3Fe2O5 and Sr2FeO3 with TN = 296 K and 179 K, respectively. A strong spin-lattice coupling SrFeO2 giving a transverse displacement of Fe and O atoms is observed at temperatures close to TN.66

The isovalent Sr-site substitution in SrFeO2 provides a substantial impact on the structure. CaFeO2 crystallizes in a distorted infinite-layer structure with Ca2+ in a six-fold coordination (vs eight-fold coordination at Sr2+), which is enabled by distortion and coherent rotation of FeO4 square-planar units (Figure 8a).67,68 The barium substitution beyond 30% induces a transition to a new structural type related to LaNiO2.5 (Figure 8b).69 The FeO4 square-planar coordination in CaFeO2 is distorted toward tetrahedral coordination, while it is rhomboidally distorted in BaFeO2. Aliovalent substitution of Sr with Nd3+, Sm3+ and Ho3+ results in a charge compensation by forming (Sr1−xLnx)Fe2+O2+y/2. The inserted oxygen induces a buckling of the FeO2 layers.70 These observations show the flexible nature of FeO4 square-planar coordination. Such a distortion is hardly found in cuprates, where only dx2−y2 orbitals are active. In other words, the ability of square-planar Fe2+ to distort largely is related to the multi-orbital nature. Fe-site substitution brings about interesting magnetic behaviour. Mn2+ substitution gives rise to a “random fan-out state” induced by site-random interlayer couplings.71 Square planar iron can also assist other transition metals to adopt square planar coordination. Several 1:1 solid solutions are reported, including SrFe0.5Ru0.5O2 with Ru2+ in an intermediate spin state (S = 1)72 and Sr2FeIrO4 with Ir2+ in a low-spin state (S = 1/2).73

Among mixed-anion compounds, oxyhydrides (oxide-hydride) have attracted considerable attention in recent years. The Oxford group successfully synthesized the first oxyhydride with a transition metal, LaSrCoO3H0.7, using a topochemical reaction of Ruddlesden-Popper LaSrCoO4 with CaH2 (Figure 9b).74 The replaced hydride anions are positioned at the equatorial site, in contrast to oxychlorides and oxybromides of this structural type.75 The double-layered analogue, Sr3Co2O4.33H0.84, was also obtained from Sr3Co2O7−δ.76 The very low valence of Co1.7+ in these compounds is understood given the presence of highly reductive H anions. Although mixed-valenced, the Co-based oxyhydrides are electronically insulating. We applied the synthetic strategy to BaTiO3 perovskite and obtained BaTiO3−xHx (x ≤ 0.6) (Figure 9a) where H and O2− are disordered at the anion site.77 ATiO3−xHx (A = Sr, Ca, Eu) were also prepared from the corresponding oxides.78,79 Unlike the cobalt-based oxyhydrides, Ti takes a ubiquitous valence of +3.4–+4,80 as seen in Magnéli phases TinO2n−1, and ATiO3−xHx exhibit a metallic conductivity.

SrVO2H, Sr2VO3H, Sr3V2O5H2, and LaSr3NiRuO4H4 have an anion-ordered perovskite-type structure with trans-MO4H2 octahedra (Figure 9).8183 The trans geometry lifts the triple degeneracy of vanadium t2g orbitals and the on-site electron-electron repulsion further splits the low-lying half-filled dxz/dyz orbitals, making these materials a Mott insulator. DFT calculations on SrVO2H show two-dimensional electronic structures.83,84 The V–H–V interaction along the c axis is much weaker than the in-plane V–O–V interaction owing to the orthogonality between V t2g and H 1s orbitals. The hydride anion can effectively block the out-of-plane interactions even under high pressure of ∼50 GPa.

High-pressure synthesis is also a useful approach to synthesize oxyhydrides.8588 A magnanese oxyhydride of LaSrMnO3.3H0.7 with the Ruddlesden-Popper perovskite structure is only accessible by taking a high-pressure route;86 otherwise, oxygen deficient oxide LaSrMnO4−δ is formed. A cubic perovskite oxyhydride SrCrO2H (Figure 9a) can be synthesized at 1000 °C and 5 GPa. This phase shows G-type spin ordering at TN ∼ 380 K,85 which is higher than those of isoelectric perovskite oxides RCr3+O3 (R = rare earth element) and any other Cr(III)-containing oxides. The enhancement of TN in SrCrO2H (vs RCrO3) is reasonably explained by the increased tolerance factor (t = 1). Such tolerance-factor control via cation/anion co-exchange (Sr2+/H \( \Leftrightarrow \) La3+/O2−) could be a promising strategy for extensive tuning of structures and resultant magnetic and other properties. Although known transition-metal oxyhydrides had been limited to those having vertex-linked octahedral networks, BaVO3−xHx synthesized under 5 GPa adopted a hexagonal structure with corner- and face-shared octahedral coordination (Figure 9e).88 For x = 0.3, H anions are located selectively at the face-shared sites, which appears to suppress electron hopping along the c axis (or confine electrons within the 2D sheet), leading to an increased electron-electron correlation as observed by NMR. A further pressurization to 7 GPa stabilizes a cubic perovskite structure (Figure 9a).

Many well-known mixed anion systems such as oxynitrides or oxyfluorides are prepared from oxide precursors, which can sometimes be a topochemical anion exchange reaction. For this review, we rephrase anion exchange reactions as topochemical, in order to emphasize that only the essential aspects of the crystal structure change under a pure anion exchange reaction. Classic examples are the topochemical fluorination of layered copper oxides as superconductors, while more recent examples are the topochemical formation of oxynitrides from perovskite precursors with NH3 treatment. For oxynitrides, NH3 is the common nitriding reagent, while fluorination can be conducted with F2, NH4F, or solid fluoropolymer powders, and hydride-based reactions usually consist of introduction by metal hydrides as we will see below.

Topochemical nitridation: While numerous oxynitrides have been reported, the compounds which are prepared by anion exchange, i.e., topochemically, are a minority. For example, the tantalate and niobate oxynitrides are typically prepared by alkali earth carbonates and oxides under flowing NH3;89,90 this may be due to the rather inflexible nature of niobate and tantalate in terms of redox, as replacing O2− with H requires a change in cation valence. With other elements, such as Fe, this is not always a problem, as the topochemical nitridation of Sr2FeMoO6 to Sr2FeMoO4.9N1.1 using NH3 has been reported.91 Additional topochemical reactions to form oxynitrides, however, seem to be possible with oxyhydride precursors as shown below.

Topochemical fluorination: With fluorination reactions, typically there can be a number of variations. For example, a one-to-one exchange of O2− and F results in reductive fluorination, often seen with low fluorine doping levels or the use of fluoropolymers under anaerobic conditions as for the synthesis of RbLaNb2O6F92 or superconducting WO2.6F0.4.93 Simple anion exchange without any redox requires two fluorides to be inserted with the removal of one oxide anion; this appears to be the case of the fluorination for some brownmillerites LaACuGaO5 (A = Sr, Ca) when using XeF2 as a fluorinating agent.94 Alternatively, oxidative fluorination can occur, where extra fluoride insertion accompanies (or even may solely occur) exchange with oxide. An early example is the use of F2 gas for fluoride insertion/exchange to convert Sr2CuO3 to the superconducting Sr2CuO2F2+δ,95 and more recent examples are the preparation of the Sr2TiO3F2 and LaSrFeO3F3 phases,96,97 both from their oxide Ruddlesden-Popper phases. Fluorination can also be conducted after removing some lattice oxide in a preliminary separate reaction, as for the synthesis of Sr7Mn4O13F2 from Sr7Mn4O15.98 These distinctions between the various types of fluorination reactions have been summarized by Slater,99 and the various oxyfluorides prepared by fluorine (exchange) reactions have been reviewed by McCabe,97 as we turn to hydride-based anion reactions as this is where the new developments are.

Topochemical exchange reactions involving hydride: Many of the recent metal oxyhydrides, such as C12A7:H,100 LaSrCoO3H0.7,75 LaSr3NiRuO4H4,101 BaTiO3−xHx,78 and Sr2VO3H87 have been prepared by anion exchange reactions of oxide precursors. The hydrogen source in this case is often a metal hydride such as NaH or CaH2, rather than H2. The use of these strongly reducing hydride agents apparently permits lower reaction temperatures (typically 200–600 °C), making the synthesis of these materials possible. The scope of various oxyhydrides5 and some studies on the formation mechanisms77,102 has been reported elsewhere, so we now turn to further reactions using oxyhydrides as a starting point.

Compared to the anions such as oxide, fluoride, or nitride, systems with hydride offer unique possibilities. Certain metals or alloys containing Mg, Ni, etc., reversibly hydrogenate/dehydrogenate to very large extents at relatively moderate temperatures compared to equivalent materials for oxygen. While this effect is not observed at the same magnitude in oxyhydrides, the hydride anion within an oxide framework is still rather easily exchangeable by most standards. Work by Kobayashi et al.,78 Yajima et al.,103 and Masuda et al.104 explicitly examine this, starting with the oxyhydride BaTiO3−xHx. When heated under inert atmospheres, BaTiO3−xHx releases the lattice hydride as H2 gas at temperatures around 400 °C,78 reminiscent of TiH2 decomposition. Heating instead under a D2 atmosphere leads to H/D exchange at similar temperatures yielding BaTiO3−xDx.78 The precise exchange temperature and activation energies involved, together with its dependence on composition have been determined by Ya et al.105 Isotope exchange of 16O/18O of course has been demonstrated,106 but usually occurs at higher temperatures, so it is evident that hydride is more readily exchanged. This is probably related to the lighter mass and smaller charge of hydride in comparison to oxide.

This facile exchange can lead to chemical transformations otherwise difficult to achieve. For example, oxynitrides are typically made by treating under NH3 at strenuous temperatures; 900 °C is necessary to prepare BaTiO2N0.1.103,104 In contrast, treating the oxyhydride BaTiO2.5H0.5 with NH3, or even N2 at 400 °C results in higher N contents (BaTiO2.5N0.3).103,104 Other than direct H/N3− exchange, the vacancies created by the leaving hydride species can also result in enhanced nitrogen content during thermal processing; this was demonstrated by Mikita et al. during the synthesis of EuTiO2N from EuTiO3−xHx.107 The anion-vacancy enhanced H/O exchange is also shown by mimicking the well-known strategy to improve ionic conductivity.108 Other than being thermolabile, hydride is also susceptible to attack from acidic groups as such as HF; and this has been utilized to prepared BaTiO2.5H0.25F0.25 from the oxyhydride BaTiO2.5H0.5 at fairly moderate conditions.104 This product is closely related to the reported BaTiO2.9F0.1, which we note is accessible only by high pressure/temperature reaction (3 GPa, 1300 °C).109

Catalysis with mixed anion systems: While many of the studies have focused on the electronic and magnetic properties of these mixed anion systems, interesting chemical applications are also possible. Controlling the oxide/nitride, or oxide/halide ratio results in a varied optical band gap, and thus providing a means for optimizing photocatalytic activity.110 Apart from these catalysts, however, one can envision the constituent anions actually being part of the catalytic cycle. For pure oxides, this is already known as they often function as good oxidation catalysts.111 This naturally leads one to suspect oxyhydrides, and oxysulfides, etc. as having activity for new chemical transformations involving hydrogen or sulfur. This has been recently discovered for BaTiO2.5H0.5 and TiH2, which have been found to be active catalysts for the Haber-Bosch synthesis of ammonia from N2 and H2.112 The active site here is a Ti center on the surface, which contrasts conventional catalysts, where a supported metal such as Fe, Ru, or Co on a support acts as the active site. Other titanium oxides are shown to be inactive towards NH3 synthesis, so it is the presence of hydride in the lattice which makes this reaction possible.112 The oxyhydride can also be used as a catalyst support for Ru, Fe, Co, or Ni particles, leading to enhanced activities for NH3 synthesis113 and CO2 methanation reactions.114 This results are also mirrored by other catalytic supports such as LiH,115 electrides (which reversible convert in-situ between hydrides)116 and LnHO;117 in all of these cases, it is noteworthy that all of the reactions involved can be viewed as hydrogenation reactions. It is hence conceivable that in the future, other catalytic reactions involving H, N, S, and O will become possible or enhanced through the use of mixed anion systems.

Clearly, global warming caused by human influence, has been accelerating its effects during the last decade. This appears to be causing disasters across the world. To stop the global warming effect, it is crucial to introduce fundamental changes in the energy supply and consumption systems, for which new materials with game-changing functions are indispensable. We are hoping that these materials can be obtained through anion-inspired chemistry.

This work has been carried out with students and postdocs in Prof. K. Yoshimura Group at the Graduate School of Science (2003–2009) and the Kageyama group (2010–) and many other collaborations. Research funding includes CREST (JPMJCR1421) and JSPS KAKENHI (JP16H6439), FIRST project, Japan Society for the Promotion of Science (JSPS) Core-to-Core Program (A) Advanced Research Networks.

Hiroshi Kageyama

Hiroshi Kageyama received his PhD (1998) from Kyoto University. Afterwards he worked at the Institute for Solid State Physics, the University of Tokyo. In 2003, he became an associate professor at Department of Chemistry, Graduate School of Science, Kyoto University. In 2010, he was appointed to a full professor at Department of Energy and Hydrocarbon Chemistry, Graduate School of Engineering, Kyoto University. His research interests concern the discovery of new inorganic compounds with novel physical and chemical properties. He is the project leader of ‘Mixed-anion Project’ of MEXT Grant-in-Aid (Scientific Research on Innovation Areas) for 2016-2021.