Supplementary Materials
Materials and Methods
The FeTiO3 starting material was synthesized via a solid-state method under
controlled gas mixtures at CSIRO Minerals (Clayton, Australia) and its chemical and
phase purity was confirmed by standard electron probe microanalysis and X-ray
diffraction. The
57
FeTiO3 sample investigated was described in Ref. 26. A FeTi2O5
sample enriched in
57
Fe was synthesized by the standard solid-state reaction method.
Stoichiometric amounts of
57
Fe (95.38%), Fe2O3 (99.998%), and TiO2 (99.998%) were
mixed and ground under ethanol in an agate mortar. The mixture was loaded into a sealed
and evacuated silica tube, which was heated at 1200 0C for 44 hours, and then
subsequently quenched to ambient conditions. X-ray diffraction (XRD) and Mössbauer
spectroscopy (MS) confirmed the product to be phase-pure with orthorhombic structure
(Cmcm, Z=4).
We studied the high pressure behavior of FeTiO3 and FeTi2O5 in a four-pin
modified Merrill-Bassett design diamond anvil cell (DAC). The DAC has 300 m culets,
with a 150 or 125-m-diameter hole in the pre-indented rhenium gasket (thickness 30-40
m). Samples were loaded in NaCl, LiF or Ne pressure transmitting medium in different
experiments. Here NaCl and LiF are also used as thermal insulators for laser heating.
Equations of state of NaCl, LiF or ruby fluorescence (2-3 ruby balls of 3-5 m in
diameter were also placed in the pressure chamber) were used for pressure determination
at ambient temperature. Pressure gradients in laser-heated samples were negligible, and
uncertainty in pressure determination at 50 GPa does not exceed 2 GPa.
The experimental procedures of our laser heated DAC have been described
previously [43,44]. High-pressure powder diffraction experiments of FeTiO3 were
conducted at beamline 13-BM-D of the Advanced Photon Source (APS), Chicago. Data
were collected with a MAR345 detector using a monochromatic X-ray beam of
wavelength = 0.3344 Å and a beam size of 5x5 μm. High-pressure XRD experiments
on FeTi2O5 were conducted at the European Synchrotron Radiation Facility (ESRF),
Grenoble and Bayerisches Geoinstitut (BGI), Germany. At ESRF, the data were collected
1
with a MAR345 detector using an X-ray beam with wavelength  = 0.4133 Å. At BGI,
the XRD data were collected using a system consisting of a high-brilliance FRD rotating
anode generator and a Bruker APEX charge-coupled device area detector. An
accelerating voltage of 55 KV and a beam current of 60 mA were applied. Mo K
radiation ( = 0.7108 Å) was focused to about 50 m FWHM beam size. The collecting
time of each pattern was 30 minutes. The collected images were integrated using the
Fit2D program in order to obtain a conventional diffraction spectrum.
Mössbauer
57
Fe spectra of
57
FeTiO3 and
57
FeTi2O5 samples at elevated pressures
before and after laser heating were collected in transmission mode on a constant
acceleration Mössbauer spectrometer using a high specific activity 57Co point source in a
Rh-matrix. The velocity scale was calibrated relative to 25-m--Fe foil. The collection
time for each spectrum was at least 8 hours. The spectra were fitted to Lorentzian lineshapes using the commercial software NORMOS written by R A Brand (distributed by
Wissenschaftliche Elektronik GmbH, Germany).
2
N
Decompression
N: NaCl
W: W üstite
N
Intensity (arb.unit)
W
N
W
N
N
W
N
N
(II)
GPa
0
9
(II)
(II)
(I)
N
W
(II)
(III)
17
(III)
26
40
W
5
6
50
W
W
N
N
7
8
9
W
N
N
10
11
12
13
W
N
14
15
W
W
N
16
17
2(°)
Figure S1: Selected XRD patterns collected on decompression of FeTiO3 after laser
heating at 2000(200) K and 53(2) GPa. The X-ray wavelength is 0.3344 Å. The
backgrounds of the diffraction spectra were subtracted. On decompression, all the
reflections can be divided into three groups: one from the pressure-transmitting medium
NaCl with strong intensities and a phase transition from Pm3m to Fm3m around 26 GPa;
a second group of reflections whose full width at half maximum (FWHM) increases upon
decompression and peaks eventually disappears below 9 GPa; a third group which
belongs to the recovered material with a cubic structure attributed to wüstite. Peak labels
are: N for NaCl and W for wüstite. The arrows are guides for the eye. The red is the first
group, the blue is the second, and the green is the third.
3
-4
-3
-2
-1
0
1
2
3
4
Relative Transimission
1.5%
(d) ambient condition after full decompression
A
E
1.3%
(c) T-quenched from 50 GPa and 2000 K
B
C
A
D
1.5%
(b) 50 GPa and 300 K befroe heating
2.0%
(a) 45 GPa and 300 K
-4
-3
-2
-1
0
1
2
3
4
-1
Velocity (mms )
Figure S2:
57
Fe Mössbauer spectra of FeTiO3 collected at different conditions. LiF was
used as the pressure-transmitting medium. “A”, “B”, “C”, “D”, and “E” denote different
spectral components. (a) 45 GPa and 300 K. Perovskite FeTiO3 shows a well-resolved
Fe2+ doublet with  = 0.81 mm s-1 and EQ = 1.92 mm s-1. (b) 50 GPa and 300 K before
laser heating. The doublet splits into more than one component that is too complex to
obtain an unambiguous fitting. (c) 43 GPa and 300 K from sample laser-heated at 50 GPa
and ~2000 K. According to the XRD results, FeTiO3 dissociates into FeO and FeTi2O5.
The corresponding Mössbauer spectrum should include the components from FeO and
FeTi2O5. Thus three Fe2+ sites and one additional Fe3+ component were used in the fitting
process. “A” is attributed to wüstite with  = 0.88 mm s-1 and EQ = 1.10 mm s-1. “B” is
4
attributed to the M2 site of post-Fpb with  = 0.88 mm s-1 and EQ = 2.26 mm s-1, and
“C” belongs to the M1 site with  = 0.98 mm s-1 and EQ = 2.97 mm s-1. There is a Fe3+
component “D” with  = -0.09 mm s-1. The portion of “A” component is 50±6%, and the
ratio of A/(B+C+D) = 1(±8%). Based on the area of every iron component and the mol
ratio of 1:1 for the dissociated product the Fe3+ component is thought to be associated
with post-Fpb. (d) Ambient conditions spectrum after recovery of the high-P, T treated
material. “A” is from wüstite with  = 0.88 mm s-1 and EQ = 1.20 mm s-1, while “E” is
attributed to amorphous FeTi2O5 with  = 0.95 mm s-1 and EQ = 1.99 mm s-1. The area
ratio of A/E=1.07(±8%) is close to 1. Compared with (c) and (d), Fe3+ disappears on
decompression to ambient conditions. At the same time, FeTi2O5 becomes amorphous.
Thus the behavior of the Fe3+ component also indirectly supports the conclusion that
most Fe3+ is associated with FeTi2O5. And no Fe3+ was observed in FeTi2O5 synthesized
from a pure precursor (Fig.S4). It means the post-Fpd should be Fe1+δTi2-δO5 in the
dissociated product, rather than FeTi2O5.
5
L(002)
WB(222)
WB(311)
L(022)
WB(220)
WB(200)
WB(111)
L(111)
Amorphous
(c)
6
7
8
9
10
11
76 GPa and 300 K
12
13
14
15
L(022)
L(111)
WH(103)
WH(102)
L(002)
L(111)
(a)
L(002)
WH(100)
(b)
L(111)
WH(101)
76 GPa and 2500 K
WH(002)
Intensity (arb. unit)
5 GPa and 300 K
16
2()
Figure S3: XRD patterns of an additional experiment at (a) 76 GPa and 300 K, (b) 76
GPa and 2500(200) K, and (c) 5 GPa and 300 K quenched from (b) conditions. The Xray wavelength is 0.3344 Å. LiF was used as pressure-transmitting medium and thermal
insulator for laser heating. The backgrounds of the diffraction spectra were subtracted.
“L” represents LiF, “WH” is the NiAs-type wüstite, and “WB” is the Fm3m phase of
wüstite. There is a bulge in lower angle of 6.5º-8.0º (c), which means amorphous. Double
upward arrows denote double characteristic peaks of post-Fpd.
6
(b) after laser heating
6
8
10
12
14
16
18
obs.
calc.
peaks
Ne(111)
Intensity (arb. unit)
Intensity (arb. unit)
obs.
calc.
peaks
6
20
8
10
Ne(002)
(a) before laser heating
12
14
16
18
20
2 ()
2 ()
(c)
 = 0.7108 Å
FeTi2O5
C2/c
Pressure
GPa
4
9
15
On decompression
Intensity / a.u.
0
22
12
13
14
15
16
17
18
19
20
21
22
23
24
2 / degree
Fig. S4 X-ray diffraction patterns of FeTi2O5. (a) At 42 GPa and 300 K before laser
heating, FeTi2O5 is a low-pressure phase Cmcm; (b) Quenched from 2000±200 K at 42
GPa, the pressure decreased to 25 GPa, and FeTi2O5 transforms into a high-pressure
phase C2/c. Ne is used as the pressure-transmitting medium. (a) and (b) Data recorded at
ESRF using  = 0.4133 Å. (c) Diffraction patterns on decompression, collected at BGI
with  = 0.7108 Å.
7
-4 -3 -2 -1
0
1
2
3
4
5
4.0%
(d) ambient condition after full decompression
C1
C3
0.4%
(c)T quenched from 2000 K at 42 GPa
(b) 42 GPa and 300 K
1.3%
Relative Transimission
C2
4.0%
(a) ambient condition
-4 -3 -2 -1
0
1
2
3
4
5
-1
Velocity (mms )
Figure S5: 57Fe Mössbauer spectra of FeTi2O5 at (a) ambient conditions as synthesized, (b)
42 GPa and 300 K, (c) 42 GPa and 300 K after laser heating around 2000 K, and (d)
ambient conditions after complete decompression to ambient conditions. The cyan
fraction “C1” belongs to the Fe2+ component located on the M1 (8f) position, the gray
fraction “C2” belongs to the Fe2+ component located on the M2 (4d) position, and the
olive “C3” belongs to the Fe2+ component of amorphous FeTi2O5. The corresponding
hyperfine parameters are listed in Table S1.
8
Intensity (arb. unit)
obs.
calc.
bckgr
diff.
post-Fpd(C2/c)
wüstite (R-3m)
NaCl (Pm3m)
5
6
7
8
9
10
11
12
13
14
15
16
2()
Fig. S6 Rietveld refinement XRD pattern from a mixture of B2-NaCl (top ticks), wüstite
(middle ticks), and post-Fpb (bottom ticks) at 40 GPa. In the fitting process a simple
model was used, i.e., all Fe cations occupy M1 (8f) sites heptahedrally coordinated by
seven oxygen ions, and Ti cations occupy M2(4d) sites surrounded by six oxygen atoms
in octahedral coordination. The lattice parameters obtained for wüstite (R-3m, Z = 3) are
a = 2.8210(6) Å and c = 7.386(3) Å and the lattice parameter for NaCl (Pm3m, Z = 1) is a
=2.9544(2) Å. The structural parameters of post-ferropseubrookite (C2/c, Z = 4) are a =
10.135(3) Å, b = 4.305(1) Å, c = 10.023(4) Å,  = 141.48(9), and M1(0.109, 0.694,
0.175), M2(0.25, 0.25, 0.5), O1(0.201, 0.479, 0.124), O2(0.113, 0.028, 0.514), O3(0,
0.501, 0.25); the uncertainties in coordinates are 0.005-0.010; Rp=7.5%, wRp=5.4%.
Here: RP   Yi (obs)  Yi (calc) / Yi (obs)
and Rwp 
 w [(Y (obs)  Y (calc)] /  w [Y (obs)] 
2 1/ 2
2
i
i
i
i
i
are the R factors, where
Yi(obs) is the observed intensity at step i, Yi(calc) is the calculated intensity and Wi is
the weight.
9
(A)
10.8
Axis length (Å)
10.6
a
4.40
4.36
10.4
4.32
b
4.28
10.2
c
4.24
10.0
4.20
9.8
10
15
20
25
30
35
40
45
50
40
45
50
Pressure (GPa)
141.7
141.6
(B)

141.5
(°)
141.4
141.3
141.2
141.1
141.0
140.9
140.8
10
15
20
25
30
35
Pressure (GPa)
Figure S7: Unit cell parameters of monoclinic post-Fpb (space group C2/c) as function of
pressure. The a axis converges toward the c axis, and  increases slightly with increasing
pressure. The a axis is the most compressible, while the b axis is the least compressible.
10
Table S1. Iron hyperfine parameters in FeTi2O5, corresponding to Fig. S4. The numbers
in parentheses are the estimated standard deviations in units of the last digit, based on
fitting statistics.
C1
(mm/s)
EQ(mm/s)
A(%)
(a)
1.030(7)
1.96(2)
38(2)
(b)
0.778(7)
1.72(2)
66(4)
(c)
0.85(2)
1.73(7)
51(9)
(d)
0.992(2)
1.58(2)
33(4)
C2
(mm/s)
EQ(mm/s)
A(%)
1.066(3)
3.04(1)
62(2)
0.867(8)
2.58(2)
34(2)
0.96(2)
2.66(7)
49(9)
1.032(3)
2.70(2)
22(3)
C3
(mm/s)
EQ(mm/s)
A(%)
1.013(2)
2.15(1)
45(6)
References
43. L. Dubrovinsky, S. K. Saxena, P. Lazor, R. Ahuja, O. Eriksson, J. M. Wills, and B.
Johansson, Nature 388: 362 (1997).
44. L. Dubrovinsky, N. A. Dubrovinskaia, S. K. Saxena, H. Annersten, E. Halenius, H.
Harryson, F. Tutti, S. Rekhi, and T. LeBihan, Science 289: 430 (2000).
11