High-pressure evolution of Fe2O3electronic structure revealed by x-ray absorption
Shibing Wang,1,2,* Wendy L. Mao,2,3,4Adam P. Sorini,2Cheng-Chien Chen,2,5Thomas P. Devereaux,2,6Yang Ding,7
Yuming Xiao,8Paul Chow,8Nozomu Hiraoka,9Hirofumi Ishii,9Yong Q. Cai,10and Chi-Chang Kao11,†
1Department of Applied Physics, Stanford University, Stanford, California 94305, USA
2SIMES, SLAC National Accelerator Laboratory, Menlo Park, California 94025, USA
3Photon Science, SLAC National Accelerator Laboratory, Menlo Park, California 94025, USA
4Department of Geological and Environmental Sciences, Stanford University, Stanford, California 94305, USA
5Department of Physics, Stanford University, Stanford, California 94305, USA
6Geballe Laboratory for Advanced Materials, Stanford University, Stanford, California 94305, USA
7HPSynC, Carnegie Institution of Washington, Washington, DC 20015, USA
8HPCAT, Carnegie Institution of Washington, Washington, DC 20015, USA
9National Synchrotron Radiation Research Center, Hsinchu 30076, Taiwan
10NSLS-II, Brookhaven National Laboratory, Upton, New York 11973, USA
11SSRL, SLAC National Accelerator Laboratory, Menlo Park, California 94025, USA
?Received 15 September 2010; published 19 October 2010?
We report the high-pressure measurement of the Fe K edge in hematite ?Fe2O3? by x-ray absorption spec-
troscopy in partial fluorescence yield geometry. The pressure-induced evolution of the electronic structure as
Fe2O3transforms from a high-spin insulator to a low-spin metal is reflected in the x-ray absorption pre-edge.
The crystal-field splitting energy was found to increase monotonically with pressure up to 48 GPa, above
which a series of phase transitions occur. Atomic multiplet, cluster diagonalization, and density-functional
calculations were performed to simulate the pre-edge absorption spectra, showing good qualitative agreement
with the measurements. The mechanism for the pressure-induced electronic phase transitions of Fe2O3is
discussed and it is shown that ligand hybridization significantly reduces the critical high-spin/low-spin transi-
DOI: 10.1103/PhysRevB.82.144428PACS number?s?: 71.70.Ch, 62.50.?p, 78.70.Dm
An archetypal 3d transition-metal oxide and important
geological compound, ?-Fe2O3?hematite? undergoes a series
of structural and electronic transitions at high pressure. At
ambient conditions, Fe2O3is an antiferromagnetic insulator
and adopts the corundum structure. This structure is main-
tained until approximately 50 GPa whereupon it transforms
to a Rh2O3?II?-type structure,1accompanied by a 10% drop
in volume. The structural transition is associated with
changes in magnetic and electronic structures. X-ray K?
emission at ambient pressure and 72 GPa show that the mag-
netic moment drops from high spin ?HS? to low spin ?LS? at
high pressure.2Conductivity measurements indicate that an
insulator to metal transition occurs between 40 and 60 GPa.3
Mössbauer spectroscopy up to 82 GPa ?Ref. 3? and synchro-
tron Mössbauer spectroscopy at 70 GPa ?Ref. 4? imply the
collapse of the magnetic moments and a nonmagnetic nature
of the HP phase.
The nature of these transitions has been a popular re-
search topic over the past decade. Based on their structural
study of the Rh2O3-II phase, Rozenberg et al.1have sug-
gested that the charge-transfer gap closure is responsible
for metallization and concurrent spin moment transition.
Combined local-density approximation and dynamical mean-
field theory calculations by Kuneš et al.5have implied that
the reduction in the Mott gap with pressure drives the
volume collapse and structure change. This idea appears to
be at odds with experimental observations of a metastable
state in which the HS and high-pressure structure occur
simultaneously.6Thus, despite many studies of the transi-
tions in Fe2O3, the nature of the evolution of the electronic
structure with pressure remains unresolved. In this paper, we
implemented experimental method and theoretical ap-
proaches bringing valuable information to the problem.
A number of spectroscopic techniques have been applied
to investigate the electronic configuration of 3d transition-
metal compounds. Photoemission and x-ray L-edge absorp-
tion provide useful information on the 3d levels of transition
metals but unfortunately, these probes cannot penetrate the
high pressure cells. X-ray absorption spectroscopy ?XAS? at
the K edge of 3d transition elements, however, operates in
the hard x-ray regime, allowing the study of the electronic
structure at high pressure.
The pre-edge region of the K-edge absorption spectrum
can be used to investigate 3d electrons of transition-metal
compounds. In Fe-bearing compounds, the pre-edge spectra
contain information about the oxidation state and local
coordination.7However, limited by the 1s core-hole lifetime
broadening, the energy resolution using a transmission ge-
ometry is not sufficient to resolve the detailed structure of
the pre-edge region. Therefore we use the partial fluores-
cence yield method for measuring absorption. Instead of col-
lecting the transmitted x-ray, the K?1emission line is mea-
sured. This method thus has a 2p core-hole lifetime
broadening of about 0.3 eV, resulting in much higher energy
Here we present the first high-pressure XAS measurement
in partial fluorescence yield on Fe2O3up to 64 GPa. The
improved resolution of the resulting spectra shows the evo-
lution of the Fe3+3d electronic structure as the material un-
PHYSICAL REVIEW B 82, 144428 ?2010?
©2010 The American Physical Society144428-1
dergoes its complex pressure-induced transitions. Previously,
Caliebe et al. applied this technique to Fe2O3, and assigned
the double-peak structure of the pre-edge to the t2gand eg
components of the 3d band8as suggested previously.9Simi-
lar methods have been used to study orbital hybridization
and spin polarization of Fe2O3?Ref. 10? and pre-edges of
other Fe-containing compounds.11
Fe2O3powder was loaded in a hydrostatic pressure trans-
mitting medium ?He or Ne? in an x-ray transparent Be gas-
ket. Ruby fluorescence was used for pressure calibration.
High-pressure XAS spectra of Fe2O3were collected at two-
third generation synchrotron facilities. In both setups, mono-
chromatic x-rays focused by Kirkpatrick-Baez mirrors were
directed through a panoramic diamond-anvil cell, and the
analyzer was fixed at 90° from the incident beam.
In the SPring-8 XAS experiment conducted at BL12XU,
we scanned the incident x-ray energy from 7110 to 7145 eV
with a step size of 0.1 eV and over the smaller range of
7112–7115 eV at 0.05 eV step size. In the APS setup at
HPCAT 16-IDD, the entire edge was scanned from 7100 to
7160 eV with a step size of 0.25 eV. The pre-edge was
scanned from 7108 to 7118 eV ?7109 to 7119 eV for 56 and
64 GPa? with a step size of 0.2 eV. For both measurements,
the partial fluorescence yield was collected with the analyz-
ers set at the Fe K?1energy ?6405.6 eV?.
Figure 1?A? shows the representative Fe K-edge XAS
spectra for Fe2O3. The partial fluorescence yield geometry
allows us to resolve the pre-edge features. At the highest
pressure in our study, we collected the K?emission spectrum
of the sample shown in Fig. 1?B?. Compared with the 0 GPa
spectrum of Badro et al., there is a dramatic reduction in the
K?? satellite peak intensity in the 64 GPa spectrum, indicating
a LS ground state.2,12
As shown in Fig. 1?A? inset, it is also observed that the
K-edge blueshifts with pressure until the phase-transition re-
gion and remain approximately constant thereafter. This shift
of K edge with pressure is also observed in other 3d
transition-metal oxides,13a result of the increase in electron
density upon compression.
Figure 2 shows the Fe K-edge pre-edge spectra of the
sample from ambient pressure to 64 GPa. The tail of the
main absorption edge was subtracted for each spectrum by
removing the K-edge absorption spectrum of Fe in the Be
gasket. The pre-edge features at ambient pressure are associ-
ated with excitations to t2gand egorbitals, split by the octa-
hedral crystal field. Our ambient pressure data can be fit with
a crystal-field splitting energy ?CFSE? of 1.4 eV, consistent
with previous observation.8,9The two-peak feature in the
pre-edge persists until 48 GPa, just before the phase transi-
tions occur. By fitting the pre-edge spectra we estimate a
monotonic increase in the CFSE to 1.85 eV at 48 GPa, as
shown in Table I. This increase is expected as the FeO6oc-
tahedra shrink with pressure, and the shorter metal-ligand
distance elevates the eglevel relative to the t2glevel.
The pre-edge spectra above the phase transitions ?i.e.,
above 48 GPa? are more complicated to interpret. The full
width at half maximum of the pre-edge features significantly
broadens and a simple assignment in terms of single particle
t2gand egtransitions is inconsistent; at such pressures, Fe2O3
is in the LS state in which egshould be empty and five of the
six t2gstates occupied. Such a single-particle configuration
should lead to relatively small ?large? t2g?eg? amplitudes, un-
FIG. 1. ?Color online? ?A? X-ray K-edge absorption spectra of
Fe2O3in partial fluorescence yield geometry at 11 and 64 GPa;
Inset: Fe K-edge position at different pressures. The edge is deter-
mined by the maximum of the first derivative of the absorption
spectra. ?B? X-ray K?emission spectra of Fe2O3at 64 and 0 GPa
from Ref. 2, showing the reduction in the spin moment. Red: high-
spin state and blue: low-spin state.
Incident Energy (eV)
Normalized Intensity (arb.unit)
FIG. 2. ?Color online? X-ray K-edge pre-edge of Fe2O3at 0, 17,
29, 40, 48, 56, and 64 GPa. The bottom three spectra are from
SPring-8 using high-resolution monochromator and the top four
spectra are from APS using diamond monochromator.
WANG et al.
PHYSICAL REVIEW B 82, 144428 ?2010?
like the features observed in the pre-edge spectra at 56 and
III. THEORETICAL INTERPRETATION
To understand the pressure dependence of the XAS, we
first used crystal-field atomic multiplet theory to calculate
the electronic structure. The relevant parameters are the
atomic t2g-egenergy-level spacing 10Dq ?Ref. 14? and the
=0.346 eV appropriate for solid-state Fe3+systems,16and
perform calculations for a range of 10Dq. The lowest two
eigenenergies for the ?1s?2?3d?5configuration are shown in
Fig. 3?a? from which a HS-LS transition is evident near
10Dq=2.2 eV. For low pressure ?low Dq? the ground state
has6A1character ?HS? and crosses over at high pressure to a
state of2T2character ?LS?.17,18
While the critical value of 10Dq determined by the atomic
multiplet calculation is larger than that suggested by the ex-
perimental t2g-egpeak splitting in Fig. 2, it is well known
that the critical 10Dq for the HS-LS transition is reduced by
the Fe-O charge-transfer processes. We perform calculations
on a FeO6octahedral cluster that explicitly includes multi-
plets, ligand hybridization and charge-transfer via the Slater-
Koster matrix elements,19,20Racah parameter A, and charge-
transfer gap energy ?. At ambient pressure, the values of the
parameters are ?in units of electron volt?: Vpd?=−1.13,
Vpd?=0.65, Vpp?=0.56, and Vpp?=−0.16, A=5.0, 10Dq
=0.96, and ?=2.7.19We have used the smaller value of Vpd?
from Ref. 19. The lowering of the critical 10Dq is illustrated
in Fig. 3?a?, which shows the energies of the HS and LS
states calculated in the FeO6cluster compared to atomic
multiplet theory as a function of 10Dq. The HS to LS tran-
sition occurs at smaller 10Dq since the hybridization most
strongly couples the d5LS state with the d6L ? LS state, lying
lower in energy than the d6L ? HS state.
These parameters yield the ambient pressure spectra
shown in Fig. 3?c?, which is in good agreement with experi-
ment ?cf. Fig. 2 and Table I?. The two spectral peaks sepa-
rated by ?1.4 eV correspond to excitations into the t2gand
egorbitals, respectively, and indicate a HS ground state, with
the observed CFSE coming from 10Dq plus a 0.45 eV cova-
lent contribution. Thus while the critical 10Dq is reduced by
the Fe-O charge-transfer processes, the ligand field splitting
due to covalency pushes up the spectral t2g-egpeak separa-
tion of the XAS spectra.8
With parameters set to reproduce ambient spectra, we
consider the pressure evolution of the HS-LS transition and
the XAS spectra. As the pressure increases, both 10Dq and
the hopping integrals increase, respectively having ?d−5and
?d−4Fe-O bond-length dependence.14,19The combined ef-
fect of pressure-dependent hopping and 10Dq is explained in
the phase diagram of Fig. 3?b?. We consider several varia-
tions in Vpd?with d as shown in Fig. 3?b?, which all indicate
that the critical pressure occurs between 52 and 55 GPa.
Although variation in the exponent of Vpd?induces a varia-
tion on the order of 5% in the predicted critical pressure it is
striking to observe that the experimentally observed limits on
the critical pressure are in general agreement with theoretical
Figures 3?c?–3?e? show the calculated pre-edge XAS
spectra from the FeO6cluster at various pressures. The spec-
trum at 48 GPa ?Fig. 3?d?? shows a clear two peak structure
in the HS state, with a t2g-egpeak separation of ?1.6 eV.
The calculated CFSE is ?15% smaller in energy than experi-
ment, which may be in part due to structural deviations from
TABLE I. Crystal-field splitting energy ?CFSE? of Fe2O3as a function of pressure.
FIG. 3. ?Color online? ?a? Energy of LS state for the single atom
multiplet calculation ?dotted line? compared with the FeO6cluster
diagonalization ?dashed line? relative to the HS state ?solid line?. ?b?
HS-LS phase diagram for Fe2O3. The dotted line shows the prob-
able trajectory of ?10Dq,Vpd?? with increasing pressure ?see text?.
??c?–?e?? K-edge pre-edge XAS spectra from the FeO6cluster cal-
culation at various pressures; EAis the Fe K-edge absorption en-
ergy. ?c? At ambient pressure, the spectrum shows distinct t2g-eg
absorption peaks separated by 1.4 eV, indicating a high-spin ground
state. ?d? At 48 GPa, the peak separation is 1.6 eV, and the ground
state still resides in the high-spin sector. ?e? At 76 GPa, the spec-
trum shows broad, multiple peaks, indicating a low-spin ground
state. All the spectra were broadened with a 0.3 eV Lorentzian.
HIGH-PRESSURE EVOLUTION OF Fe2O3…
PHYSICAL REVIEW B 82, 144428 ?2010?
octahedral symmetry giving inequivalent Fe-O bonds not in-
cluded in the cluster calculation,1as well as the overall un-
certainty in cluster parameters. Figure 3?e? shows the calcu-
lated XAS spectra at 76 GPa. The high-pressure spectra have
multiple-peak features indicating a LS ground state; this
qualitative change in character of the ground state is reflected
as a qualitative change in the calculated spectra. It is the
simple transformation properties ?A1? of the HS state that
allow the XAS to be interpreted in terms of single-particle t2g
and eglevels; the final state, with one additional d electron,
transforms as A1??T2?E?=T2?E mimicking the single-
particle t2gand eglevels. On the other hand, addition of a d
electron to the LS state yields T2??T2?E?=A1?E ?T1
?T2?T1?T2resulting in more peaks than would be ex-
pected based on a single-particle interpretation.
While an insulator-metal transition is not necessarily con-
comitant with a change in the local-spin configuration ?and
vice versa?, a low-spin metallic state is always expected at a
high enough pressure. In this regime, we use the all-electron
FEFF code21–23to calculate the high-pressure Fe K-edge XAS
for a large cluster of 152 atoms in the high-pressure struc-
ture. Figure 4 shows the calculated pre-edge XAS, having
broad pre-edge features in qualitative agreement with the ex-
periment at and above 56 GPa.
We last turn to the electronic phase-transition mechanism.
Badro et al. have shown the coexistence of HS and Rh2O3-II
structure indicating that the electronic transition cannot drive
the structural transition. Kuneš et al. divided the electronic
transition into a Mott gap closing and a HS-LS gap closing,
and estimated the respective regimes of stability via a local
“density-based interaction.” Here we have indicated the im-
portance of atomic multiplets and ligand hybridization. Our
results indicate the location of the HS-LS transition can be
well described withinthe charge-transfermultiplet-
hybridization cluster approach and reasonable choices for the
pressure dependence of the cluster parameters. The reduction
in the critical pressure for the HS-LS transition in compari-
son with atomic multiplet theory due to ligand hybridization
is seen to be significant. These results lead to the prediction
that the critical pressure occurs between 52 and 55 GPa, at
values of 10Dq much smaller than would be expected from
atomic multiplet theory based on the experimental spectra.
While our cluster calculation cannot address in detail the
closing of a bulk Mott gap, the observed reduction in the
HS-LS transition pressure leads us to suggest that the physics
of a local HS-LS transition should be strongly reconsidered
as the key ingredient giving the evolution of spectral features
observed in the pre-edge XAS spectra with pressure.
In summary, we measured x-ray absorption spectra of
Fe2O3up to 64 GPa, and experimentally resolved the crystal-
field splitting and its pressure dependence through the metal-
insulator transition. The CFSE increases from 1.41 eV at
ambient conditions to 1.85 eV at 48 GPa. The pre-edge fea-
tures change drastically at higher pressures corresponding to
the range where a number of electronic and structural tran-
sitions have been reported. We constructed the phase dia-
gram for Fe2O3which shows that the changes in multiplet
structure and hybridization are important for a quantitative
estimate of the critical pressure. Based on considerations of
local cluster physics, excellent agreement between the ob-
served pressure dependence of the experimental and calcu-
lated spectra were obtained.
The authors thank W. Harrison, S. Johnston, E. Kaneshita,
and B. Moritz for helpful discussions, and thank H.-k. Mao
and J. Shu on experiments. S.W. and W.L.M. are supported
by the NSF-Geophysics under Grant No. EAR-0738873 and
Department of Energy through DOE-NNSA?CDAC? under
Grant No. DE-AC02-76SF00515. A.P.S., C.C.C., and T.P.D.
are supported by the U.S. Department of Energy under Con-
08ER46540 ?CMSN?. Y.D. is supported by EFree funded by
DOE ?Grant No. DE-SC0001057?. The experiment at
SPring-8 was performed under the approval of JASRI ?Grant
No. 2007A4264? and NSRRC ?Grant No. 2006-3-112-3?.
Portions of this work were performed at HPCAT ?Sector 16?,
Advanced Photon Source ?APS?, Argonne National Labora-
tory. HPCAT is supported by DOE-BES, DOE-NNSA, NSF,
and the W.M. Keck Foundation. APS is supported by DOE-
BES under Contract No. DE-AC02-06CH11357.
*Corresponding author; firstname.lastname@example.org
†Also at NSLS, Brookhaven National Laboratory, Upton, New York
1G. Kh. Rozenberg, L. S. Dubrovinsky, M. P. Pasternak, O. Naa-
man, T. Le Bihan, and R. Ahuja, Phys. Rev. B 65, 064112
2J. Badro, V. V. Struzhkin, J. Shu, R. J. Hemley, H.-k. Mao, C.-c.
Kao, J.-P. Rueff, and G. Shen, Phys. Rev. Lett. 83, 4101 ?1999?.
3M. P. Pasternak, G. Kh. Rozenberg, G. Yu. Machavariani, O.
Naaman, R. D. Taylor, and R. Jeanloz, Phys. Rev. Lett. 82, 4663
4S.-H. Shim, A. Bengtsonb, D. Morganb, W. Sturhahnc, K. Cat-
allia, J. Zhaoc, M. Lerchec, and V. Prakapenkae, Proc. Natl.
Acad. Sci. U.S.A. 106, 5508 ?2009?.
FIG. 4. ?Color online? Ab initio calculations with the FEFF soft-
ware of the pre-edge region of the Fe K-edge XAS spectra for the
high-pressure metallic Rh2O3?II?-type structure of Fe2O3. EFis the
WANG et al.
PHYSICAL REVIEW B 82, 144428 ?2010?
5J. Kuneš, Dm. M. Korotin, M. A. Korotin, V. I. Anisimov, and P.
Werner, Phys. Rev. Lett. 102, 146402 ?2009?.
6J. Badro, G. Fiquet, V. V. Struzhkin, M. Somayazulu, H.-k. Mao,
G. Shen, and T. Le Bihan, Phys. Rev. Lett. 89, 205504 ?2002?.
7M. Wilke, F. Farges, P.-E. Petit, G. E. Brown, Jr., and F. Martin,
Am. Mineral. 86, 714 ?2001?.
8W. A. Caliebe, C.-C. Kao, J. B. Hastings, M. Taguchi, A. Kotani,
T. Uozumi, and F. M. F. de Groot, Phys. Rev. B 58, 13452
9G. Dräger, R. Frahm, G. Materlik, and O. Brümmer, Phys. Status
Solidi B 146, 287 ?1988?.
10P. Glatzel, A. Mirone, S. G. Eeckhout, M. Sikora, and G. Giuli,
Phys. Rev. B 77, 115133 ?2008?.
11J.-P. Rueff, L. Journel, P.-E. Petit, and F. Farges, Phys. Rev. B
69, 235107 ?2004?.
12J.-P. Rueff and A. Shukla, Rev. Mod. Phys. 82, 847 ?2010?.
13A. Y. Ramos, H. C. N. Tolentino, N. M. Souza-Neto, J.-P. Itié, L.
Morales, and A. Caneiro, Phys. Rev. B 75, 052103 ?2007?.
14J. H. Van Vleck, J. Chem. Phys. 7, 72 ?1939?.
15G. Racah, Phys. Rev. 62, 438 ?1942?.
16D. M. Sherman and T. David Waite, Am. Mineral. 70, 1262
?1985?; S. Brice-Profeta, M.-A. Arrio, E. Tronc, N. Menguy, I.
Letard, C. Cartier dit Moulin, M. Nogués, C. Chanéac, J.-P.
Jolivet, and Ph. Sainctavit, J. Magn. Magn. Mater. 288, 354
17Y. Tanabe and S. Sugano, J. Phys. Soc. Jpn. 9, 766 ?1954?.
18J. Zaanen, G. A. Sawatzky, and J. W. Allen, Phys. Rev. Lett. 55,
19W. A. Harrison, Elementary Electronic Structure ?World Scien-
tific, Singapore, 2004?.
20The experimental t2g/egpeak ratio is smaller than unity, indicat-
ing other secondary excitation channels, such as electric dipole
transitions ?Refs. 24 and 25? or excitations to p density of states
?Ref. 10?. Including these effects can modify the calculated in-
tensity ratio but will not change the generic features.
21A. L. Ankudinov, B. Ravel, J. J. Rehr, and S. D. Conradson,
Phys. Rev. B 58, 7565 ?1998?; A. L. Ankudinov, C. E. Bouldin,
J. J. Rehr, J. Sims, and H. Hung, ibid. 65, 104107 ?2002?.
22Y. Ohta, T. Tohyama, and S. Maekawa, Phys. Rev. B 43, 2968
23J. Kuneš, A. V. Lukoyanov, V. I. Anisimov, R. T. Scalettar, and
W. E. Pickett, Nature Mater. 7, 198 ?2008?.
24M.-A. Arrio, S. Rossano, Ch. Brouder, L. Galoisy, and G. Calas,
Europhys. Lett. 51, 454 ?2000?.
25T. Yamamoto, X-Ray Spectrom. 37, 572 ?2008?.
HIGH-PRESSURE EVOLUTION OF Fe2O3…
PHYSICAL REVIEW B 82, 144428 ?2010?