Possible Origin of the Absence of Magnetic Order in LiOsO₃: Spin–Orbit Coupling Controlled Ground State

Abstract

LiOsO 3 is the first experimentally confirmed polar metal with ferroelectric-like distortion. One puzzling experimental fact is its paramagnetic state down to very low temperature with negligible magnetic moment, which is anomalous considering its 5d 3 electron configuration since other osmium oxides (e.g., NaOsO 3 ) with 5d 3 Os ions are magnetic. Here the magnetic and electronic properties of LiOsO 3 are re-investigated carefully using the first-principles density functional theory. The calculations reveal that the magnetic state of LiOsO 3 can be completely suppressed by the spin-orbit coupling. The subtle balance between significant spin-orbit coupling and weak Hubbard U of 5d electrons can explain both the nonmagnetic LiOsO 3 and magnetic NaOsO 3 . This work provides a reasonable understanding of the long-standing puzzle of magnetism in some osmium oxides.Correlated electron systems with appreciable Coulomb repulsion are one of the most attractive platforms for accessing a series of emerging physical properties such as metal-insulator transition (MIT), superconductivity, colossal magnetoresistance, multiferroicity, and so on, which are often technologically useful. Such a Coulomb repulsion is usually characterized by the on-site Hubbard U. On the other hand, spin-orbit coupling (SOC) in condensed matters is becoming highly concerned, evidenced within a lot of emergent quantum materials such as topological insulators, Weyl semi-metals, and Kitaev systems, [1][2][3][4] where SOC can be the core ingredient of physics underlying their novel physical phenomena. On one hand, SOC can be strong for heavy ions, and even comparable to U for 5d electrons. For example, SOC is believed to play a decisive role in determining the unconventional properties in those 5d 5 transition metal oxides, [5] such as the J eff ¼ 1/2 Mott state for iridates (Ir 4þ ). [6] On the other hand, the wavefunctions of 5d electrons are more extended than those of 3d and 4d electrons, which effectively reduces the on-site Coulomb repulsion U. Therefore, the competitive and/or cooperative effect of SOC plus Coulomb repulsion provide a unique playground for novel 5d electronic properties.Here we consider a specific 5d perovskite system: LiOsO 3 , which is known as the first experimentally confirmed polar metal with ferroelectric-like structural transition, [7] as predicted by Anderson and Blount. [8] Certainly, the major concern with such an unusual ferroelectric-like metallic state is not only the potential functionality but more importantly possible competition and coupling between the ferroelectricity and metallicity which are usually mutually exclusive.A well-known but yet unsolved puzzling issue of LiOsO 3 is its magnetic ground state. In LiOsO 3 , each Os ion is surrounded by an oxygen octahedron, which splits Os's 5d orbitals into the t 2g and e g sectors by the crystalline field. In the ideal limit, the three 5d electrons of Os 5þ ion will occupy the t 2g orbitals in the half filling manner. If the SOC effect is negligible, the half-filled...

Strain‐Induced Slater Transition in Polar Metal LiOsO3

“…Previous works based on both local density approximation (LDA) + U and LDA + dynamic meaning field theory (DMFT) both suggested that the ground state of LiOsO3 is just located near the critical point of metal-insulator transition [10]. If the Coulomb repulsion of Os 5+ is weak enough, the strong hybridization between Os's 5d and O's 2p orbitals and spin-orbit coupling (SOC) will suppress the local magnetic moment in LiOsO3 completely and then the ground state should be metallic.…”

“…The Coulomb repulsion in correlated electron systems is usually characterized by the on‐site Hubbard U . As the Coulomb repulsion should be insignificant in 5d systems because of the spatially extended 5d orbitals, we have carefully investigated the ground state of LiOsO 3 by performing two different types of DFT + U methods: the LDA + U method introduced by Liechtenstein et al in which the exchange splitting J and the Hubbard U are considered separately and the simplified local spin density approximate (LSDA) + U method introduced by Dudarev et al which only needs a parameter U eff = U − J . In the LDA + U case, the suitable value of Hubbard U in LiOsO 3 should be in the range of 1.0–2.0 eV, but in LSDA + U case, the value of U eff should be small enough (≈0 eV), i.e., a bare LSDA calculation is a proper choice.…”

“…We take the hexagonal cell of LiOsO 3 in the noncentrosymmetric phase of R 3 c as the reference structure, as shown in Figure a. First, previous works have proved that the magnetic phase of LiOsO 3 could only be NM or G‐AFM whatever the condition is . So in the following, we only consider the competition between the NM phase and G‐AFM phase.…”

“…[2,[7][8][9] Previous works based on both local density approximation (LDA) þ U and LDA þ dynamic meaning field theory (DMFT) suggested that the ground state of LiOsO 3 is just located near the critical point of metal-insulator transition. [10] If the Coulomb repulsion of Os 5þ is weak enough, the strong hybridization between Os's 5d and O's 2p orbitals and spin-orbit coupling (SOC) will suppress the local magnetic moment in LiOsO 3 completely and then the ground state should be metallic. But if the Coulomb repulsion is significant, a G-type antiferromagnetic (G-AFM) structure would be established with every nearest-neighboring spins oppositely aligned, and meanwhile LiOsO 3 will become an insulator.…”

“…[11] Up to now, all experimental evidences showed that LiOsO 3 should be a nonmagnetic (NM) weak-correlated metal, which is different from other 5d 3 osmium oxides such as NaOsO 3 , Cd 2 OsO 7 , etc. [12][13][14] As clarified in a previous study, [10] the main factor that makes LiOsO 3 different from NaOsO 3 is the difference between their lattice structures. The Os-O-Os network is more compact in LiOsO 3 than that in NaOsO 3 , so the 5d-2p hybridization is stronger and the effective hopping between the nearest-neighbor Os's 5d orbitals is larger in LiOsO 3 than that in NaOsO 3 .…”