PHYSICAL REVIEW MATERIALS 4, 034405 (2020)

Origin of itinerant carriers in antiferromagnetic LaFe1−x Mox O3 studied by x-ray spectroscopies

Dibya Phuyal ,1,* Soham Mukherjee,1 S. K. Panda,2 Somnath Jana,1,† Carlo U. Segre ,3 Laura Simonelli,4
Sergei M. Butorin ,1 Håkan Rensmo,1 and Olof Karis1
1
Division of Molecular and Condensed Matter Physics, Department of Physics, Uppsala University, Box 516, SE-75121, Uppsala, Sweden
2
Department of Physics, Bennett University, Greater Noida 201310, Uttar Pradesh, India
3
CSRRI and Department of Physics, Illinois Institute of Technology, Chicago, Illinois 60616, USA
4
CELLS-ALBA Synchrotron, E-08290 Cerdanyola del Vallés, Barcelona, Spain

(Received 22 October 2019; revised manuscript received 10 February 2020; accepted 18 February 2020;
published 13 March 2020)

We report on the electronic structure of doped LaFeO3 at the crossover from an insulating-to-metallic
phase transition. Comprehensive x-ray spectroscopic methodologies are used to understand core and valence
electronic structure as well as crystal structure distortions associated with the electronic transition. Despite the
antiferromagnetic (AFM) ordering at room temperature, we show direct evidence of itinerant carriers at the
Fermi level revealed by resonant photoemission spectroscopy (RPES) at the Mo L3 edge. RPES data taken at
the Fe L3 edge show spectral weight near the valence band edge and significant hybridization with O 2p states
required for AFM ordering. Resonant inelastic x-ray scattering spectra taken across Fe L2,3 edges show electron
correlation effects (U) driven by Coulomb interactions of d electrons as well as broad charge-transfer excitations
for x  0.2 where the compound crosses over to a metallic state. Site substitution of Fe by Mo ions in the
Fe-O6 octahedra enhances the separation of the two Fe-O bonds and Fe-O-Fe bonding angles relative to the
orthorhombic LaFeO3 , but no considerable distortions are present to the overall structure. Mo ions appear to
be homogeneously doped, with average valency of both metal sites monotonically decreasing with increasing
Mo concentration. This insulator-to-metal phase transition with AFM stability is primarily understood through
intermediate interaction strengths between correlation (U) and bandwidth (W) at the Fe site, where an estimation
of this ratio is given. These results highlight the important role of extrinsic carriers in stabilizing a unique phase
transition that can guide future efforts in antiferromagnetic-metal spintronics.

DOI: 10.1103/PhysRevMaterials.4.034405

I. INTRODUCTION structure [8,9]. In oxide perovskites with the generic ABO3

formula, the transition metal (TM) dopant (B ) at the B site
Chemical substitution has proven to be a useful avenue
is often used as a design principle to stabilize carriers from
to create unusual ground states in oxide materials [1,2]. Per-
an insulating to a metallic ground state [10]. These complex
ovskite oxide compounds are known to host numerous novel
transitions, described in the Zaanen-Sawatsky-Allen scheme
ground states, such as charge ordering, orbital ordering, spin
[11], are typically characterized by the relative strengths of
ordering, etc. [1]. The interplay between the charge, orbital,
the on-site Coulomb correlation energy (U), bandwidth (W)
and spin degrees of freedom—tuned via chemical doping—
of the d shell of the transition metal, and the hybridization
has resulted in a wealth of interesting phenomena such as
strength between the TM 3d and O 2p states (t pd ). The
metal-insulator transitions, colossal magnetoresistance, high
competition between localization due to on-site Coulomb
TC superconductivity, multiferroicity, etc. [2–7]. These aspects
correlations between electrons and the kinetic energy gain due
make the prospect of identifying new electronic phases in
to band formation lies at the heart of the diverse electronic and
oxide materials very intriguing and a focus of intense re-
magnetic properties of transition metal oxides.
search. A common approach for the discovery of new oxide
In a previous study, a metalliclike behavior was reported
materials is through doping a host material, specifically tar-
for the insulating LaFeO3 (LFO) when doped with Mo ions
geted to tune between different electronic phases [3] and local
[12]. The antiferromagnetic (AFM) insulator first moves to a
small gap Mott insulator by increasing the Mo doping up to
*
[email protected]
20%. At doping levels above 20%, it shows signatures of a
†
Present address: Methods and Instrumentation for Synchrotron crossover from insulating to itinerant while maintaining anti-
Radiation, Helmholtz-Zentrum Berlin, 12489 Berlin, Germany. ferromagnetic (AFM) order, where Fe ions also show charge
disproportionation [12]. In the parent compound LaFeO3 , Fe
Published by the American Physical Society under the terms of the sites are in the +3 oxidation state. The substitution of Fe at the
Creative Commons Attribution 4.0 International license. Further B site by Mo ions in LaFe1−x Mox O3 (LFMO) is intimately
distribution of this work must maintain attribution to the author(s) linked to equal contributions of Fe2+ and Fe3+ , similar to
and the published article’s title, journal citation, and DOI. Funded that found in nickelates [13–16], and in related SrFeO3−δ [17]
by Bibsam. and La1/3 Sr2/3 FeO3 [18] systems. More interestingly, both

2475-9953/2020/4(3)/034405(9) 034405-1 Published by the American Physical Society

DIBYA PHUYAL et al. PHYSICAL REVIEW MATERIALS 4, 034405 (2020)

experimental and theoretical findings on a 25% Mo doping II. EXPERIMENT

sample establish a system that shows an orbital-selective
A. Solid-state synthesis
Mott state, where the metallic character at the Fermi level
emerges from Mo t2g states, while occupied Fe 3d states Polycrystalline ceramics were prepared by the solid-
remain positioned several eV below the Fermi level (EF ) [12]. state synthesis procedure. Stoichiometric ratios of preheated
The electronic charge disproportionation is described by two La2 O3 , Fe2 O3 , and MoO3 powders were thoroughly mixed
valences at the Fe site: Fe1 (3+) and Fe2 (2+) respectively, and the obtained fine powder was subsequently annealed at
where Fe-O bond distances are disproportionate due to nearby 1200 ˚C in a 99% Ar/1%H2 gas for 12 h and cooled to room
Mo ions but not sufficiently to drive a change in crystal temperature at 5 ˚C /min. X-ray diffraction (PANalytical MRD
structure. By increasing site substitution with Mo ions, i.e., II) was used to check the phase integrity for any impurities.
doping extrinsic carriers, the Fe2+ content increases contin- Finally, the powder was pressed to 5 GPa to form a 5-mm-
uously until a crossover from insulator to metal transition diameter pellet and subsequently sintered for 6 h at 1200 ˚C
(20% Mo) is reached. The average valence of Fe at x = 0.20 in the 99% Ar/1%H2 gas environment.
and x = 0.25 suggests an equal number of Fe3+ and Fe2+ .
Further increase of Mo doping does not seem to change the Fe B. Photoemission and high-resolution hard
valence. The additional electrons lead to a measureable band x-ray absorption near edge structure
with dominant Mo character. The electronic structure reported Photoelectron spectroscopy measurements were performed
for this compound is very rich in detail, and merits further at the Diamond Light Source using the beamline I09 (Did-
experimental information to point to competing interactions cot, UK). All data were recorded at room temperature with
and substantiate the materials’ reported physical properties. a base pressure of 2.0 × 10−10 mbar. Both soft and hard
To this purpose, we combine several x-ray spectroscopies x-ray photoelectron spectroscopies (HAXPES) were collected
to address the nature of the electronic structure and iden- in resonant condition across the Fe L3 and Mo L3 edges,
tify the role of correlations as well as any structural distor- respectively. The soft x-ray energy was selected through a
tions as a function of doping in the LaFe1−x Mox O3 system. plane grating monochromator, whereas a Si (111) double-
We investigate the effects of electron doping at three dif- crystal monochromator was used for hard x-ray energies. An
ferent doping levels: a parent compound (x = 0, LaFeO3 ), EW-4000 photoelectron analyzer (VG Scienta) with a slit
and low (x = 0.10, LaFe0.90 Mo0.10 O3 ) and high (x = 0.25, width set to 0.2 mm was used to record spectra at normal
LaFe0.75 Mo0.25 O3 ) doping, with the latter that is the max- emission. The corresponding x-ray absorption was measured
imum solid solubility limit in these systems. We employ in total electron yield (TEY) mode. The binding energy scale
resonant photoelectron spectroscopy (RPES), which provides was calibrated by setting the Au 4 f core level and EF of
an element-selective projection to the partial density of states a gold reference foil to 84.0 and 0.0 eV respectively. The
(DOS) [19,20] and specifically accesses the character of d total instrumental resolution for the PES measurements was
states near the Fermi level (EF ) and their hybridization with kept below 250 meV at the selected soft and hard x-ray
O 2p states. The transition from an insulating to a metallic energies. High-resolution absorption spectra were collected
phase is unambiguously identified through the appearance of at the Fe K edge at the CLÆSS beamline [21] of the ALBA
appreciable spectral DOS at the Fermi level (EF ). The corre- Synchrotron (Barcelona, Spain). The spectra were recorded
sponding resonant inelastic x-ray scattering (RIXS) spectra, by selecting the Kβ1,3 emission line (7058 eV) and scanning
performed with excitations across the Fe L3 edge, allow us the incoming energy across the Fe K edge by means of a
to follow the resonant electronic excitations as a function Si(311) double-crystal monochromator. The energy in the
of doping. As we progress from the insulating phase with high-resolution fluorescence detected x-ray absorption near
x  0.15 into the metallic phase with x  0.2, excitations edge Structure (HERFD-XANES) was selected using a Si
evolve from narrow energy loss features, characteristic of (333) dynamical bent diced analyzer crystal and an energy
insulating systems, to broad interband excitations, that are dispersive one-dimensional(1D) detector in Rowland circle
instead characteristic of metallic samples. geometry (Rowland radius = 1 m). The overall energy reso-
Moreover, chemical speciation, including changes to av- lution was around 0.9 eV, full width at half maximum of the
erage formal valency of the metal sites, degree of noncen- quasielastic line collected over a tape foil.
trosymmetricity of the metal-O6 environment, and short-range
ordering of the B sites in the LaFe1−x Mox O3 system, were
investigated using extended x-ray absorption fine structure C. XES and RIXS
(EXAFS). Our EXAFS results show that the Mo substitution Fe L x-ray absorption (XAS), emission (XES), and res-
at the Fe site is homogeneous. Mo doping systematically onant inelastic x-ray scattering (RIXS) measurements were
decreases the effective formal valency of both Fe and Mo performed at the 8.0.1 beamline at the Advanced Light Source
ions, without destroying the centrosymmetric nature of the (Berkeley, CA). The x-ray absorption spectrum was acquired
metal-O6 cage. The analysis from these combined techniques in both total electron yield (TEY) and total fluorescence yield
reveals the origin of itinerant carriers in a unique antifer- (TFY) modes. All measurements were performed at room
romagnetic LaFe1−x Mox O3 system. These findings not only temperature under ultrahigh-vacuum conditions (10−10 mbar)
provide insights into the origins of itinerant carriers in a with the incident beam 45˚ to the surface. The spectra have
magnetic system, but also point to distinctive properties of been normalized to the beam flux measured by a gold mesh
complex oxides. and calibrated based on a TiO2 (anatase) standard. The XES
data were energy calibrated using metallic emission lines as a

034405-2
ORIGIN OF ITINERANT CARRIERS IN … PHYSICAL REVIEW MATERIALS 4, 034405 (2020)

FIG. 1. HAXPES data measured at 2500 eV on the

LaFe1−x Mox O3 series. Both insulating LaFeO3 (red triangles)
and x = 0.10 (blue diamond) systems exhibit gaps between the
valence band and the EF , whereas for the x = 0.25 composition
(green circle), there is clear density of states at EF , characteristic of
a metal.

reference (Zn Lα,β , Mn Lα,β , Fe Lα,β ). The RIXS spectra were

taken across the Fe L3 absorption resonance energy with a
total instrumental resolution of 350 meV as measured from
the elastic profile [22–24]. FIG. 2. (a) The absorption spectrum for x = 0.25 measured at the
Mo L3 edge. The dots (•) indicate the excitation energies used to col-
D. Extended x-ray absorption fine structure (EXAFS) lect photoemission spectra. (b) RPES spectra for LaFe0.75 Mo0.25 O3
showing three enhanced spectral features that relate to Mo 4d contri-
A dual detector setup was used in simultaneous EXAFS bution in the valence band.
measurements in transmission and fluorescence modes at the
Fe K edge and the Mo K edge. The measurements were
performed at room temperature (300 K) at the Sector 10 (BE). However, the tail of the valence band still terminates
(MRCAT) bending magnet beamline at the Advanced Photon below EF with negligible intensity at 0 eV, consistent with
Source, Argonne National Laboratories (USA). For each Fe K its insulating properties. For the x = 0.25 sample, there is a
and Mo K measurement, corresponding elemental foils were nominal increase of states near 0.5 eV, with a sizable intensity
simultaneously measured to track any energy shifts in the at EF , in parallel with the transition to a metallic state.
beam during the data collection. Detailed fits to the data were In order to map the different spectral contributions in the
carried out up to the maximum usable k values (10.0 Å−1 for valence band, resonant photoemission at the Mo L3 edge
Fe K and 12.0 Å−1 for Mo K) using the standard EXAFS probes the direct Mo 4d states. Figure 2 shows the x-ray ab-
equation with the help of the ARTEMIS program [25], which sorption [panel (a)] and resonant photoemission spectra [panel
uses FEFF 6.0 code [26] to calculate the scattering amplitudes (b)] taken across the Mo L3 edge for the x = 0.25 compound.
and phases for all possible scattering events. The details of the The absorption spectrum corresponds to the Mo 2p → 4d
data analysis fits and results are tabulated in the Supplemental transitions, in close agreement with previously reported Mo
Material [27]. L edge spectra of Mo-based oxide perovskites [30–33]. The
valence band resonant photoemission is obtained by tuning
the excitation photon energy at selected energies (marked
III. RESULTS
with dots) across the Mo L3 absorption edge. The measured
Figure 1 shows the HAXPES valence band measured at an resonant photoemission spectra exhibit four distinct features
incident energy of 2500 eV for LaFe1−x Mox O3 compounds that are enhanced when tuned to resonant excitation energies
(x = 0.0, 0.10, and 0.25). The spectra were energy calibrated in Fig. 2(b). The spectral region around 0.5 eV BE shows an
relative to Au 4 f7/2 and normalized to the La 5p3/2 peak intense and narrow peak (A) that is enhanced monotonically
(17.7 eV). At high photon energies, the probing depth is rel- as the excitation energy approaches the absorption maximum
atively high and signatures from the surface do not dominate at 2523.5 eV, and decreases as the incident energy is tuned
the spectra. A close inspection of the valence band near the past the resonance condition. Feature A can be assigned to
Fermi level (EF ) shows an increase in spectral weight as a Mo 4d t2g states [12], where the itinerant electron in the
function of Mo content. For x = 0 (red curve), a sizable gap partially filled t2g states gives rise to the itinerant character
between the valence band and the EF is clearly observable, for x = 0.25. Feature B at ∼2 eV BE also shows resonant
in agreement with previous soft x-ray studies on LaFeO3 behavior, particularly at the excitation energy corresponding
[28,29]. For the x = 0.10 doped sample (blue curve), there to the absorption threshold (hv = 2522.2 eV), which points
is a nominal increase of states near 0.8 eV in binding energy to Mo d states that lie just below the Fermi level. The 4d

034405-3
DIBYA PHUYAL et al. PHYSICAL REVIEW MATERIALS 4, 034405 (2020)

FIG. 4. (a) Fe L2,3 XAS measured in TFY mode for LaFeO3

(red), x = 0.10 (blue), and x = 0.25 (green) samples. The tick marks
indicate the excitation energies used for RIXS spectra. (b) Two-
dimensional RIXS intensity map measured across the entire L2,3 edge
for the x = 0.25 sample displayed against the excitation (vertical
axis) and emission (horizontal axis) photon energy.

spectral weight at EF has no significant contributions from Fe

3d states and that metallicity can be attributed solely to Mo
4d states, forming an impurity band. The structures towards
the higher binding energy side, F and G, are dominated by the
O 2p-Fe 3d hybridized states. Moreover, these measurements
reveal features D and E having their maximum resonant
intensity at different incident energies. The inset in Fig. 3(b)
shows a narrow range of the valence band expanded at a
few selected energies (703.0, 707.5, 709.0, and well above
FIG. 3. (a) The Fe L2,3 XAS spectrum for x = 0.25 with a strong resonance at 712.1 eV). When the incident energy is tuned
shoulder centered at ∼707.5 eV. The tick marks indicate the ex- to the maximum of the shoulder peak at 707.5 eV [Fig. 3(a)],
citation energies used for RPES spectra. (b) RPES data collected an enhancement in feature D just below the EF is seen (feature
at several incident energies with four principal structures that are D, at 0.8 eV BE) and points to Fe1 t2g character according to
enhanced as compared to the off-resonance case (703.3 eV, yellow our previous calculations [12]. Conversely, at the maximum of
curve). An expanded view in the inset shows features that are related the absorption spectrum (709.0 eV), the structure E near 3 eV
to the charge disproportionated Fe states. The red curve taken at the gains substantial intensity in contrast to the off-resonance
shoulder peak is enhanced near 0.5 eV and is representative of Fe2+ energy (703.3 eV) and is associated with Fe2 t2g states. We ar-
states while the features near 3 eV are resonantly enhanced when gue the x = 0.25 compound to be a charge disproportionated
taken at the absorption maximum. system of Fe2+ (Fe1) and Fe3+ (Fe2) nominal charge states
with narrow band partially filled t2g states with contributions
at different energy positions [12]. One set of states originates
transition metal oxides (TMOs) generally have significant from Fe2+ , whose t2g states are shown to lie closer to the
bandwidths (W) and small electron correlation U due to the Fermi level, and the other from Fe3+ whose states contribute
large spatial extent of 4d orbitals. The broad region between 4 further below EF . Since feature D structure is enhanced at
and 8 eV (C) corresponds to the Mo 4d-O 2p admixture as was the low-energy shoulder feature in the Fe L XAS spectrum
shown by band structure calculations in our previous study [see Fig. 3(a)], peak D can be ascribed to Fe1 and feature
[12]. The discussion here primarily pertains to resonances E to Fe2 t2g states. The variations in spectral intensities at
near EF , and the spectra in Fig. 2(b) qualitatively show the a fixed incoming energy can thus be directly linked to the
emergence of new Mo 4d states and lend evidence for a different valency and charge distribution of Fe sites in the
delocalized character and specifically establish the metallicity compound.
in this material for doping levels at x = 0.25.
A similar experiment has been performed with photon
energies across the Fe L3 edge to probe the partial spectral A. Fe L RIXS
weight of Fe 3d states in the valence band. The corresponding RIXS can offer valuable insights into the insulating-to-
x-ray absorption spectrum collected in total electron yield metallic transition and the correlation effects within the Fe
(TEY) mode is shown in Fig. 3(a). The selected energies d band [22,23]. RIXS measurements were carried out by
for the RPES valence band data are marked with an asterisk tuning the incident photon energy across the Fe L edge for
and corresponding RPES spectra shown in Fig. 3(b). The the three compounds presented here. An overview of the
off-resonance condition is fixed at 703.3 eV and the on- Fe L edge XAS spectra collected in total fluorescence yield
resonance condition is determined from the peak position (TFY) mode is shown in Fig. 4(a). The spectral profiles are
of the Fe L3 XAS edge maximum. The spectrum shows a similar for all samples, as is the position of the absorption
significant intensity enhancement for several features (labeled maximum at 709 eV for LaFeO3 and x = 0.10, which gets
D–G) and follows a photon energy dependence. The intensity shifted slightly to 708.5 eV for the x = 0.25 sample. The
at EF shows no measurable changes, indicating that the partial lower-energy shoulder peak near 707 eV is more prominent

034405-4
ORIGIN OF ITINERANT CARRIERS IN … PHYSICAL REVIEW MATERIALS 4, 034405 (2020)

FIG. 5. Fe L2,3 edge RIXS spectra for (a) the undoped parent LaFeO3 , (b) x = 0.10 Mo-doped sample, and (c) the metallic crossover at
x = 0.25 at selected photon energies. The vertical dashed lines indicate energy positions of inelastic loss features and arrows are used to guide
the linearly dispersing fluorescencelike structure.

for x = 0.25 and is attributed largely to Fe2+ character [12]. submanifolds that can be separated to distinct energies of the
The full RIXS map for the x = 0.25 sample was taken across 2p-3d (t2g ) and 2p-3d (eg ) resonant processes. The excitations
the Fe L2,3 edges and is shown in Fig. 4(b). The RIXS map within the t2g manifold will occur at a lower energy than
taken across the Fe L2,3 edge (vertical axis) consists of three scattering involving eg levels, with the energy difference cor-
main regions. These are (from high to low emission energy) responding to the t2g and eg splitting. This allows us to assign
the Fe L2,3 edge RIXS, the Fe L2,3 /M1 , and O K XES. One can the two d-d features for LaFeO3 and x = 0.10 samples to
see that the Fe L2,3 edge RIXS signals consist of energy-loss intra-t2g processes (−1.6 eV) and t2g-eg processes (−3.1 eV).
features (Fe 3d-3d orbital excitations) that are parallel to the Excitations involving delocalized character linearly disperse
elastic line, while the “normal” fluorescence features have with incident energy (hνi ) [22,23]. Dispersing features on an
fixed emission energies. energy loss scale establish their origin as fluorescentlike decay
Figure 5 shows the measured RIXS spectra for the x = 0, of the core hole rather than Raman-type loss features. Overall,
0.10, and 0.25 samples measured at various photon energies the low- and high-energy inelastic features for both (x = 0 and
across both L3 and L2 absorption thresholds as indicated by x = 0.10) data sets show similar trends in energy positions in
tick marks in Fig. 4(a). The RIXS spectra are plotted as a both d-d and CT excitations.
function of energy loss, which is defined as the difference RIXS on LaFe0.75 Mo0.25 O3 [Fig. 5(c)] clearly shows strong
between incident and emitted photon energies. All spectra fluorescence contributions at low energy losses that are super-
were measured using the same acquisition time and have imposed on top of the d-d excitations, which is a characteristic
been normalized to the incident flux. There are several well- feature of a metallic sample [16,38], where low-energy excita-
resolved structures that can be identified based on the energy tions across the Fermi level are possible. Our first observation
loss of the excitations involved. First, the elastic line at 0 is that the spectral weight in the fluorescence features has
eV energy loss comes from the recombination of the 2p5 3d 6 increased relative to the spectral weight in the Raman modes
intermediate state to recreate the ground state, 2p6 3d 5 . Sec- for the insulating samples. Fluorescence features are already
ondly, at low energy losses, both the two insulating LaFeO3 present at the onset of absorption (705.8 eV), common to
[Fig. 5(a)] and x = 0.10 [Fig. 5(b)] samples exhibit three metallic and semimetallic systems [39]. We also observe
constant energy loss features centered around −1.6, −3.1, and that at excitation energies past the L3 edge, the CT and d-d
−5.1 eV (indicated by dashed lines). The higher-energy loss elastic features become evident as the fluorescence structures
feature at −5.1 eV is readily distinguished as a charge-transfer become well separated. However, the respective measured
(CT) excitation and corresponds to O → Fe ligand-to-metal peaks for x = 0.25 are broadened likely due to an increase in
charge transfer to hybridized Fe 3d-O 2p states. The two the bandwidth of these electronic excitations. Further support
lower-energy features (<4 eV) can be assigned to intrasite for this interpretation is provided from photoemission spectra
Fe d-d excitations. These are typical of strongly correlated that reveal both Fe d and Mo d states at 0.8 eV below the
oxides and are observed in many other similar iron-based Fermi level [see Figs. 2(b) and F3(b)]. The similarity between
compounds [34–37]. The fixed energy loss features are as- the resonance behavior and energy positions of the two d-d
signed to electronic excitations that involve localized Fe d peaks in both insulating (LaFeO3 and x = 0.10) and metallic
states. The 3d manifold is nominally split into t2g and eg (x = 0.25) compounds strongly suggests that it has the same

034405-5
DIBYA PHUYAL et al. PHYSICAL REVIEW MATERIALS 4, 034405 (2020)

FIG. 6. RIXS spectra taken at selected excitation energies for

x = 0.25 sample. The linearly dispersing fluorescence lines reveal FIG. 7. (a) Normalized Fe K HERFD-XANES (300 K) for the
several Raman features that point to strong electron-electron correla- LaFe1−x Mox O3 series; inset shows the absorption edge position
tions within the Fe d manifold. systematically shifting to lower energies with increasing Mo content.
(b) Enhanced preedge features. (c) Fit to the x = 0.2 sample. (d)
Formal valence of Fe and (e) Mo decreasing with Mo doping. (f)
Number of Mo ions (nMo ) around each Mo (blue triangles) and Fe
origin in both compounds, namely, a local d-d excitation that
(red circles) site.
points to electronic correlation.
In order to further disentangle the origin of the observed
Raman features in the x = 0.25 metallic sample, we closely the valence band has contributions from both Fe 3d and Mo
examine its RIXS data. Since the inelastic signal at the L3 4d, as highlighted in the photoemission part of this study.
resonance is actually a superposition of fluorescence, Raman This can lead to hybridization between the Fe d and the high
d-d excitations, and continuum excitations across the Fermi binding energy part of the Mo d bands as was proposed for
level, we focus on a few measurements using excitation en- FeTiO3 [42], TbMnO3 [43], and several double-perovskite
ergy below, at, and above the L3 resonance where the Raman compounds [44,45]. The presence of partially filled Mo d
and fluorescence components are well separated as shown in states extends across the Fermi level to open additional path-
Fig. 6. The excitation energies correspond to the baseline of ways for the neutral charge fluctuations. The two TMs’ octa-
the XAS peak (705.8 eV), the low-energy shoulder peak that hedra always share one oxygen at a corresponding vertex, and
corresponds to the predominantly Fe2+ L3 peak (707.0 eV), thus the intersite d-d transfer can proceed via the hybridized
the largely Fe3+ site at Fe L3 absorption maximum (708.6 eV), O 2p states. However, calculations have to be developed to
and well above the resonance (715.8 eV). All spectra are capture the full understanding of the observed data.
normalized to the intensity of the strong fluorescence peak
that shifts linearly with excitation energy. The low-energy loss
spectral components (<5 eV) (<5 eV) can still be tracked B. Fe K HERFD-XANES, Fe K XAFS, and Mo K XAFS
at higher excitation energies, as evident from the 715.8-eV Figure 7(a) shows the normalized Fe K HERFD-XANES
spectra taken in between the L3 and L2 absorption edges. for the LaFe1−x Mox O3 series. With increasing Mo content,
Furthermore, RIXS spectra for the metallic sample exhibit the absorption edge position systematically shifts to lower
clear low-energy loss intensities below 1.5 eV, providing a energies [inset of Fig. 7(a)], suggesting a decrease in the
fingerprint to closing of the insulating gap. This signature has formal valence of Fe. Figure 7(b) shows an expanded view
also been observed in several insulating-to-metallic transitions of the well-resolved preedge features. One can observe how
such as in NdNiO3 [16] and La1−x Srx MnO3 [40]. The spectra the intensity of these peaks decreases going from a pure Fe3+
taken at 707 and 708.6 eV show a complex profile that is state (LaFeO3 , in black) to the doped compounds, which are
an amalgamation of band fluorescence, electron-hole pair a mixture of Fe3+ and Fe2+ . Linear combination fits to the
continuum excitations across the Fermi level, and phenomena preedge help us to quantify the ratio of Fe2+ and Fe3+ in the
related to the presence of different Fe sites in the metallic samples, considering x = 0.0 (pure Fe3+ state) and x = 0.25
phase. We rationalize this position the following way. An (Fe2+ :Fe3+ = 1 : 1) as the end members. A representative fit
intersite d-d transition between the charge-disproportionated of the series over −20 to −5 eV across the Fe K absorption
Fe sites can occur since both sites exist in a similar high- edge of the x = 0.2 sample is shown Fig. 7(c). The formal
spin alignment (S = 2 for Fe2+ and S = 5/2 for Fe3+ ) valence per Fe site can then be calculated as a weighted
[41]. Another tangible link to the observed d-d transition can average of the number of Fe2+ and Fe3+ ions occurring in a
occur from the Fe-Mo intersite excitations mediated by the given composition. Figure 7(d) shows how the formal valence
oxygen hybridization since the high binding energy part of of Fe in the LaFe1−x Mox O3 series systematically decreases as

034405-6
ORIGIN OF ITINERANT CARRIERS IN … PHYSICAL REVIEW MATERIALS 4, 034405 (2020)

a function of Mo doping, suggesting an increasing number

of Fe2+ sites, thereby shifting the Fe K absorption edge
position to lower energies [inset of Fig. 7(a)]. Accounting
for overall charge neutrality allows us to extract the average
valency per Mo site, which also decreases with increasing
Mo concentration. It is important to note here that the fit to
Fe K HERFD-XANES for the x = 0.2 sample yields only
7% ± 2% of the x = 0.0 sample (pure Fe3+ ), suggesting the
systems x = 0.2 and x = 0.25 to be quite similar, as reported
previously [12]. Further electron doping can now fill up only
the Mo bands, and hence a larger reduction of Mo valency for
the x = 0.25 system in Fig. 7(e). These extra electrons at the FIG. 8. (a) Bond angles for both Fe and Mo measured by
Mo sites must be associated with the observed metallicity, as EXAFS. By increasing site substitution by Mo, the Fe-O-Fe bond
identified from our RPES and RIXS measurements. angle relaxes and continuously increases as a function of doping. (b)
The cooperative tilting of the octahedra is also reduced by increases
Local structural effects such as clustering, phase segre-
in Mo doping.
gation, and B-site ordering that can strongly influence bulk
transport properties were also addressed using conventional
EXAFS technique. Details of the analysis and structural pa- overlap with oxygen pπ orbitals that produce a strong
rameters extracted are tabulated in the Supplemental Material antiferromagnetic exchange [47]. The increase in TM-O-TM
[27]. The key findings are (a) Mo exists in a single environ- angle and the concomitant increase in d-d hopping should cor-
ment (absence of isosbestic points in Mo K XANES), (b) Mo respond to an increase in the bandwidth and antiferromagnetic
gets doped at Fe sites and not interstices, (c) no evidence exchange. Both insulating and metallic compounds maintain
of Mo clustering or growth of Mo-rich phases, and (d) high the same crystal structure and preserve their symmetry.
degree of local homogeneity. The fraction of Fe and Mo In order to understand the change in the strength of the
ions per B site around a Fe and Mo absorber were estimated on-site Coulomb interaction strength in going from LaFeO3
independently and plotted in Fig. 7(f), which roughly parallels to LaMoO3 , we have computed the values of bare and par-
the nominal fraction estimated from global stoichiometry of tially screened Coulomb interaction (v and U) within the
the LaFe1−x Mox O3 samples. constrained random-phase approximation (cRPA) [48–57] as
implemented in the WIEN2K code [58] (for technical details
IV. DISCUSSION
see Supplemental Material [27]). The results of our calcula-
tions (see Table S3 in the Supplemental Material [27]) show
An AFM ground state generally favors more localized that this method yielded a U = 4.41 eV for the 3d states in
electrons while FM favors an itinerant [46]. The significant LaFeO3 , and U = 2.37 eV for the 4d states in LaMoO3 . Sig-
hybridization of the localized Fe states with O 2p states nificantly, these values would clearly suggest an appreciable
allows for an energy gain due to the delocalization of Mo d decrease of about 2.0 eV in the calculated strength of U in
states. The 4d TMOs generally create the energy bands with going from the 3d to 4d states. In addition, basic electronic
significant bandwidths (W) due to the larger spatial extent structure (see Supplemental Material [27]) also indicates d-
of the 4d orbitals, thereby decreasing the overall correlation bandwidth (W) enhancement from around 4 to 6 eV upon
effects. An AFM state can be realized if the value of U is going from 3d states in LaFeO3 to 4d states in LaMoO3 ,
sufficiently small and comparable with the bandwidth [10,43]. respectively. The behavior of correlated systems depends on
A transition from an AFM insulator to an AFM metal is thus a delicate balance between the band energy and the effective
expected to occur as U/W is varied within this picture. The interaction of the correlated electrons. In the present case,
3
half-filled t2g orbitals present in the Fe3+ ions strongly favor U/W approximately changes from 1.1 to 0.4, indicating a
the anion-mediated superexchange and lead to an AFM state. 63% decrement for the 4d states compared to the 3d states.
By doping, we introduce additional electrons occupying the Therefore, our theoretical calculations clearly indicate that the
minority spin t2g states that drive the system to approach a one-electron bandwidth W of the 4d states in the x = 0.25
low U limit. For the x = 0.25 compound, Fe ions maintain sample would be large, and with that, the ratio of the on-site
their AFM characteristics despite the presence of Mo impurity Coulomb repulsion U to W is less than the critical value
bands. (U/W )c , which is likely to favor a metallic phase.
The changes to the electronic structure are not
accompanied by appreciable lattice distortions, similar to
V. CONCLUSIONS
that found in charge-disproportionated La0.33 Sr0.67 FeO3
[45]. The bond lengths and lattice distortions to the Fe-O A combined x-ray spectroscopic investigation was used to
polyhedra, which affect the electron hopping between reveal a detailed picture of the evolution of electronic structure
sites and in turn modify the size of W, vary as a result across the insulator-to-metal transition in LaFe1−x Mox O3 .
of Mo doping. By site substitution of Mo, the Fe-O-Fe Our RPES valence band data suggest that states occurring
bond angle increases from 163˚ to 175˚ for LaFeO3 at the EF are due entirely to itinerant Mo 4d states for
and metallic x = 0.25, respectively, and is shown in the x = 0.25 sample. The addition of delocalized Mo d
Fig. 8(a). This also decreases the cooperative tilting of the states and Mo-O hybridization is likely to increase the
octahedra [Fig. 8(b)], and increases the localized d-electron effective bandwidth necessary for itinerant character in the

034405-7
DIBYA PHUYAL et al. PHYSICAL REVIEW MATERIALS 4, 034405 (2020)

metallic sample. RPES taken across the Fe L3 edge suggests U and W lies in an intermediate range of interaction strengths
different spectral contributions at the edge of the valence band in LaFe0.75 Mo0.25 O3 to be simultaneously AFM and metallic.
stemming from two different Fe sites present in the metallic
LaFe1−x Mox O3 (x = 0.2 and 0.25) compound. RIXS spectra
taken at the Fe L edge for doped and undoped samples show
ACKNOWLEDGMENTS
strong electron correlation (U) for all three compounds, with
d-d and charge-transfer features that are well resolved. The We acknowledge Advanced Light Source at Lawrence
electron-doped metallic compound, with Mo concentration at Berkeley National Lab and Diamond Light Source for alloca-
x = 0.25, shows broad spectral structures that are associated tion of beam time. This work was supported financially in part
with a metallic state and are evidence for closing of the by the Swedish Energy Agency (Grant No. P43549-1), the
insulating gap. The wide bandwidth low-energy structure in Swedish Research Council (Grants No. 2016-4524 and No.
the RIXS spectra evidences both Fe inter- and intrasite excita- 2018-04330), the Knut and Alice Wallenberg Foundation, and
tions (between the charge ordered Fe sites) as well as possible the Tryggers Foundation (Grant No. CTS-17:376). MRCAT
intersite Fe-Mo excitations. These results show a suppression operations are supported by the Department of Energy and
of itinerant electron behavior and the corresponding relevance MRCAT member institutions. This research used resources
of the localized electron picture for the Fe sites in all com- of the Advanced Photon Source, a US Department of Energy
pounds. The Fe-O hybridization strength as shown through (DOE) Office of Science User Facility operated for the DOE
photoemission experiments is likely to maintain the AFM Office of Science by Argonne National Laboratory under
ordering present in the metallic sample. This balance between Contract No. DE-AC02-06CH11357.

[1] M. Imada, A. Fujimori, and Y. Tokura, Rev. Mod. Phys. 70, [18] J. Herrero-Martin, G. Subias, J. Garcia, J. Blasco, and
1039 (1998). M. Concepción Sánchez, Phys. Rev. B 79, 045121
[2] M. B. Salamon and M. Jaime, Rev. Mod. Phys. 73, 583 (2001). (2009).
[3] J. Q. Li, Y. Matsui, S. K. Park, and Y. Tokura, Phys. Rev. Lett. [19] S. J. Oh, J. W. Allen, I. Lindau, and J. C. Mikkelsen, Phys. Rev.
79, 297 (1997). B 26, 4845 (1982).
[4] C. N. R. Rao and A. Arulraj, Curr. Opin. Solid State Mater. Sci. [20] L. H. Tjeng, C. T. Chen, J. Ghijsen, P. Rudolf, and F. Sette,
3, 23 (1998). Phys. Rev. Lett. 67, 501 (1991).
[5] N. P. Armitage, P. Fournier, and R. L. Greene, Rev. Mod. Phys. [21] L. Simonelli, C. Marini, W. Olszewski, M. Ávila Pérez, N.
82, 2421 (2010). Ramanan, G. Guilera, V. Cuartero, and K. Klementiev, Cogent
[6] M. Coey, Nature 430, 155 (2004). Phys. 3, 1 (2016).
[7] D. V. Efremov, J. Van Den Brink, and D. I. Khomskii, Nat. [22] A. Kotani and S. Shin, Rev. Mod. Phys. 73, 203 (2001).
Mater. 3, 853 (2004). [23] L. J. P. Ament, M. van Veenendaal, T. P. Devereaux, J. P.
[8] A. Cammarata and J. M. Rondinelli, Appl. Phys. Lett. 108, Hill, and J. van den Brink, Rev. Mod. Phys. 83, 705
213109 (2016). (2011).
[9] P. Limelette, A. Georges, D. Jérome, P. Wzietek, P. Metcalf, and [24] K. Ishii, T. Tohyama, and J. Mizuki, J. Phys. Soc. Jpn. 82,
J. M. Honig, Science 302, 89 (2003). 021105 (2013).
[10] S. K. Pandey, P. Mahadevan, and D. D. Sarma, Europhys. Lett. [25] B. Ravel and M. Newville, J. Synchrotron Radiat. 12, 537
117, 57003 (2017). (2005).
[11] J. Zaanen, G. A. Sawatzky, and J. W. Allen, Phys. Rev. Lett. 55, [26] S. I. Zabinsky, J. J. Rehr, A. Ankudinov, R. C. Albers, and M.
418 (1985). J. Eller, Phys. Rev. B 52, 2995 (1995).
[12] S. Jana, S. K. Panda, D. Phuyal, B. Pal, S. Mukherjee, A. Dutta, [27] See Supplemental Material at http://link.aps.org/supplemental/
P. A. Kumar, D. Hedlund, J. Schött, P. Thunström, Y. Kvashnin, 10.1103/PhysRevMaterials.4.034405 for details for EXAFS
H. Rensmo, M. V. Kamalakar, C. U. Segre, P. Svedlindh, K. analysis fits and parameters used, and technical details for
Gunnarsson, S. Biermann, O. Eriksson, O. Karis, and D. D. Hubbard U calculations.
Sarma, Phys. Rev. B 99, 075106 (2019). [28] H. Wadati, D. Kobayashi, H. Kumigashira, K. Okazaki, T.
[13] M. Medarde, C. Dallera, M. Grioni, B. Delley, F. Vernay, J. Mizokawa, A. Fujimori, K. Horiba, M. Oshima, N. Hamada,
Mesot, M. Sikora, J. A. Alonso, and M. J. Martínez-Lope, Phys. M. Lippmaa, M. Kawasaki, and H. Koinuma, Phys. Rev. B 71,
Rev. B 80, 245105 (2009). 035108 (2005).
[14] J. Shi, Y. Zhou, and S. Ramanathan, Nat. Commun. 5, 4860 [29] J. E. Kleibeuker, Z. Zhong, H. Nishikawa, J. Gabel, A. Müller,
(2014). F. Pfaff, M. Sing, K. Held, R. Claessen, G. Koster, and G.
[15] S. Johnston, A. Mukherjee, I. Elfimov, M. Berciu, and G. A. Rijnders, Phys. Rev. Lett. 113, 237402 (2014).
Sawatzky, Phys. Rev. Lett. 112, 106404 (2014). [30] M. Besse, V. Cros, A. Barthélémy, H. Jaffrès, J. Vogel, F.
[16] V. Bisogni, S. Catalano, R. J. Green, M. Gibert, R. Scherwitzl, Petroff, A. Mirone, A. Tagliaferri, P. Bencok, P. Decorse, P.
Y. Huang, V. N. Strocov, P. Zubko, S. Balandeh, J. M. Triscone, Berthet, Z. Szotek, W. M. Temmerman, S. S. Dhesi, N. B.
G. Sawatzky, and T. Schmitt, Nat. Commun. 7, 1 (2016). Brookes, A. Rogalev, and A. Fert, Europhys. Lett. 60, 608
[17] M. Reehuis, C. Ulrich, A. Maljuk, C. Niedermayer, B. (2002).
Ouladdiaf, A. Hoser, T. Hofmann, and B. Keimer, Phys. Rev. [31] V. Kanchana, G. Vaitheeswaran, M. Alouani, and A. Delin,
B 85, 184109 (2012). Phys. Rev. B 75, 220404(R) (2007).

034405-8
ORIGIN OF ITINERANT CARRIERS IN … PHYSICAL REVIEW MATERIALS 4, 034405 (2020)

[32] J.-S. Kang, H. Han, B. W. Lee, C. G. Olson, S. W. Han, K. H. H. Ishii, K. D. Tsuei, and T. J. Yang, Phys. Rev. B 82, 094442
Kim, J. I. Jeong, J. H. Park, and B. I. Min, Phys. Rev. B 64, (2010).
024429 (2001). [44] T. Berlijn, P. C. Snijders, O. Delaire, H.-D. Zhou, T. A. Maier,
[33] R. Thomas, J. Kas, P. Glatzel, M. Al Samarai, F. M. F. De Groot, H.-B. Cao, S.-X. Chi, M. Matsuda, Y. Wang, M. R. Koehler, P.
R. Alonso Mori, M. Kavčič, M. Zitnik, K. Bucar, J. J. Rehr, and R. C. Kent, and H. H. Weitering, Phys. Rev. Lett. 118, 077201
M. Tromp, J. Phys. Chem. C 119, 2419 (2015). (2017).
[34] Z. Cui, X. Zhai, Y. D. Chuang, H. Xu, H. Huang, J. Wang, [45] J. Matsuno, T. Mizokawa, A. Fujimori, Y. Takeda, S. Kawasaki,
Z. Fu, R. Peng, J. Guo, and Y. Lu, Phys. Rev. B 95, 205102 and M. Takano, Phys. Rev. B 66, 193103 (2002).
(2017). [46] T. Moriya and Y. Takahashi, Annu. Rev. Mater. Sci. 14, 1
[35] T. Nomura, Y. Harada, H. Niwa, K. Ishii, M. Ishikado, S. (1984).
Shamoto, and I. Jarrige, Phys. Rev. B 94, 035134 (2016). [47] D. I. Khomskii, Transition Metal Compounds (Cambridge Uni-
[36] C. Monney, A. Uldry, K. J. Zhou, A. Krzton-Maziopa, E. versity Press, Cambridge, 2014).
Pomjakushina, V. N. Strocov, B. Delley, and T. Schmitt, Phys. [48] F. Aryasetiawan, M. Imada, A. Georges, G. Kotliar, S.
Rev. B 88, 165103 (2013). Biermann, and A. I. Lichtenstein, Phys. Rev. B 70, 195104
[37] S. M. Butorin, J. Electron Spectrosc. Relat. Phenom. 110, 213 (2004).
(2000). [49] L. Vaugier, H. Jiang, and S. Biermann, Phys. Rev. B 86, 165105
[38] D. E. McNally, X. Lu, J. Pelliciari, S. Beck, M. Dantz, M. (2012).
Naamneh, T. Shang, M. Medarde, C. W. Schneider, V. N. [50] G. Kresse and J. Hafner, Phys. Rev. B 47, 558 (1993).
Strocov, E. V. Pomjakushina, C. Ederer, M. Radovic, and T. [51] P. E. Blöchl, Phys. Rev. B 50, 17953 (1994).
Schmitt, npj Quantum Mater. 4, 1 (2019). [52] C. Martins, M. Aichhorn, L. Vaugier, and S. Biermann, Phys.
[39] C. Monney, K. J. Zhou, H. Cercellier, Z. Vydrova, M. G. Rev. Lett. 107, 266404 (2011).
Garnier, G. Monney, V. N. Strocov, H. Berger, H. Beck, T. [53] Y. Nomura, M. Kaltak, K. Nakamura, C. Taranto, S. Sakai, A.
Schmitt, and P. Aebi, Phys. Rev. Lett. 109, 047401 (2012). Toschi, R. Arita, K. Held, G. Kresse, and M. Imada, Phys. Rev.
[40] K. Ishii, T. Inami, K. Ohwada, K. Kuzushita, J. Mizuki, Y. B 86, 085117 (2012).
Murakami, S. Ishihara, Y. Endoh, S. Maekawa, K. Hirota, and [54] S. K. Panda, I. Dasgupta, E. Şaşoğlu, S. Blugel, and D. D.
Y. Moritomo, Phys. Rev. B 70, 224437 (2004). Sarma, Sci. Rep. 3, 2995 (2013).
[41] S. Grenier, J. P. Hill, V. Kiryukhin, W. Ku, Y. J. Kim, K. J. [55] B. Amadon, T. Applencourt, and F. Bruneval, Phys. Rev. B 89,
Thomas, S. W. Cheong, Y. Tokura, Y. Tomioka, D. Casa, and T. 125110 (2014).
Gog, Phys. Rev. Lett. 94, 047203 (2005). [56] A. van Roekeghem, L. Vaugier, H. Jiang, and S. Biermann,
[42] A. Agui, T. Uozumi, M. Mizumaki, and T. Käämbre, Phys. Rev. Phys. Rev. B 94, 125147 (2016).
B 79, 092402 (2009). [57] S. K. Panda, H. Jiang, and S. Biermann, Phys. Rev. B 96,
[43] J. M. Chen, J. M. Lee, S. W. Huang, K. T. Lu, H. T. Jeng, 045137 (2017).
C. K. Chen, S. C. Haw, T. L. Chou, S. A. Chen, N. Hiraoka, [58] K. Schwarz and P. Blaha, Comput. Mater. Sci. 28, 259 (2003).

034405-9