PHYSICAL REVIEW B 94, 235158 (2016)

Ba2 NiOsO6 : A Dirac-Mott insulator with ferromagnetism near 100 K

Hai L. Feng,1,2,* Stuart Calder,3 Madhav Prasad Ghimire,4,5,† Ya-Hua Yuan,1,6 Yuichi Shirako,7,‡ Yoshihiro Tsujimoto,1
Yoshitaka Matsushita,8 Zhiwei Hu,2 Chang-Yang Kuo,2 Liu Hao Tjeng,2 Tun-Wen Pi,9 Yun-Liang Soo,9,10 Jianfeng He,1,6
Masahiko Tanaka,11 Yoshio Katsuya,11 Manuel Richter,4,12 and Kazunari Yamaura1,6
1
Research Center for Functional Materials, National Institute for Materials Science, 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan
2
Max Planck Institute for Chemical Physics of Solids, Nöthnitzer Strasse 40, 01187 Dresden, Germany
3
Quantum Condensed Matter Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA
4
Leibniz Institute for Solid State and Materials Research, IFW Dresden, P.O. Box 270116, D-01171 Dresden, Germany
5
Condensed Matter Physics Research Center, Butwal-13, Rupandehi, Lumbini, Nepal
6
Graduate School of Chemical Sciences and Engineering, Hokkaido University, North 10 West 8, Kita-ku, Sapporo,
Hokkaido 060-0810, Japan
7
Department of Chemistry, Gakushuin University, 1-5-1 Mejiro, Toshima-ku, Tokyo 171-8588, Japan
8
Materials Analysis Station, National Institute for Materials Science, 1-2-1 Sengen, Tsukuba, Ibaraki 305-0047, Japan
9
National Synchrotron Radiation Research Center, Hsinchu 30076, Taiwan
10
Department of Physics, National Tsing Hua University, Hsinchu 30013, Taiwan
11
Synchrotron X-ray Station at SPring-8, National Institute for Materials Science, Kouto 1-1-1, Sayo-cho, Hyogo 679-5148, Japan
12
Dresden Center for Computational Materials Science, DCMS, TU Dresden, D-01069 Dresden, Germany
(Received 6 October 2016; published 28 December 2016)

The ferromagnetic semiconductor Ba2 NiOsO6 (Tmag ∼ 100 K) was synthesized at 6 GPa and 1500 °C. It
crystallizes into a double perovskite structure [F m-3m; a = 8.0428(1) Å], where the Ni2+ and Os6+ ions are
perfectly ordered at the perovskite B site. We show that the spin-orbit coupling of Os6+ plays an essential role in
opening the charge gap. The magnetic state was investigated by density functional theory calculations and powder
neutron diffraction. The latter revealed a collinear ferromagnetic order in a >21 kOe magnetic field at 5 K. The
ferromagnetic gapped state is fundamentally different from that of known dilute magnetic semiconductors such as
(Ga,Mn)As and (Cd,Mn)Te (Tmag < 180 K), the spin-gapless semiconductor Mn2 CoAl (Tmag ∼ 720 K), and the
ferromagnetic insulators EuO (Tmag ∼ 70 K) and Bi3 Cr3 O11 (Tmag ∼ 220 K). It is also qualitatively different from
known ferrimagnetic insulators and semiconductors, which are characterized by an antiparallel spin arrangement.
Our finding of the ferromagnetic semiconductivity of Ba2 NiOsO6 should increase interest in the platinum group
oxides, because this alternative class of materials should be useful in the development of spintronic, quantum
magnetic, and related devices.

DOI: 10.1103/PhysRevB.94.235158

I. INTRODUCTION produced by synthesis of the Heusler alloy, Mn2 CoAl, which

exhibits spin polarized transport at room temperature and holds
The study of dilute magnetic semiconductors has been
promise for spintronic applications [6]. In 2007, a Co-doped
motivated since the 1970s by the prospect that materials
TiO2 film was claimed to show FM order at room temperature
with tight coupling between magnetic and semiconducting
[7]. However, the essential mechanism of its ferromagnetism
properties may be useful in developing spintronic and related
is still debated [8,9]. Attempts to apply the popular FM
devices [1–3]. Early studies doped magnetic impurities into
insulator EuO (Tmag ∼ 70 K) [10] to technological use are in
semiconducting hosts such as the III-V, II-VI, and IV semi-
progress. La2 NiMnO6 [11,12] and Bi2 NiMnO6 [13–15] have
conductors. Successful synthesis of a paramagnetic (II,Mn)-VI
been studied, because they have the potential to act as a FM
semiconductor was reported in 1978 followed by discovery
insulator in practical applications. Bi3 Cr2.91 O11 recently has
of long-range ferromagnetic (FM) order at ∼10 K by Ohno
been suggested to be a FM insulator with Tmag ∼ 220 K [16]. In
et al. in 1992 [4]. The magnetic ordering temperature (Tmag )
this study, we successfully synthesize a FM oxide, Ba2 NiOsO6 ,
increased to 100 K by 1998 [3], but the Tmag of most
in which the charge gap is produced by a mechanism
dilute magnetic semiconductors to date remains at or below
fundamentally different from that of any proposed to date
∼180 K [5].
for known FM insulators and semiconductors such as above.
The variety of magnetic semiconductors has increased in
The unconventional origin of this property suggests significant
concert with ongoing studies of dilute magnetic semiconduc-
feasibility in designing a practical room-temperature FM
tivity. For example, a “spin-gapless semiconductor” has been
insulator because a few 3d-5d double perovskite oxides
(DPOs) related to Ba2 NiOsO6 also exhibit magnetic order
far above room temperature (Table S1 in the Supplemental
*
[email protected] Material [17]) [18] as well as related 4d and 5d oxides SrTcO3
†
[email protected] (Tmag ∼ 1000 K) [19] and NaOsO3 (Tmag = 410 K) [20], yet
‡ magnetic interactions are still poorly understood.
Present Address: Department of Crystalline Materials Science,
Nagoya University, Furo-cho, Chikusa-ku, Nagoya, Aichi 464-8603, We report here the high-pressure, high-temperature syn-
Japan. thesis of the cubic DPO, Ba2 NiOsO6 , and its characterization

2469-9950/2016/94(23)/235158(9) 235158-1 ©2016 American Physical Society

HAI L. FENG et al. PHYSICAL REVIEW B 94, 235158 (2016)

by measurement of bulk magnetic properties and by pow-

der neutron and x-ray diffraction. A hexagonal phase of
Ba2 NiOsO6 (P -3m1; a = 5.73 Å and c = 14.07 Å) was syn-
thesized without high pressure in 1980, but its magnetic
properties were not investigated [21]. Density functional
theory (DFT) calculations show that spin-orbit coupling (SOC)
plays an essential role in generating the charge gap. To
our knowledge, cubic Ba2 NiOsO6 differs qualitatively from
most ferrimagnetic (FIM) semiconductors, in which spins
are ordered in an antiparallel fashion. Our findings point to
an alternative class of FM insulators driven by a platinum
group element that may be useful in the synthesis of practical
magnetic semiconductors.
FIG. 1. The Rietveld refinement of the SXRD profile collected
II. EXPERIMENT at room temperature for Ba2 NiOsO6 . The expected Bragg reflections
are marked by ticks for Ba2 NiOsO6 (top) and NiO (bottom). The
Materials synthesis. Polycrystalline Ba2 NiOsO6 was syn-
estimated mass proportion was Ba2 NiOsO6 :NiO = 0.976 : 0.024.
thesized by solid-state reaction at high pressure from powders The insets depict the room temperature Ni-L2,3 XAS spectra of
of BaO2 (99%, Kojundo Chemical Lab. Co., Ltd., Japan), Ba2 NiOsO6 and NiO (Ni2+ reference) and the Os-L3 XAS spectra
NiO (99.99%, Kojundo Chemical Lab. Co., Ltd., Japan), of Ba2 NiOsO6 and Sr2 FeOsO6 (Os5+ reference) and their second
Os (99.95%, Heraeus Materials Technology, Germany), and derivative.
KClO4 (>99.5%, Kishida Chemical Co., Ltd., Japan). The
powders were thoroughly mixed in a stoichiometric ratio
(BaO2 /NiO/Os/KClO4 = 2 : 1 : 1 : 0.25) and sealed in a Pt Taiwan. The Ni-L2,3 spectrum of NiO and the Os-L3 spectrum
capsule within an Ar-filled glove box. The sealed capsule of Sr2 FeOsO6 [28,29] were measured simultaneously as
was gradually and isotropically compressed in a belt-type references for Ni2+ and Os5+ , respectively.
press (Kobe Steel, Ltd., Japan) to a pressure of 6 GPa. It Other pieces of the Ba2 NiOsO6 pellet were subjected to
was then heated at 1500 °C for 1 h, while maintaining the electrical resistivity (ρ) and specific heat (Cp ) measurements.
capsule pressure. After heating, the sample was quenched to The ρ measurement was conducted at 100–300 K using a
100 °C or below within several seconds followed by release four-point method with a physical properties measurement
of the pressure over a few hours. The resulting pellet was system from Quantum Design, Inc. Electrical contacts were
polycrystalline, dense, and well sintered. prepared longitudinally on a polycrystalline piece with Pt
X-ray and neutron diffraction analysis. A piece of wires and silver paste. The dc gauge current was 0.1 mA.
Ba2 NiOsO6 pellet was finely ground and rinsed with water The temperature dependence of Cp was measured at 2–
to remove residual KCl. The final fine black powder was 300 K by a thermal relaxation method. The contribution
investigated by synchrotron x-ray diffraction (SXRD) at the from residual KCl to the raw Cp data was subtracted using
BL15XU beamline facility, SPring-8, Japan. SXRD data tabulated data [30] and the mole ratio (KCl:Ba2 NiOsO6 =
were collected at ambient temperature using a high-precision 0.25 : 1).
Debye-Scherrer type diffractometer [22] and the “MYTHEN” An amount of rinsed powder was subjected for magnetic
one-dimensional detector [23]. The radiation wavelength was studies; the magnetic susceptibility (χ ) of Ba2 NiOsO6 powder
0.652 98 Å, which was confirmed by a CeO2 reference. The was measured with a magnetic property measurement system
SXRD pattern was analyzed by the Rietveld method using from Quantum Design, Inc. conducted in field-cooled (FC)
“RIETAN-VENUS” software [24,25] Neutron powder diffrac- and zero-field-cooled (ZFC) modes between 2 and 390 K
tion (ND) was performed on a 5-g sample of Ba2 NiOsO6 in fixed magnetic fields of 0.1, 10, and 50 kOe. Isothermal
powder using the HB-2A diffractometer at the High Flux magnetization was measured at 5, 50, and 200 K in magnetic
Isotope Reactor at the Oak Ridge National Laboratory at fields between −50 kOe and +50 kOe with the same
temperatures down to 4 K. Two wavelengths were selected apparatus. Note that the mass fraction of NiO was 2.4%
using a Ge monochromator. A wavelength of 1.54 Å was according to the SXRD analysis (see Fig. 1), indicating a
used to provide sufficient reflections for an accurate crystal small impurity level. Besides, NiO is an established antiferro-
structure determination. A 2.41 Å wavelength was used for magnetic (AFM) insulator; therefore, an AFM contribution
the magnetic structure determination in both zero field and at such a low concentration to the bulk properties was
applied fields up to 45 kOe. The ND data were analyzed using expected to be negligible. We hence did not correct raw
the Rietveld refinement program FULLPROF [26]. The magnetic magnetic data of the magnetic susceptibility and magnetization
structural representation analysis was performed using SARAH measurements.
[27]. Theoretical calculations. DFT calculations were carried
Materials properties characterization. X-ray absorption out using the full-potential linearized augmented plane wave
spectroscopy (XAS) at the Ni-L2,3 and Os-L3 edges was method as implemented in the WIEN2K code [31]. The atomic
conducted at the BL07C and BL08B beamlines using the total sphere radii (RMT ) were fixed at 2.5, 2.1, 2.0, and 1.63 Bohr
electron yield and transmission method, respectively, in the radii for Ba, Ni, Os, and O, respectively. The linear tetrahedron
National Synchrotron Radiation Research Center (NSRRC) in method with 3000 (1000, 500) k points in the entire Brillouin

235158-2
Ba2 NiOsO6 : A DIRAC-MOTT INSULATOR WITH . . . PHYSICAL REVIEW B 94, 235158 (2016)

zone was employed for the reciprocal space integrations in the in Ba2 CaOsO6 [41] and Ba2 CuOsO6 [42], respectively, rather
case of a cubic unit cell containing one chemical unit cell (and than 1.963 Å for the Os5+ -O bond in Ba2 YOsO6 [43]. This
two for tetragonal and four for orthorhombic unit cells). The suggests that Os is nearly hexavalent. The bond valence sum
standard generalized gradient approximation (GGA) in the (BVS) of Os was 5.72 [44], which also suggests hexavalency.
parametrization of Perdew, Burke, and Ernzerhof (PBE-96) The Ni-O bond length was 2.078(2) Å, which is comparable
[32] was used in the initial calculations. The GGA + U to 2.097 Å for the Ni2+ -O bond in Ba2 NiWO6 [45] and
functional with double counting corrections according to distinctively different from 1.960 and 1.976 Å for the Ni3+ -O
Anisimov et al. [33] (a variant of the so-called fully localized bond in LaNiO3 [46] and LiNiO2 [47], respectively. (We
limit with J = 0) was applied for the main investigations. could not find any references for Ba2 Ni3+ MO6 in accessible
Values of the screened Coulomb interaction U = 5 and databases or the literature.) The BVS of Ni was 2.07,
2 eV were used for the Ni and Os atoms, respectively. The which indicates that Ni is nearly divalent. To ensure charge
calculations were tested by varying these values between compensation in bulk, the valence distribution should be
4 and 7 eV for Ni and 1 and 4 eV for Os, respectively. nominally +2 and +6 for Ni and Os, respectively. Therefore,
6
These ranges are comparable to values reported for 3d and the formal electronic configurations should be 3d 8 (t2g eg2 ;
5d oxides ∼2−10 eV for 3d and ∼1−3 eV for 5d) [34–36]. S = 1) for Ni, provided that the exchange splitting is smaller
SOC was considered via a second variational step using the 2
than the crystal field splitting (CFS), and 5d 2 (t2g eg0 ; S = 1)
scalar relativistic eigenfunctions as basis. All calculations were for Os. (Here and throughout the paper, we use the term
conducted using the experimental lattice parameters obtained CFS in the usual meaning that includes both electrostatic and
at 4 K. Parts of the calculations were double checked with the hybridization effects.) The parameters R0 (Ni2+ ) = 1.675 Å,
full-potential local-orbital (FPLO) code [37]. Here, the linear R0 (Os6+ ) = 1.925 Å, and B = 0.37 were used in the BVS
tetrahedron method was employed with an 8 × 8 × 6 k mesh. calculations [44].
Results obtained from the WIEN2K code were found to agree To further examine the valence states of Ni and Os in
well with those from the FPLO code. Ba2 NiOsO6 , we measured the Ni-L2,3 and Os-L3 XAS spectra
together with those of reference materials (see the insets in
Fig. 1). The identical energy positions and analogous multiplet
III. RESULTS AND DISCUSSION
spectral structures of the Ni-L2,3 edges of Ba2 NiOsO6 and
The crystal structure of Ba2 NiOsO6 was refined by the NiO demonstrate the same Ni2+ valence state and NiO6 local
Rietveld technique on an SXRD profile collected at room symmetry in both compounds. In contrast, a spectral shift at
temperature. The SXRD pattern is shown in Fig. 1, and the the Os-L3 edge of approximately 1.0(1) eV to higher energy
refined crystallographic parameters are listed in Table I. The from Sr2 FeOsO6 to Ba2 NiOsO6 indicates an increased valency
results indicate that Ba2 NiOsO6 crystallizes in a fully ordered of 1 from Os5+ to Os6+ in the latter [48–51]. The energy shift,
double perovskite structure, which is cubic with space group the CFS, and the relative t2g and eg related spectral weights
F m-3m (similar to many other DPOs) [38,39]. The degree of can be clearly discerned in the second derivative spectra at
order of the Ni and Os atoms was carefully investigated in the bottom of the right inset in Fig. 1. The eg related peak is
the refinement procedure. First, the atoms were assumed to shifted by 1.3(1) eV to higher energy in Ba2 NiOsO6 relative
occupy the 4a and 4b crystallographic sites randomly and to Sr2 FeOsO6 , and the spectral weight of the t2g related peak
partially at equimolar ratio. Then, a fully ordered model,
in which Ni remained solely at the 4b site and Os at the
4a site, was tested. The fully ordered model was found to
provide the best refinement quality. We therefore conclude
that Ba2 NiOsO6 has fully ordered Ni and Os atoms, similar
to Sr2 NiOsO6 and Ca2 NiOsO6 [40]. The order has been
attributed to distinctive differences between the charge, size,
and electronic configuration of the Ni and Os atoms [39,40].
The local coordination at the magnetic Ni and Os ions
was investigated. The Os-O bond length [1.943(2) Å] was
similar to values of 1.948 and 1.955 Å for the Os6+ -O bond

TABLE I. Atomic positions and thermal parameters of

Ba2 NiOsO6 . Note: Space group: F m-3m; lattice parameter a =
8.0428(1) Å, Z = 4; dcal = 7.9102 g cm−3 . R indices were Rp =
2.82% and Rwp = 3.98%.

2
Atom Site Occupancy x y z B (Å )
Ba 8c 1 0.25 0.25 0.25 0.36(6) FIG. 2. Structural comparison of the series A2 NiOsO6 (A = Ca,
Os 4a 1 0 0 0 0.11(4) Sr, Ba). The crystal structure symmetry changes from monoclinic
Ni 4b 1 0.5 0 0 0.19(2) (P 21 /n) to cubic (F m-3m) via a tetragonal (I 4/m) symmetry. Green
O 24e 1 0.2415(3) 0 0 0.40(2) balls represent A atoms; reddish and bluish octahedra represent NiO6
and OsO6 .

235158-3
HAI L. FENG et al. PHYSICAL REVIEW B 94, 235158 (2016)

increases reflecting a smaller number of t2g holes [48] on Os5+

3 2
(t2g ) compared to Os6+ (t2g ). The second derivative spectra
show also a larger CFS for Os6+ in Ba2 NiOsO6 compared
with that of Os5+ in Sr2 FeOsO6 as found in observations of Ir
oxides [48].
We have compared the crystal structure of Ba2 NiOsO6 with
those of A2 NiOsO6 (A = Ca, Sr) [40]. The crystal structure
symmetry changes from monoclinic (P 21 /n; A = Ca) to
cubic (F m-3m; Ba) via a tetragonal (I 4/m; Sr) symmetry,
as shown in Fig. 2. An analogous symmetry change has
been observed for A2 FeMoO6 (A = Ca, Sr, Ba) [38], which
suggests that the size of A plays a role in determining the
lattice symmetry. The angle of the magnetic Os-O-Ni bond
is altered from 151◦ (Ca2 NiOsO6 , monoclinic) [40] to 180◦
(Ba2 NiOsO6 , cubic) via 166◦ /180◦ (Sr2 NiOsO6 , tetragonal)
[40]. The angle variation seems to have a large impact on
the magnetic properties. For example, Ca2 NiOsO6 is canted
AFM below 175 K, whereas Sr2 NiOsO6 is AFM below 50 K.
When the size of A increases, the magnitude of the FM
interaction increases as evidenced by the increase in Weiss
temperature from 27 K in Sr2 NiOsO6 (see Table S1 [17]) to
113(1) K in Ba2 NiOsO6 [see the inset of Fig. 3(a)]. The Curie
constant deduced from the 50 kOe FC curve for Ba2 NiOsO6
is 1.48(2) emu mol−1 K−1 , which corresponds to an effective
moment (μeff ) of 3.46(2) μB per formula unit (f.u.). The Weiss
temperature and the Curie constant deduced from the 0.1 kOe
FC curve were 121(2) K and 1.28(2) emu mol−1 K−1 [μeff =
3.21(2) μB perf.u.], respectively, consistent with the absence
of a marked change.
Although the FM interaction is large in Ba2 NiOsO6 , an
AFM-like peak was observed in both the ZFC and FC curves
FIG. 3. (a) Temperature dependence of the magnetic susceptibil-
at ∼32 K (Tmag ) (Fig. S1 [17]). However, no noticeable peak
ity of Ba2 NiOsO6 measured at 50 kOe. The inset shows the inverse
is present in the 50 kOe curves [Fig. 3(a)], which suggests
magnetic susceptibility. The solid line is the Curie-Weiss fit to the
a possible change of the magnetic order. To investigate the
curve. (b) Isothermal magnetization loop of Ba2 NiOsO6 at 5 K in
magnetism further, we conducted an isothermal magnetization comparison with the loops for Ca2 NiOsO6 (taken from Ref. [39])
measurement at a temperature below Tmag . The magnetization and Sr2 NiOsO6 (estimated from the magnetic susceptibility data in
of Ba2 NiOsO6 reaches approximately 2.3 μB f.u.−1 at 50 kOe Ref. [39]).
[at 5 K, Fig. 3(b)]. Intriguingly, the magnetic loop indicates a
noticeable metamagnetic transition at approximately ±21 kOe
for Ba2 NiOsO6 . For comparison, magnetic loops collected at
5 K for Sr2 NiOsO6 and Ca2 NiOsO6 are shown to emphasize
the anomalous behavior of Ba2 NiOsO6 . A similar metamag-
netic transition was observed for another DPO, R2 CoMnO6
(R = rare-earth element). However, this was argued to be
caused by the presence of significant antisite disorder [52–55].
Since such disorder is absent in Ba2 NiOsO6 , the origin of
metamagnetism of Ba2 NiOsO6 should be different.
We also conducted specific heat measurements near Tmag
with and without application of a magnetic field (Fig. S2
[17]). The zero-magnetic field curve shows a broad cusp
at ∼32 K, whereas the cusplike top shifts slightly toward
lower temperature in an applied field of 10 kOe. In magnetic
fields of 30 and 50 kOe, the cusp disappears completely.
The disappearance is likely connected to the absence of an
AFM-like peak on the χ -T curves at 50 kOe.
The temperature dependence of the electrical resistivity FIG. 4. The temperature dependence of ρ for polycrystalline
of polycrystalline Ba2 NiOsO6 is shown in Fig. 4 . At room Ba2 NiOsO6 . The inset shows the Arrhenius plot of the data and
temperature the value is ∼120  cm, which is more than the fitting line, which yield an activation energy of 0.31 eV. Data
three orders of magnitude greater than the expected value measured in a magnetic field of 50 kOe also are presented.

235158-4
Ba2 NiOsO6 : A DIRAC-MOTT INSULATOR WITH . . . PHYSICAL REVIEW B 94, 235158 (2016)

[17]). The modulated AFM order transformed to a collinear

FM order in a magnetic field above 21 kOe at 5 K. For example,
the difference between the zero-field and non-zero-field ND
profiles clearly revealed the presence of FM order [Fig. 6(a)],
which gradually developed with increasing magnetic field
[Fig. 6(b)]. Figure 6(c) shows the applied magnetic field
dependence of the magnetic peak intensity at 2θ = 30◦ , which
indicates that the critical field (Hc ) is 21 kOe at 5 K. The
magnetic field observations throughout the ND measurements
accord well with the metamagnetic transition found in the
isothermal magnetization measurements. The FM order also
was refined by analysis of the ND profile collected at 45
kOe. All Ni and Os magnetic moments lie along the [100]
or equivalent [010] or [001] direction [Fig. 6(d)] indicating
the common plane of the modulated AFM and FM orders.
The magnetic moments in the FM state at 5 K were
estimated to be 2.13(8) and 0.97(5) μB per Ni and Os,
respectively. Although the Ni moment is comparable to the
spin only moment estimated from a simplified model, the Os
moment is significantly smaller. This is most likely due to
the impact of the SOC of Os, as suggested in Ref. [59], and
to the extended orbitals of the 5d atom leading to increased
hybridization, as suggested in Ref. [60]. The magnetization of
FIG. 5. Rietveld refinement of the ND profiles (λ = 1.54 Å) of
Ba2 NiOsO6 at 5 K and 50 kOe is approximately 2.3 μB f.u.−1
Ba2 NiOsO6 collected at temperatures of (a) 50 K and (b) 4 K. The
expected Bragg reflections are marked by ticks for Ba2 NiOsO6 (top),
[Fig. 3(b)], although the ND analysis yields approximately
NiO (middle), and the magnetic lattice of NiO (bottom). 3 μB f.u.−1 in total. This indicates incomplete saturation in
a 50 kOe field, which is evident from the relatively large
gradient, dM/dH , at this strength [see Fig. 3(b)]. Further
for a metallic polycrystalline oxide. Electrical resistivity high-field-magnetization studies may confirm the ND result.
increased upon cooling and exceeded the instrumental limit We have analyzed the electronic and magnetic states of
at temperatures below 105 K indicating a semiconductorlike Ba2 NiOsO6 by DFT. We calculated the total energies of
behavior. To estimate the activation energy of conduction, data supercells with magnetic alignments corresponding to one
points were plotted on a logarithmic scale as a function of FM, two AFM, and three FIM orderings (see Fig. 7). Using
inverse temperature and fit to the Arrhenius equation. The fit up and down spins to indicate the magnetic moment on the
provided an activation energy of 0.31 eV (see the inset of Ni1, Ni2, Os1, and Os2 atoms, the orderings investigated are
Fig. 4). In a field of 50 kOe, the electrical resistivity remained FM1-↑↑↑↑, AF1-↑↓↑↓, AF2-↑↓↓↑, FI1-↑↑↓↓, FI2-↑↓↑↑,
high and exceeded the instrumental limit at low temperatures and FI3-↑↑↑↓, respectively. These arrangements are consid-
as in zero field (see the plot of solid circles). ered in tetragonal supercells (space group 123, P 4/mmm)
The magnetic order of Ba2 NiOsO6 was investigated by containing two chemical unit cells. We double checked the
ND at low temperatures. Figures 5(a) and 5(b) show the ND calculations using orthorhombic supercells (space group 47,
patterns (λ = 1.54 Å) collected at 50 and 4 K, respectively. The Pmmm) with four chemical unit cells and pseudocubic lattice
nuclear lattice was well refined by the room-temperature lattice parameters a = b = c. The two settings produced nearly
model both above and below the AFM-like peak temperature identical results (with differences below 4 meV f.u.−1 ) for
(∼32 K), which indicates the absence of a lattice symmetry calculations conducted without SOC. With SOC, the total
change down to 5 K regardless of the presence of the magnetic energies obtained with the tetragonal setting did not fulfill
transition. Although the electronic configuration of Os6+ the expected symmetry relations. Thus, we report only results
2
(5d t2g ) implies a possible orbital order below Tmag [56,57]; obtained with the orthorhombic setting. The latter was double
any indication of reduced lattice symmetries lower than cubic checked by performing GGA + U + SOC calculations in a
was undetected within the instrument resolution. For example, cubic cell with one chemical unit for the FM1 case. The
a refinement attempt with a tetragonal space group of I 4/m magnetic anisotropy energies were 18 meV f.u.−1 (orthorhom-
did not converge to stable lattice parameters. A robust cubic bic setting; see Table II) and 13 meV f.u.−1 (cubic setting).
symmetry lattice, which argues against orbital ordering, was From this comparison and the small differences between the
found in a similar Os6+ compound, Ba2 CaOsO6 [58]. The orthorhombic and tetragonal settings, we assume that the other
low-temperature crystallographic parameters are summarized energy differences in Table II have a similar precision of ca.
in Table S2 [17]. 5 meV f.u.−1 .
After changing λ to 2.41 Å, we conducted ND mea- By comparing the total energies of the different magnetic
surements to investigate the magnetic order. The difference arrangements, we find that the FM1 order has the lowest
between the profiles at 5 and 75 K clearly revealed the total energy regardless of the calculation scheme (Table II).
presence of modulated AFM order as shown in Fig. S4 The energy difference to the next lowest order (AF2) is very
(detailed analysis is provided in the Supplemental Material small, ∼16 meV per f.u. in GGA + U + SOC. The difference

235158-5
HAI L. FENG et al. PHYSICAL REVIEW B 94, 235158 (2016)

FIG. 6. (a) Difference between the zero-field profile and the profile collected in a magnetic field of 45 kOe at 5 K. (b) Evolution of a magnetic
peak with increasing magnetic field at 5 K. (c) Magnetic field dependence of the intensity of the magnetic peak at 2θ = 30◦ (λ = 2.41 Å) at a
fixed temperature of 5 K. The solid line and curve are a guide for the eyes. (d) Magnetic order model of the FM state of Ba2 NiOsO6 depicted
from the analysis of the ND profiles. The magnetic easy axis lies in the cubic [100], [010], or [001] direction.

is almost comparable to the precision of the method, which for GGA + U and a gapped state for GGA + U + SOC. For
implies a competition between the FM1 and AF2 orders. a more detailed view, we considered the partial DOS (Fig. S7
Other possible magnetic orders (not shown here) were found [17]). The Ni 3d band is fully occupied in one spin channel
3
at much higher energies and were excluded from further and partially occupied (nominally t2g eg0 ) in the other, both
investigation. with and without SOC. Due to the combined effect of CFS
The FM1 alignment is identical to the experimental FM and U = 5 eV, the unoccupied Ni eg states lie about 2 eV
state in an external magnetic field [Fig. 6(d)]. Figure 8 shows above EF . The O 2p orbitals hybridize strongly with the Ni 3d
the DOS of FM1 for cases (a) without and (b) with SOC. The orbitals below −1.5 eV (the valence region) and with the Os
band structures in Figs. 9(a) and 9(b) show a metallic state 5d orbitals.
The most interesting band located at the Fermi level is
composed mainly of Os 5d states (Fig. S7 [17]). We performed
a decomposition into ml-states (Fig. S8 [17]), which confirms
the expected t2g character of the band. Without SOC [Fig. 9(a)],
the Os t2g states form two band complexes, each containing

TABLE II. Relative total energies (meV f.u.−1 ) calculated within

GGA, GGA + U , and GGA + U + SOC schemes for Ba2 NiOsO6 .
For the latter scheme, the orientations of the magnetic moments are
given with respect to the cubic axes. Results for the [110] orientation
were obtained only for the three spin arrangements with the lowest
energies in GGA or GGA + U .

FM1 AF1 AF2 FI1 FI2 FI3

FIG. 7. Left: Sketch of the tetragonal supercell (space group 123,
P 4/mmm) used for DFT calculations of possible magnetic arrange- GGA 0 85 40 115 61 58
ments. The cubic [110] direction is indicated. Right: Schematic of GGA + U 0 73 45 82 58 40
the magnetic arrangements. Red and blue arrows indicate magnetic GGA + U + SOC [001] 18 68 41 91 55 53
moments on the Ni/Os atoms. The magnetic easy axis was found GGA + U + SOC [100] 18 48 25 91 37 52
to lie along the cubic [110] direction for FM1 and AF2 by GGA + U + SOC [110] 0 16 28
GGA + U + SOC.

235158-6
Ba2 NiOsO6 : A DIRAC-MOTT INSULATOR WITH . . . PHYSICAL REVIEW B 94, 235158 (2016)

FIG. 10. A schematic view of the d band DOS structure of the

FM1 state of Ba2 NiOsO6 . The partial DOS for (a) Ni 3d and (b) Os 5d
FIG. 8. The total DOS (per f.u.) for spin-up (↑) and spin-down
bands and (c) the total DOS. The symbols s and f represent the spin
(↓) channels of the FM1 state of Ba2 NiOsO6 obtained in the (a)
exchange splitting and crystal field splitting (including correlation
GGA + U and (b) GGA + U + SOC schemes. The vertical dotted
effects), respectively.
line indicates EF = 0.

three individual bands: a spin-up complex around the Fermi certain value of U . A schematic illustration of the mechanism
level and a spin-down complex around 1 eV. The former is is depicted in Fig. 10. For the sake of clarity, details of the
partially occupied with two electrons per Os. It cannot be split DOS are ignored or changed, and the oxygen DOS is omitted.
without breaking the nonrelativistic cubic symmetry, such that
application of U does not open a gap. With SOC [Fig. 9(b)],
IV. CONCLUSION
the t2g states are split into two 8 states and one 7 state.
However, SOC alone is not sufficient to open a gap, because the Cubic Ba2 NiOsO6 was synthesized by solid-state reaction
band dispersion is greater than the spin-orbit splitting, which at 6 GPa and 1500 ◦ C. SXRD and ND revealed cubic lattice
leads to an (indirect) overlap of the band states. Although symmetry (F m-3m) with lattice parameter a = 8.0428(1) Å
we could not directly measure the strength of the SOC from at room temperature [8.0298(1) Å at 4 K]. XAS of the
the calculation, a comparison between the degree of splitting Ni and Os core levels indicated that the formal charges on
of the t2g band within the GGA and GGA + SOC schemes Ni and Os are essentially +2 and +6, respectively. Thus,
roughly estimates the spin-orbit splitting to be ∼0.17 eV. Due the formal electronic configurations should be 3d 8 (t2g 6
eg2 ;
to symmetry reduction upon inclusion of SOC, application S = 1) for Ni and 5d (t2g eg ; S = 1) for Os. The combined
2 2 0
of Coulomb corrections may increase the splitting between evaluation of theoretical and experimental results establish
occupied and unoccupied bands, which will open the gap at a Ba2 NiOsO6 as a FM insulator with Tmag of ∼100 K. SOC plays
an essential role in opening the charge gap. A metamagnetic
transition was found at 5 K, where a modulated AFM state
transforms to the FM state at a moderate magnetic field (Hc ∼
21 kOe at 5 K) without altering the semiconducting property.
DFT calculations suggest that FM and helical AFM orders
have comparable total energies and are responsible for the
metamagnetic transition. We conclude that both Coulomb and
relativistic effects must be considered equally in Ba2 NiOsO6 .
The situation is similar to the cases of the platinum group
oxide Sr2 IrO4 [36,61,62] and α-RuCl3 [63], although these
are essentially AFM materials.
1
Recently, a similar material, Ba2 NaOsO6 , having 5d 1 (t2g
tg0 ) configuration of Os, has theoretically been studied, suggest-
ing that it is an uncommon FM insulator [64]. Although the
Weiss temperature of Ba2 NaOsO6 is weakly negative (−10 K),
it might be the first material being clarified as a FM Dirac-Mott
insulator (Tmag = 6.8 K). The present compound is likely the
FIG. 9. Band structures of the FM1 state of Ba2 NiOsO6 within (a) second case; however, the FM properties are dramatically
GGA + U (red: spin up; black: spin down) and (b) GGA + U + SOC. improved as evidenced by such things as the substantially
Dashed horizontal lines at zero correspond to EF . positive Weiss temperature (+113 K) and increased Tmag near

235158-7
HAI L. FENG et al. PHYSICAL REVIEW B 94, 235158 (2016)

100 K. Thus, the cubic double perovskite reported here could ACKNOWLEDGMENTS
be useful to deepen understanding of the nature of the FM
The authors would like to thank the staff of BL15XU,
Dirac-Mott insulator.
NIMS, and SPring-8 for their help at the beamline. The
Cubic Ba2 NiOsO6 should increase general interest in
SXRD measurements were performed under the approval
platinum group oxides. Platinum group metals and their
of the NIMS Beamline Station (Proposals No. 2014A4504,
complexes have been used in many practical applications,
NO. 2014B4501, No. 2015A4502, and No. 2016B4504). The
because of their excellent catalytic activities [65]. However,
XAS measurements were supported by Chin-Wen Pao. M.P.G
electrical properties and other practical features are not well
thanks the Alexander von Humboldt Foundation for financial
known apart from high electrical conductivity and robust
support through the Georg Forster Research Fellowship
corrosion resistance. Recent studies of Ir and Os oxides have
Program. M.P.G. thanks K. Koepernik and R. Laskowski
shown that the significant SOC and extended 5d orbitals in the
for helpful discussion, and M.R. thanks M. Knupfer, U.
oxide produce outstanding electric and magnetic properties
Rößler, and H. Rosner for helpful discussion. M.P.G. and M.R.
as found in Slater insulators [20,66], Dirac-Mott insulators
thank U. Nitzsche for technical assistance. This research was
[36,61], and ferroelectric metals [67]. The largest reported
supported in part by the World Premier International Research
spin-phonon-electron coupling was recently discovered in
Center of the Ministry of Education, Culture, Sports, Science
an osmium oxide, which further illustrates this potential of
and Technology (MEXT) of Japan and the Japan Society for
high-valent platinum group oxides [68]. Growing interest in
the Promotion of Science (JSPS) through a Grant-in-Aid for
unique physical properties, as exemplified by the FM semicon-
Scientific Research (Grants No. 25289233, No. 15K14133,
ductivity of cubic Ba2 NiOsO6 , should motivate further study of
and No. 16H04501). The research conducted at ORNL’s
platinum group oxides. Our report of cubic Ba2 NiOsO6 heralds
High Flux Isotope Reactor was sponsored by the Scientific
an alternative class of FM insulator oxides, which may be
User Facilities Division, Office of Basic Energy Sciences, US
useful in developing a practical magnetic semiconductor that
Department of Energy.
can be employed in spintronic and quantum magnetic devices.

[1] T. Dietl, Nat. Mater. 9, 965 (2010). [17] See Supplemental Material at http://link.aps.org/supplemental/
[2] A. H. MacDonald, P. Schiffer, and N. Samarth, Nat. Mater. 4, 10.1103/PhysRevB.94.235158 for list of double perovskite
195 (2005). oxides, A2 MOsO6 , in which A = Ca, Sr, Ba, and M = a
[3] H. Ohno, Science 281, 951 (1998). 3d-transition metal.
[4] H. Ohno, H. Munekata, T. Penney, S. von Molnár, and L. L. [18] D. Serrate, J. M. De Teresa, and M. R. Ibarra, J. Phys.: Condens.
Chang, Phys. Rev. Lett. 68, 2664 (1992). Matter. 19, 023201 (2007).
[5] M. Wang, R. P. Campion, A. W. Rushforth, K. W. Edmonds, [19] E. E. Rodriguez, F. Poineau, A. Llobet, B. J. Kennedy, M.
C. T. Foxon, and B. L. Gallagher, Appl. Phys. Lett. 93, 132103 Avdeev, G. J. Thorogood, M. L. Carter, R. Seshadri, D. J. Singh,
(2008). and A. K. Cheetham, Phys. Rev. Lett. 106, 067201 (2011).
[6] S. Ouardi, G. H. Fecher, C. Felser, and J. Kübler, Phys. Rev. [20] Y. G. Shi, Y. F. Guo, S. Yu, M. Arai, A. A. Belik, A. Sato, K.
Lett. 110, 100401 (2013). Yamaura, E. Takayama-Muromachi, H. F. Tian, H. X. Yang, J.
[7] Y. Matsumoto, M. Murakami, T. Shono, T. Hasegawa, T. Q. Li, T. Varga, J. F. Mitchell, and S. Okamoto, Phys. Rev. B
Fukumura, M. Kawasaki, P. Ahmet, T. Chikyow, S.-y. 80, 161104 (2009).
Koshihara, and H. Koinuma, Science 291, 854 (2001). [21] U. Treiber and S. Kemmler-Sack, Z. Anorg. Allg. Chem. 470,
[8] M. Parras, Á. Varela, R. Cortés-Gil, K. Boulahya, A. Hernando, 95 (1980).
and J. M. González- Calbet, J. Phys. Chem. Lett. 4, 2171 (2013). [22] M. Tanaka, Y. Katsuya, and A. Yamamoto, Rev. Sci. Instrum.
[9] S. R. Shinde, S. B. Ogale, J. S. Higgins, H. Zheng, A. J. Millis, 79, 075106 (2008).
V. N. Kulkarni, R. Ramesh, R. L. Greene, and T. Venkatesan, [23] M. Tanaka, Y. Katsuya, Y. Matsushita, and O. Sakata, J. Ceram.
Phys. Rev. Lett. 92, 166601 (2004). Soc. Jpn. 121, 287 (2013).
[10] N. Tsuda, K. Nasu, A. Fujimori, and K. Siratori, Electronic [24] K. Momma and F. Izumi, J. Appl. Crystallogr. 41, 653
Conduction in Oxides (Springer, Berlin, 2013). (2008).
[11] R. I. Dass and J. B. Goodenough, Phys. Rev. B 67, 014401 [25] F. Izumi and K. Momma, Solid State Phenom. 130, 15 (2007).
(2003). [26] J. Rodrı́guez-Carvajal, Physica B 192, 55 (1993).
[12] R. I. Dass, J. Q. Yan and J. B. Goodenough, Phys. Rev. B 68, [27] A. S. Wills, Physica B 276-278, 680 (2000).
064415 (2003). [28] A. K. Paul, M. Reehuis, V. Ksenofontov, B. H. Yan, A. Hoser,
[13] E. Langenberg, I. Fina, J. Ventura, B. Noheda, M. Varela, and J. D. M. Tobbens, P. M. Abdala, P. Adler, M. Jansen, and C. Felser,
Fontcuberta, Phys. Rev. B 86, 085108 (2012). Phys. Rev. Lett. 111, 167205 (2013).
[14] Y. Du, Z. X. Cheng, X. L. Wang, P. Liu, and S. X. Dou, J. Appl. [29] A. K. Paul, M. Jansen, B. H. Yan, C. Felser, M. Reehuis, and P.
Phys. 109, 07B507 (2011). M. Abdala, Inorg. Chem. 52, 6713 (2013).
[15] M. Azuma, K. Takata, T. Saito, S. Ishiwata, Y. Shimakawa, and [30] Y. S. Touloukian and E. H. Buyco, Specific Heat Nonmetallic
M. Takano, J. Am. Chem. Soc. 127, 8889 (2005). Solids (IFI/Plenum, New York, 1970).
[16] W. Yi, Y. Matsushita, A. Sato, K. Kosuda, M. Yoshitake, and A. [31] P. Blaha, K. Schwarz, G. K. H. Madsen, D. Kvasnicka, and J.
A. Belik, Inorg. Chem. 53, 8362 (2014). Luitz, wien2k, version 14.2, An Augmented Plane Wave+Local

235158-8
Ba2 NiOsO6 : A DIRAC-MOTT INSULATOR WITH . . . PHYSICAL REVIEW B 94, 235158 (2016)

Orbitals Program for Calculating Crystal Properties (Technis- [52] M. K. Kim, J. Y. Moon, H. Y. Choi, S. H. Oh, N. Lee,
che Universitat Wien, Austria, 2001). and Y. J. Choi, J. Phys.: Condens. Matter. 27, 426002
[32] J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, (2015).
3865 (1996). [53] J. Blasco, L. Garcia-Munoz, J. Garcia, J. Stankiewicz, G. Subias,
[33] V. I. Anisimov, I. V. Solovyev, M. A. Korotin, M. T. Czyzyk, C. Ritter, and J. A. Rodriguez-Velamaz, Appl. Phys. Lett. 107,
and G. A. Sawatzky, Phys. Rev. B 48, 16929 (1993). 012902 (2015).
[34] Y. Yuan, H. L. Feng, M. P. Ghimire, Y. Matsushita, Y. Tsujimoto, [54] L. Y. Wang, W. P. Zhou, D. H. Wang, Q. Q. Cao, Q. Y. Xu, and
J. He, M. Tanaka, Y. Katsuya, and K. Yamaura, Inorg. Chem. Y. W. Du, J. Appl. Phys. 117, 17D914 (2015).
54, 3422 (2015). [55] H. S. Nair, T. Chatterji, and A. M. Strydom, Appl. Phys. Lett.
[35] H. B. Wang, S. S. Zhu, X. D. Ou, and H. Wu, Phys. Rev. B 90, 106, 022407 (2015).
054406 (2014). [56] M. A. Laguna-Marco, P. Kayser, J. A. Alonso, M. J. Martı́nez-
[36] B. J. Kim, H. Jin, S. J. Moon, J. Y. Kim, B. G. Park, C. S. Leem, Lope, M. van Veenendaal, Y. Choi, and D. Haskel, Phys. Rev. B
J. Yu, T. W. Noh, C. Kim, S. J. Oh, J. H. Park, V. Durairaj, G. 91, 214433 (2015).
Cao, and E. Rotenberg, Phys. Rev. Lett. 101, 076402 (2008). [57] S. I. Csiszar, M. W. Haverkort, Z. Hu, A. Tanaka, H. H. Hsieh,
[37] K. Koepernik and H. Eschrig, Phys. Rev. B 59, 1743 (1999). H. J. Lin, C. T. Chen, T. Hibma, and L. H. Tjeng, Phys. Rev.
[38] R. P. Borges, R. M. Thomas, C. Cullinan, J. M. D. Coey, Lett. 95, 187205 (2005).
R. Suryanarayanan, L. Ben-Dor, L. Pinsard-Gaudart, and A. [58] C. M. Thompson, J. P. Carlo, R. Flacau, T. Aharen, I. A. Leahy,
Revcolevschi, J. Phys.: Condens. Matter. 11, L445 (1999). J. R. Pollichemi, T. J. S. Munsie, T. Medina, G. M. Luke, J.
[39] M. T. Anderson, K. B. Greenwood, G. A. Taylor, and K. R. Munevar, S. Cheung, T. Goko, Y. J. Uemura, and J. E. Greedan,
Poeppelmeier, Prog. Solid State Chem. 22, 197 (1993). J. Phys.: Condens. Matter. 26, 306003 (2014).
[40] R. Macquart, S. J. Kim, W. R. Gemmill, J. K. Stalick, Y. Lee, T. [59] H. Matsuura and K. Miyake, J. Phys. Soc. Jpn. 82, 073703
Vogt, and H. C. zur Loye, Inorg. Chem. 44, 9676 (2005). (2013).
[41] K. Yamamura, M. Wakeshima, and Y. Hinatsu, J. Solid State [60] A. E. Taylor, R. Morrow, D. J. Singh, S. Calder, M. D. Lumsden,
Chem. 179, 605 (2006). P. M. Woodward, and A. D. Christianson, Phys. Rev. B 91,
[42] H. L. Feng, M. Arai, Y. Matsushita, Y. Tsujimoto, Y. H. Yuan, 100406 (2015).
C. I. Sathish, J. F. He, M. Tanaka, and K. Yamaura, J. Solid State [61] B. J. Kim, H. Ohsumi, T. Komesu, S. Sakai, T. Morita, H. Takagi,
Chem. 217, 9 (2014). and T. Arima, Science 323, 1329 (2009).
[43] E. Kermarrec, C. A. Marjerrison, C. M. Thompson, D. D. [62] G. Jackeli and G. Khaliullin, Phys. Rev. Lett. 102, 017205
Maharaj, K. Levin, S. Kroeker, G. E. Granroth, R. Flacau, Z. (2009).
Yamani, J. E. Greedan, and B. D. Gaulin, Phys. Rev. B 91, [63] K. W. Plumb, J. P. Clancy, L. J. Sandilands, V. V. Shankar, Y.
075133 (2015). F. Hu, K. S. Burch, H. Y. Kee, and Y. J. Kim, Phys. Rev. B 90,
[44] I. D. Brown, Chem. Rev. 109, 6858 (2009). 041112 (2014).
[45] D. E. Cox, G. Shirane, and B. C. Frazer, J. Appl. Phys. 38, 1459 [64] S. Gangopadhyay and W. E. Pickett, Phys. Rev. B 91, 045133
(1967). (2015).
[46] H. Falcón, A. E. Goeta, G. Punte, and R. E. Carbonio, J. Solid [65] H. Renner, G. Schlamp, I. Kleinwächter, E. Drost, H. M.
State Chem. 133, 379 (1997). Lüschow, P. Tews, P. Panster, M. Diehl, J. Lang, T. Kreuzer, A.
[47] A. Hirano, R. Kanno, Y. Kawamoto, Y. Takeda, K. Yamaura, M. Knödler, K. A. Starz, K. Dermann, J. Rothaut, R. Drieselmann,
Takano, K. Ohyama, M. Ohashi, and Y. Yamaguchi, Solid State C. Peter, and R. Schiele, in Ullmann’s Encyclopedia of Industrial
Ionics 78, 123 (1995). Chemistry, edited by Barbara Elvers (Wiley-VCH Verlag GmbH
[48] L. S. I. Veiga, G. Fabbris, M. van Veenendaal, N. M. Souza-Neto, & Co. KGaA, Berlin, 2000).
H. L. Feng, K. Yamaura, and D. Haskel, Phys. Rev. B 91, 235135 [66] S. Calder, V. O. Garlea, D. F. McMorrow, M. D. Lumsden, M.
(2015). B. Stone, J. C. Lang, J. W. Kim, J. A. Schlueter, Y. G. Shi, K.
[49] J.-M. Chen, Y.-Y. Chin, M. Valldor, Z. Hu, J.-M. Lee, S.-C. Yamaura, Y. S. Sun, Y. Tsujimoto, and A. D. Christianson, Phys.
Haw, N. Hiraoka, H. Ishii, C.-W. Pao, K.-D. Tsuei, J.-F. Lee, Rev. Lett. 108, 257209 (2012).
H.-J. Lin, L.-Y. Jang, A. Tanaka, C.-T. Chen, and L. H. Tjeng, [67] Y. G. Shi, Y. F. Guo, X. Wang, A. J. Princep, D. Khalyavin,
J. Am. Chem. Soc. 136, 1514 (2014). P. Manuel, Y. Michiue, A. Sato, K. Tsuda, S. Yu, M. Arai,
[50] T. Burnus, Z. Hu, H. Wu, J. C. Cezar, S. Niitaka, H. Takagi, C. Y. Shirako, M. Akaogi, N. L. Wang, K. Yamaura, and A. T.
F. Chang, N. B. Brookes, H. J. Lin, L. Y. Jang, A. Tanaka, K. Boothroyd, Nat. Mater. 12, 1024 (2013).
S. Liang, C. T. Chen, and L. H. Tjeng, Phys. Rev. B 77, 205111 [68] S. Calder, J. H. Lee, M. B. Stone, M. D. Lumsden, J. C. Lang,
(2008). M. Feygenson, Z. Zhao, J. Q. Yan, Y. G. Shi, Y. S. Sun, Y.
[51] Z. Hu, M. S. Golden, S. G. Ebbinghaus, M. Knupfer, J. Fink, F. Tsujimoto, K. Yamaura, and A. D. Christianson, Nat. Commun.
M. F. de Groot, and G. Kaindl, Chem. Phys. 282, 451 (2002). 6, 8916 (2015).

235158-9