Electronically Driven Structural Distortions in Lithium Intercalates of the n = 2 Ruddlesden-Popper-Type Host Y2Ti2O 5S2: Synthesis, Structure, and Properties of Li xY2Ti2O5S2 (0 < x < 2)

Article abstract of DOI:10.1021/ja037763h

Lithium intercalation into the oxide slabs of the cation-deficient n = 2 Ruddlesden-Popper oxysulfide Y₂Ti₂O₅S ₂ to produce Li_xY₂Ti₂O ₅S₂ (0 < x < 2) is described. Neutron powder diffraction measurements reveal that at low levels of lithium intercalation into Y₂Ti₂O₅S₂, the tetragonal symmetry of the host is retained: Li_0.99(5)Y₂Ti ₂O₅S₂, I4/mmm, a = 3.80002(2) A, c = 22.6396(2) A, Z = 2. The lithium ion occupies a site coordinated by four oxide ions in an approximately square planar geometry in the perovskite-like oxide slabs of the structure. At higher levels of lithium intercalation, the symmetry of the cell is lowered to orthorhombic: Li_0.99(5)Y ₂Ti₂O₅S₂, Immm, a = 3.82697(3) A, b = 3.91378(3) A, c = 22.2718(2) A, Z = 2, with ordering of Li⁺ ions over two inequivalent sites. At still higher levels of lithium intercalation, tetragonal symmetry is regained: Li _1.52(5)Y₂Ti₂O₅S₂, I4/mmm, a = 3.91443(4) A, c = 22.0669(3) A, Z = 2. A phase gap exists close to the transition from the tetragonal to orthorhombic structures (0.6 < x < 0.8). The changes in symmetry of the system with electron count may be considered analogous to a cooperative electronically driven Jahn-Teller type distortion. Magnetic susceptibility and resistivity measurements are consistent with metallic properties for x > 1, and the two-phase region is identified as coincident with an insulator to metal transition.

Full text of DOI:10.1021/ja037763h

Published on Web 01/31/2004
Electronically Driven Structural Distortions in Lithium
Intercalates of the n ) 2 Ruddlesden-Popper-Type Host
Y2Ti2O5S2: Synthesis, Structure, and
Properties of LixY2Ti2O5S2 (0 < x < 2)
Geoffrey Hyett, Oliver J. Rutt, Zolta´n A. Ga´l, Sophie G. Denis,
Michael A. Hayward, and Simon J. Clarke*
Contribution from the Inorganic Chemistry Laboratory, Department of Chemistry,
UniVersity of Oxford, South Parks Road, Oxford, OX1 3QR, U.K.
Received August 5, 2003; E-mail: simon.clarke@chem.ox.ac.uk
Abstract: Lithium intercalation into the oxide slabs of the cation-deficient n ) 2 Ruddlesden-Popper
oxysulfide Y2Ti2O5S2 to produce LixY2Ti2O5S2 (0 < x < 2) is described. Neutron powder diffraction
measurements reveal that at low levels of lithium intercalation into Y2Ti2O5S2, the tetragonal symmetry of
the host is retained: Li0.30(5)Y2Ti2O5S2, I4/mmm, a ) 3.80002(2) Å, c ) 22.6396(2) Å, Z ) 2. The lithium
ion occupies a site coordinated by four oxide ions in an approximately square planar geometry in the
perovskite-like oxide slabs of the structure. At higher levels of lithium intercalation, the symmetry of the
cell is lowered to orthorhombic: Li0.99(5)Y2Ti2O5S2, Immm, a ) 3.82697(3) Å, b ) 3.91378(3) Å, c ) 22.2718(2)
Å, Z ) 2, with ordering of Li+ ions over two inequivalent sites. At still higher levels of lithium intercalation,
tetragonal symmetry is regained: Li1.52(5)Y2Ti2O5S2, I4/mmm, a ) 3.91443(4) Å, c ) 22.0669(3) Å, Z ) 2.
A phase gap exists close to the transition from the tetragonal to orthorhombic structures (0.6 < x < 0.8).
The changes in symmetry of the system with electron count may be considered analogous to a cooperative
electronically driven Jahn-Teller type distortion. Magnetic susceptibility and resistivity measurements are
consistent with metallic properties for x > 1, and the two-phase region is identified as coincident with an
insulator to metal transition.
such as K3C60,13 are superconductors. Layered oxides such as
Introduction
the numerous and technologically important Ruddlesden-
Intercalation is important in the modification of materials
properties.1,2 Reductive topotactic insertion of alkali metals into
transition-metal oxides3 and sulfides4 has been explored exten-
sively in the context of producing electrochromic devices and
in battery technology.5-7 Other classes of solid such as
fullerenes,8 nitrides,9 oxyhalides,10 and nitride halides have also
been investigated as hosts for intercalation, and the lithium
intercalates of the layered group 4 nitride halides â-ZrNCl11
and â-HfNCl12 and the fullerides of the heavier alkali metals,
Popper (R-P) phases (An+1MnX3n+1),14 which consist of per-
ovskite slabs n octahedra thick separated by so-called rock salt
layers, accept nonredox intercalants to increase the separation
of the slabs15 or exfoliate them.16 Some R-P phases, such as
Sr3Ru2O7,17 undergo topotactic oxidatiVe intercalation of fluorine
into the rock salt layers. Oxysulfides are a still rare class of
solid-state compound in which the different chemical require-
ments of the two anions often result in materials with layered
structures. The properties of these compounds are expected to
provide new opportunities in solid-state science, and recent
interest has been kindled in the layered LnCuOS compounds,18
which consist of alternate LnO fluorite and CuS antifluorite
layers. The groups of Tremel19 and Meerschaut20 have reported
independently the layered oxysulfides with general formula
(1) Murphy, D. W. AdVances in the Synthesis and ReactiVity of Solids:
Research Annual; JAI Press: Greenwich, CT, 1991; Vol. 1, p 237.
(2) Rouxel, J.; Tournoux, M. Solid State Ionics 1996, 84, 141.
A
(3) Chippindale, A. M.; Dickens, P. G.; Powell, A. V. Prog. Solid State Chem.
1994, 21, 133.
(4) Eichhorn, B. W. Progress in Inorganic Chemistry; Karlin, K. D., Ed.;
Wiley: New York, 1994; Vol. 42.
(5) Bruce, P. G. Chem. Commun. 1997, 1817.
(6) Murphy, D. W.; Sunshine, S. A.; Zahurak, S. M. NATO AdV. Sci. Inst.
Ser., Ser. B 1987, 172, 173.
(13) Hebard, A. F.; Rosseinsky, M. J.; Haddon, R. C.; Murphy, D. W.; Glarum,
S. H.; Palstra, T. T. M.; Ramirez, A. P.; Kortan, A. R. Nature 1991, 350,
600.
(7) Murphy, D. W. NATO AdV. Sci. Inst. Ser., Ser. E 1985, 101, 181.
(8) (a) Rosseinsky, M. J.; Murphy, D. W.; Ramirez, A. P.; Fleming, R. M.;
Zhou, O. Springer Ser. Solid-State Sci. 1993, 117, 11. (b) Murphy, D. W.;
Rosseinsky, M. J. NATO AdV. Sci. Inst. Ser., Ser. B 1993, 305, 73.
(9) Bowman, A.; Mason, P. V.; Gregory, D. H. Chem. Commun. 2001, 1650.
(10) (a) Jacobson, A. J. In Solid State Chemistry: Compounds; Cheetham, A.
K., Day, P., Eds.; Oxford University Press: New York, 1992, pp 182-
233. (b) Halbert, T. R. Intercalation Chem. 1982, 375-403.
(11) Yamanaka, S.; Kawaji, H.; Hotehama, K.; Ohashi, M. AdV. Mater. 1996,
8, 771.
(14) Ruddlesden, S. N.; Popper, P. Acta Crystallogr. 1958, 11, 541.
(15) Mohan Ram, R. A.; Clearfield, A. J. Solid State Chem. 1994, 112, 288.
(16) Schaak, R. E.; Mallouk, T. E. Chem. Mater. 2000, 12, 3429.
(17) Li, R. K.; Greaves, C. Phys. ReV. B 2000, 62, 3811.
(18) Ueda, K.; Takafuji, K.; Hosono, H. J. Solid State Chem. 2003, 170, 182.
(19) Goga, M.; Seshadri, R.; Ksenofontov, V.; Gu¨tlich, P.; Tremel, W. Chem.
Commun. 1999, 979.
(20) Boyer, C.; Deudon, C.; Meerschaut, A. C. R. Acad. Sci. Paris, Ser. II 1999,
2, 93.
(12) Yamanaka, S.; Hotehama, K.; Kawaji, H. Nature 1998, 392, 73.
9
1980
J. AM. CHEM. SOC. 2004, 126, 1980-1991
10.1021/ja037763h CCC: $27.50 © 2004 American Chemical Society

Electronically Driven Distortions in Li Intercalates
A R T I C L E S
filled recirculating drybox with a combined O2 and H2O content of
less than 5 ppm. Orange Y2Ti2O5S2 and brown Nd2Ti2O5S2 were
prepared on the 2-10 g scale by reacting stoichiometric quantities of
Y2O3 or Nd2O3, TiO2, and TiS2 at 1100 °C in sealed silica tubes. The
lanthanide oxides (Y2O3: Aldrich 99.99%; Nd2O3: ALFA 99.99%)
were dried at 900 °C for 24 h in air and then removed to the drybox,
TiO2 (Aldrich 99.9+%) was dried at 250 °C in air, and TiS2 was
prepared by reacting Ti (ALFA 99.99%, dehydrided) with S (ALFA
99.9995%) at 600 °C for 3-4 days in evacuated silica tubes.
(Caution: the temperature was raised from 400 to 600 °C over 24 h to
avoid a buildup of sulfur pressure.) The starting materials were ground
together thoroughly and loaded into silica tubes which had been baked
at 900 °C under a vacuum of 2 × 10-2 mbar for several hours to remove
adsorbed moisture. This prebaking step was important to avoid
contamination of the oxysulfides with Ln2Ti2O7 phases (especially in
the case of Y2Ti2O5S2). The tubes were sealed under a vacuum of 2 ×
10-2 mbar, and the brown or orange oxysulfides were obtained phase
pure according to laboratory powder X-ray diffraction by heating the
tubes at 1100 °C for 3-4 days. Lithium intercalation was carried out
in three different ways using either lithium vapor or reducing solutions
containing the Li+ ion. First, 2-5 g of the oxysulfide powders were
reacted with excess n-BuLi (Aldrich, 2.5 M in hexane; Li/Ti ratios of
between 7:1 and 10:1) at between 20 and 50 °C for 1-60 days under
a nitrogen atmosphere. The samples were filtered and washed three
times with distillation-dried hexane and then dried under vacuum prior
to storage in the drybox. In the second low-temperature approach,
applied to Y2Ti2O5S2, excess lithium (Li/Ti ratios of between 1:1 and
10:1) was dissolved in 30-50 cm3 of liquid ammonia (BOC 99.98%
and dried over sodium) in one arm of a glass “H”-cell30 (designed to
withstand an internal pressure of up to 15 atm) and poured onto 0.5-3
g of Y2Ti2O5S2 located in the other arm. The solution was stirred
overnight at -78 °C after which the powder had changed in color from
orange to black. The unreacted metal-ammonia solution was separated
from the solid by means of a porous glass filter separating the two
arms of the “H”-cell, and the solid was washed by condensing the
ammonia back onto the solid and repeating the filtration step until the
liquid in contact with the solid product was colorless. The ammonia
was then removed to another vessel, and the “H”-cell was evacuated
to 2 × 10-2 mbar and removed to the drybox prior to extraction of the
solid product. In some cases, this treatment was then repeated with
fresh Li/NH3 solution for several periods of 24 h. If temperatures around
0 °C (Caution: the glassware must be designed to withstand the vapor
pressure of ammonia and the pressure of any evolved H2) were used
for the intercalation using Li/NH3, the LixY2Ti2O5S2 material formed
in the reaction was found to catalyze the decomposition of the solution
to LiNH2 and H2 over a period of a few hours (indeed, decomposition
of alkaline earth/ammonia solutions was rapid even at -78 °C, and
this route cannot be used for intercalation of the alkaline earths). These
samples prepared using “chimie douce” approaches were used for
neutron powder diffraction investigations and are summarized in Table
1. Alternatively, lithium vapor was used as the reducing agent: up to
2 g of the oxysulfide powders were placed in nickel tubes (Ni 222
alloy, 99.9% pure, 10 cm long, 9-mm inner diameter, 0.5-mm wall
thickness, and sealed at one end) and a smaller stainless steel tube (5-
mm inner diameter, open at one end) containing some lithium pieces
(Li/Ti ratios of between 2:1 and 10:1) was placed on top of the powder.
This assembly was used so that the oxysulfide powder was in contact
only with lithium vapor and not the liquid. The nickel tube was sealed
by arc-welding under argon (purified using a Ti getter at 800 °C), and
the whole was sealed in an evacuated silica jacket to avoid oxidation
of the nickel at elevated temperatures. These tubes were placed upright
in a muffle furnace and heated at 400-700 °C for periods of 1-6
weeks.
Ln2Ti2O5S2 (Ln ) Pr - Er, Y). These compounds resemble
the n ) 2 R-P phases, but the 12-coordinate site in the
perovskite-like slabs is vacant, and the ordering between oxide
and sulfide results in the oxide ions being confined to the
perovskite slabs, while the sulfide ions occupy the rock salt
layers. These compounds contain a reducible transition-metal
cation, and their structural features resemble those of the ReO3
structure as well as those of layered transition-metal sulfides,
both of which will accept reducing intercalants. We have
recently demonstrated21 that sodium may be inserted either into
the 12-coordinate site in the perovskite-like oxide slab (under
thermodynamic control), producing R-NaxY2Ti2O5S2 (0 < x e
1), or into a tetrahedral site in the sulfide layer (under kinetic
control), producing â-NaY2Ti2O5S2 (with no apparent range of
compositions), which is to our knowledge the first example of
reductive intercalation into the rock salt layers of a Ruddlesden-
Popper-type phase. The two phases of formula NaY2Ti2O5S2
(Ti oxidation state of +3.5) are both black with dominant Pauli
paramagnetism, and detailed investigations of the R-NaxY2-
Ti2O5S2 (0 e x e 1.0) intercalates22 and the intercalates with
alkali metals inserted into the sulfide layers23 are reported
elsewhere. Here we describe the effect on the crystal structures
of the insertion of variable amounts of lithium into the vacant
sites in the oxide slabs of Y2Ti2O5S2. Some initial measurements
on lithium intercalates of Nd2Ti2O5S2 are reported, but the
yttrium system is reported in detail as this is the only member19
for which measurements of magnetic properties are not com-
plicated by the presence of a paramagnetic lanthanide ion. The
smaller size of lithium means that, unlike in the sodium case,
there is no kinetic barrier to insertion into the perovskite blocks
even well below room temperature, and this is always favored
over insertion into the rock salt-type sulfide layers. Furthermore,
the lithium ion is accommodated in a different site in the oxide
layers to that occupied by the sodium ion in the analogous
sodium intercalates,21 and lithium intercalates LixLn2Ti2O5S2 for
0 < x e 1.85(5) (i.e., approaching full occupancy of the lithium
sites (x ) 2) and with Ti oxidation states approaching +3) are
chemically accessible. The details of the structures of these
intercalates are strongly dependent on electron count. This work
is complementary to the many investigations of reduced titanium
oxides including Ti2O3,24 LnTiO3 (Ln ) Ce - Lu, Y),25
La1-xSrxTiO3,26 NaTiO2,27 and the superconducting spinel
LiTi2O4,28,29 which show the interplay between crystal structure,
electronic structure, and temperature for reduced titanate systems
in the proximity of the metal-insulator boundary.
Experimental Section
Synthesis. The Li intercalates are all extremely air-sensitive and
oxidize in a matter of seconds when exposed to air. All manipulations
of solids were therefore carried out in a Glovebox Technology argon-
(21) Denis, S. G.; Clarke, S. J. Chem. Commun. 2001, 2356.
(22) Clarke, S. J.; Denis, S. G.; Rutt, O. J.; Hill, T. L.; Hayward, M. A.; Hyett,
G.; Ga´l, Z. A. Chem. Mater. 2003, 15, 5065.
(23) Rutt, O. J.; Hill, T. L.; Ga´l, Z. A.; Hayward, M. A.; Clarke, S. J. Inorg.
Chem. 2003, 42, 7906.
(24) Mott, N. F. J. Phys. 1981, 42, 277.
(25) Greedan, J. E. J. Less-Common Met. 1985, 14, 335.
(26) Hays, C. C.; Zhou, J.-S.; Markert, J. T.; Goodenough, J. B. Phys. ReV. B
1999, 60, 10367.
(27) Clarke, S. J.; Fowkes, A. J.; Harrison, A.; Ibberson, R. M.; Rosseinsky,
M. J. Chem. Mater. 1998, 10, 372.
(28) Johnston, D. C.; Prakash, H.; Zachariasen, W. H.; Viswanathan, R. Mater.
Res. Bull. 1973, 8, 777.
(29) Harrison, M. R.; Edwards, P. P.; Goodenough, J. B. Philos. Mag. B 1985,
52, 679.
(30) Wayda, A. L.; Dye, J. L. J. Chem. Educ. 1985, 62, 356.
9
J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004 1981

A R T I C L E S
Hyett et al.
Table 1. Summary of Synthetic Methods Used To Obtain Materials LixY2Ti2O5S2
x in LixY Ti2O S
synthetic method
Li/Ti mole ratio
T/°C
duration
structural analysis
color
2
5
2
0.30(5)
0.66(5)a
0.75(5)a
0.95(5)
0.99(5)
1.14(5)
1.30(5)
1.50(5)
1.52(5)
1.85(5)
n-BuLi/hexane
n-BuLi/hexane
n-BuLi/hexane
n-BuLi/hexane
n-BuLi/hexane
n-BuLi/hexane
Li/NH3(l)
n-BuLi/hexane
Li/NH3(l)
Li/NH3(l)
7:1
10:1
7:1
7:1
10:1
7:1
2:1
10:1
1:1
20
20
20
50
25
50
-78
50
24 hours
7 days
7 days
24 hours
7 days
7 days
D2B
POLARIS
D2B
D2B
POLARIS
D2B
dark green
dark green
dark green/black
black
black
black
black
black
black
black
4 × 48 hours
bnot measured
POLARIS
POLARIS
POLARIS
60 days
-78
-78
24 hours
3 × 24 hours
2:1
a
b
Two-phase materials. Orthorhombic lattice parameters derived from PXRD data.
Chemical Analysis. Analysis for lithium was carried out using a
Thermo Elemental Atomscan 16 ICP analyzer. The lithium ions were
leached out of the intercalates by shaking the materials with deionized
water.
showed that the samples, which have small paramagnetic moments,
usually contained minuscule amounts of ferromagnetic impurities
(probably nickel), which saturated at fields above 1 T. Corrections were
carried out for core diamagnetism using standard tables.32
Diffraction Measurements. Powder X-ray diffraction (PXRD)
measurements to assess phase purity and determine lattice parameters
were carried out using a Siemens D5000 diffractometer operating in
Debye-Scherrer geometry with Cu KR1 radiation selected using a
Ge(111) monochromator. The samples were ground with amorphous
boron (1:1 mass ratio) to limit sample absorption and sealed in 1-mm
diameter glass capillaries. Powder neutron diffraction (PND) measure-
ments were carried out using the diffractometer POLARIS at the ISIS
Facility, Rutherford Appleton Laboratory, U.K., or the high-resolution
diffractometer D2B at the Institut Laue-Langevin, Grenoble, France.
On POLARIS, samples of between 2 and 5 g in mass were measured
in the d-spacing range 0.4 < d < 8 Å by means of three banks of
detectors located at scattering angles 2θ of 35° (3He tube detector),
90° (ZnS scintillator), and 145° (3He tube detector, highest resolution
bank: ∆d/d ) 5 × 10-3) for a total integrated proton current at the
production target of approximately 500/m µAhr where m is the mass
in grams of the sample. On D2B samples of between 4 and 5 g in
mass were measured in the 2θ range 5-150° using neutrons of
wavelength 1.595 Å selected using a 28-crystal Ge(335) monochromator
(0.83 < d < 18.3 Å). The diffraction pattern was measured using 64
3He tube detectors spaced at intervals of 2.5° in 2θ, and the entire
detector bank was moved over this interval in steps of 0.05°.
Measurements were made over periods of 6-8 h. Low-temperature
measurements on D2B or POLARIS were carried out at 2 or 5 K
achieved using an ILL “orange” cryostat. One sample was additionally
measured on POLARIS at up to 400 °C in a Rutherford-Appleton
Laboratory-designed furnace (RAL F2) with a cylindrical vanadium
heating element and an evacuated sample space. All samples for neutron
diffraction measurements were contained in 10-mm diameter thin-walled
vanadium cans, which were sealed in the drybox with indium gaskets
(or with a soft copper gasket in the case of measurements above room
temperature). Rietveld refinement against PND data was carried out
using the GSAS suite of programs.31 Refinement against POLARIS
data was carried out using all three detector banks simultaneously.
Magnetic Susceptibility Measurements. Measurements were car-
ried out using a Quantum Design MPMS2 SQUID magnetometer in
the temperature range 5-320 K and at magnetic fields of up to 5 T.
Material (30-120 mg) was loaded into predried (100 °C) gelatine
capsules, which were sealed with superglue immediately after removal
from the drybox, and then immediately loaded into the magnetometer
to avoid degradation of the samples due to the permeability of the
gelatine capsules. Measurements of the susceptibility were made by
measuring the magnetic moment of the paramagnetic samples as a
function of temperature at fields of either 3 and 4 T or 4 and 5 T and
determining the gradient of moment against field in this region. This
was necessary because measurements of sample moment against field
Results and Discussion
Synthesis. The powdered samples turned from orange/brown
to dark green after approximately 1 h of stirring with n-BuLi at
room temperature and to black after several days (Table 1).
When Li/NH3 solution, a considerably stronger reducing agent
than n-BuLi,6 was used as the reductant the solid turned black
after a few hours even at -78 °C. The PXRD patterns of
samples of LixY2Ti2O5S2 covering the range 0 < x < 1.85(5)
(Table 1) obtained by these methods were all indexed on a unit
cell very similar in dimensions to that of the starting materials,
and the intensity distributions of the diffraction patterns were
also similar, indicating that insertion of lithium is topotactic.
Compounds with compositions LixY2Ti2O5S2 with 0 < x < 0.6
or x > 1.5 were indexed on a body-centered tetragonal cell while
materials with intermediate compositions 0.7 < x < 1.3 were
indexed on a body-centered orthorhombic cell with aorth ≈ borth
≈ atet, corth ≈ ctet. Compositions LixY2Ti2O5S2 with 0 < x e
1.5 were all achieved by reaction with n-BuLi in hexane:
compositions may be controlled in this range by changing the
reaction time and temperature as summarized in Table 1.
Lithium intercalation using Li/NH3 solution is much more rapid,
and LixY2Ti2O5S2 with x ) 1.5 was achieved in only 24 h at
-78 °C, in contrast to the 60 days at 50 °C required using
n-BuLi. In the n-BuLi reactions, the reducing nature of the
intercalate leads to side reactions producing LiH when the
highest Li content (x ≈ 1.5) materials obtainable by this method
are synthesized. The more reducing Li/NH3 solution enabled
Li contents of up to x ) 1.85(5) to be achieved using three 24
h treatments of the solid at -78 °C with fresh Li/NH3 solution
and intermediate workup. However, the LixY2Ti2O5S2 products
with x approaching 2 catalyze the decomposition of the
metastable Li/NH3 solution to LiNH2 and H2. Considerable
amounts of LiNH2 were therefore obtained in addition to the
highest Li content materials. LiNH2 is only sparingly soluble
in liquid ammonia and was not washed out during workup of
the reaction. Subsequent washing of the sample with tetrahy-
drofuran which had been dried immediately prior to use by
distillation over potassium metal enabled all the LiNH2 to be
removed (as judged by comparison of the measured Li content
of the washed sample, with the expected value based on the
measured lattice parameters) with no apparent deintercalation
(31) Larson, A.; von Dreele, R. B. The General Structure Analysis System; Los
(32) Magnetic properties of transition metal compounds; Landolt-Bornstein New
Alamos National Laboratory: Los Alamos, NM, 1985.
Series II/10, supplement 2; Springer-Verlag: Berlin, 1979.
9
1982 J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004

Electronically Driven Distortions in Li Intercalates
A R T I C L E S
Table 2. Results of Rietveld Refinements of the Structures of Single-Phase LixY2Ti2O5S2 Materials against Neutron Powder Diffraction Data
Collected at ILL (instrument D2B) or ISIS (instrument POLARIS)
compound
Li0.30(5)Y Ti2O S
Li0.99(5)Y Ti2O S
Li1.52(5)Y Ti2O S
Li1.85(5)Y Ti2O S a
2 5 2
2
5
2
2
5
2
2
5
2
instrument
T/K
D2B (1.595 Å)
298
POLARIS
298
POLARIS
298
POLARIS
298
space group
a/Å
b/Å
I4/mmm
3.80002(2)
a
22.6396(2)
326.919(5)
27
2.448
0.0515
0.0477
Immm
I4/mmm
3.91831(3)
a
I4/mmm
3.94260(8)
a
3.83189(3)
3.91881(3)
22.3007(2)
333.876(6)
98
2.261
0.0149
0.0161
c/Å
22.0652(3)
338.770(8)
77
21.9578(9)
341.31(2)
88
V/Å3
variables
ø2
0.716b
0.909c
wRp
0.0151
0.0398
0.0074
0.0856
R(F2)
a
b
c
Contains substantial amount of LiNH2. ø2 < 1 due to slight undercollection of data. ø2 < 1 due to large incoherent background arising from LiNH2
impurity.
of lithium from the oxysulfide phase (lattice parameters before
and after washing were unchanged). The products of insertion
of Li vapor at temperatures of 400-500 °C were similar to those
obtained by the reaction with n-BuLi. At reaction temperatures
of 600 °C and above, additional phases including YTiO3 were
present in the PXRD patterns. The soft route using n-BuLi
allows much more control of the composition and allows for
the synthesis of larger quantities of higher purity material than
does the vapor intercalation route. Consequently, materials made
by the soft routes were used for the structural and property
measurements.
Structures. The refinements of the structures of the lithium
intercalates of Y2Ti2O5S2 and Nd2Ti2O5S2 were carried out using
the structures of the parent oxysulfides as starting models with
lattice parameters for each intercalate derived from laboratory
X-ray powder diffraction measurements. Analysis of the result-
ing Fourier difference map for a tetragonal compound of
stoichiometry Li1.38(4)Nd2Ti2O5S2 indicated that in contrast to
the analogous sodium intercalates,21,22 in which Na+ is located
in the center of the 12-coordinate sites in the oxide slabs, Li+
was located in the center of the 4-coordinate, approximately
square “windows” separating pairs of adjacent 12-coordinate
vacancy sites. There are two such sites per 12-coordinate
vacancy in Ln2Ti2O5S2, and thus the limiting stoichiometry for
Li intercalation is Li2Ln2Ti2O5S2 (TiIII) which is almost achiev-
able in the case of Li1.85(5)Y2Ti2O5S2 using Li/NH3 solution.
Rietveld analysis of the structures of two orthorhombic com-
pounds Li0.95(5)Y2Ti2O5S2 and Li1.14(5)Y2Ti2O5S2 using data
collected on D2B was initially carried out assuming that the
two now-inequivalent four-coordinate Li+ sites were equally
occupied. Refinement of the fractional occupancies of these two
sites showed instead that only the smaller and squarer of the
two sites is occupied by Li for compositions up to LixY2Ti2O5S2
with x ≈ 1. The orthorhombic distortion at certain levels of Li
intercalation (and, hence, electron count as discussed further
below) therefore results in ordering of the Li+ cations over the
available sites. Once the location of Li in the tetragonal and
orthorhombic phases had been determined, most of the refine-
ments proceeded without incident. The very wide range of
d-spacings available on POLARIS allowed for reliable refine-
ment of anisotropic displacement ellipsoids for all atoms.
Refinement of anisotropic ellipsoids for Li in Li0.99(5)Y2Ti2O5S2
led to a decrease in the agreement factor ø2 to 2.262 from a
value of 2.598 in the case of isotropic Li ellipsoids. The
displacement ellipsoids for Li in all these intercalates are
extremely elongated toward the centers of the 12-coordinate
Figure 1. Results of Rietveld refinements of the structures of Li0.99(5)Y2-
Ti2O5S2 (top) and Li1.52(5)Y2Ti2O5S2 (bottom) against POLARIS data (145°
detector bank). The measured (points), calculated (line), and difference
(lower line) profiles are shown. Tick marks indicate allowed reflections
for the LixY2Ti2O5S2 phase (lower set), 1.7(1) mole % of a Y2Ti2O7 impurity
(middle set), and vanadium sample container (upper set). The insets show
magnifications of the low d-spacing regions.
vacancy sites. This is suggestive of high Li+ ion mobility in
the solids, which is consistent with the much greater air
sensitivity of these compounds when compared with the
analogous sodium intercalates, from which the sodium ions may
only be leached using dilute nitric acid solution.21 The mobility
of the lithium ions is under investigation. The refinement results
for the range of compositions from Li0.30(5)Y2Ti2O5S2 to
Li1.85(5)Y2Ti2O5S2 are presented in Table 2. Figure 1 compares
the calculated and observed diffractograms following Rietveld
refinement of the structures of Li0.99(5)Y2Ti2O5S2 and Li1.52(5)Y2-
Ti2O5S2 at 295 K against POLARIS data. The compound
Li1.85(5)Y2Ti2O5S2, prepared using Li/NH3 solution, contained
a considerable amount of LiNH2 (as discussed above), which
was not noticeable in the X-ray diffraction pattern. This material
9
J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004 1983

A R T I C L E S
Hyett et al.
or by using anisotropic strain broadening terms in the description
of the profile function were all unsatisfactory. Annealing of the
sample Li0.66(5)Y2Ti2O5S2 at 380 °C for 4 days in a predried
evacuated silica ampule produced a material which was appar-
ently tetragonal according to laboratory X-ray powder diffraction
with a c lattice parameter almost unchanged compared with the
unannealed material and with an a parameter close to the mean
of a and b in the apparently orthorhombic unannealed material.
Investigation of the annealed sample of Li0.66(5)Y2Ti2O5S2 on
POLARIS showed that neither a single tetragonal phase in I4/
mmm (ø2 ) 8.26; wRp ) 0.0370) nor a single orthorhombic
phase in Immm (ø2 ) 1.898; wRp ) 0.0177) could adequately
account for the data, but the data could be modeled satisfactorily,
according to visual inspection and the numerical agreement
factors, using approximately equal proportions of a tetragonal
phase and an orthorhombic phase (ø2 ) 1.431; wRp ) 0.0153).
The two-phase refinement was carried out with anisotropic
displacement ellipsoids for all atoms except Ti, with ellipsoid
parameters of corresponding atoms in the two phases constrained
as far as possible to be equal while respecting the different
symmetries (a reasonable assumption based on the similarity
of the two phases and required to enable refinement of the
relative phase fractions), and with a similar constraint applied
to peak profile parameters pertaining to the two lower-resolution
data banks. All other structural parameters and the profile
parameters for the more-discerning high-resolution (2θ ) 145°)
data bank were independent for the two phases. The low
scattering length of Li (-1.90 fm) relative to the other nuclei
in the sample (Y ) 7.75 fm, Ti ) -3.30 fm, O ) 5.805 fm, S
) 2.847 fm), the intrinsically highly anisotropic nature of the
Li ellipsoid, the various constraints applied, and correlations
between refined parameters meant that although the Li contents
of the two phases refined stably and reflected the overall Li
content of the sample (although they were not constrained to
do so), the Li content of the tetragonal phase was lower than
expected (0.46(2)), and the Li content of the orthorhombic phase
was higher than expected (1.04(4)) based on the lattice
parameters and bond lengths of the two phases (see, for example,
Figure 2). The refinement results for this two-phase refinement
are shown in Figure 3 and Table 4. This two-phase behavior
which was observed, although not investigated in detail, for
LixNd2Ti2O5S2 phases of similar composition, is indicative of
a composition gap between tetragonal materials and those which
adopt the orthorhombic structure. When materials with com-
positions in the gap region are made at room temperature, they
apparently consist of more than two phases: annealing results
in the formation of one tetragonal phase and one Li-richer
orthorhombic phase, but single-phase material within certain
composition limits (see below) is apparently not stable with
respect to disproportionation at room temperature. The structural
parameters for the tetragonal and orthorhombic phases which
are assumed to represent the edges of the phase gap region are
presented in Table 5. The sample Li0.66(5)Y2Ti2O5S2 was
measured at elevated temperatures to determine whether a single
phase with this composition was attainable via comproportion-
ation. In the refinements against room temperature data, the
tetragonal 200 and 204 reflections were particularly poorly
accounted for using a single tetragonal phase. At 200 °C, the
full widths at half-maximum of these reflections had 74% of
their room temperature values, while the manifold of the most
Figure 2. Lattice parameters and unit cell volume for LixY2Ti2O5S2 (0 <
x e 2) as functions of x. a (b) and b (O) (upper) and c (b) and volume (O)
(lower). The shaded area represents the two-phase region. The lines are
guides to the eye.
was measured on POLARIS prior to washing with THF, and
LiNH2 was found to dominate the neutron diffraction pattern.
In this case a two-phase (scattering fractions: Li2Y2Ti2O5S2:
28%, LiNH2: 72%) refinement using previously published
structural data for LiNH233 was satisfactory but did not permit
stable refinement of anisotropic displacement ellipsoids for the
atoms in Li1.85(5)Y2Ti2O5S2. Li was assumed, following the
results of refinements for the other compounds, to be the only
significantly anisotropic atom in this phase, and its ellipsoid
was fixed at the dimensions resulting from the refinement for
Li1.52(5)Y2Ti2O5S2. The cell parameters ((a + b)/2 for the
orthorhombic phases) vary approximately linearly with x as
shown in Figure 2. Atomic parameters are presented in Table
3, and other structural data are described in detail below.
Materials with compositions around Li0.7Y2Ti2O5S2, prepared
using n-BuLi (stirred at 20 °C for 7 days), appeared to be
orthorhombic when investigated by laboratory X-ray powder
diffraction; however, refinements in the space group Immm of
single orthorhombic phases for the sample of composition
Li0.75(5)Y2Ti2O5S2 (lattice parameters a ) 3.8186(1) Å, b )
3.8758(1) Å, c ) 22.400(1) Å) against D2B data and for the
sample of composition Li0.66(5)Y2Ti2O5S2 (lattice parameters a
) 3.8194(1) Å, b ) 3.8608(1) Å, c ) 22.417(1) Å) against
POLARIS data were unsatisfactory. Attempts to account for
the data by lowering the symmetry to one of the monoclinic
space groups I2/m, C2/c, P21/n, or the triclinic space group I1h
(33) Jacobs, H.; Juza, R. Z. Anorg. Allg. Chem. 1972, 391, 271.
9
1984 J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004

Electronically Driven Distortions in Li Intercalates
A R T I C L E S
Table 3. Atomic Parameters for Single-Phase Samples LixY2Ti2O5S2 Obtained from Refinement against Neutron Powder Diffraction Data
atom/site
refined parameter
Y Ti2O S a
Li0.30(5)Y Ti2O S
Li0.99(5)Y Ti2O S
Li1.52(5)Y Ti2O S
Li1.85(5)Y Ti2O S
2
5
2
2
5
2
2
5
2
2
5
2
2
5
2
I4/mmm
I4/mmm
Immm
I4/mmm
I4/mmm
Y (0 0 z)
z
0.333589(9)
0.53(1)
0.07867(2)
0.39(1)
0.099313(8)
0.57(1)
0.33378(4)
0.49(2)
0.08035(8)
0.35(8)
0.09905(4)
0.59(2)
0.33471(1)
0.59(1)
0.08458(2)
0.31(2)
0.10243(2)
0.64(1)
0.09366(2)
0.76(2)
0.78(2)
0.20472(3)
0.54(2)
4.16(2)
1.002(8)
0.33555(2)
0.61(1)
0.08758(4)
0.36(2)
0.09732(2)
1.15(2)
0.33637(7)
0.45(2)
0.0888(1)
0.56(5)
0.09730(7)
1.33(3)
100 × (Uiso,eq/Å2)
Ti (0 0 z)
z
100 × (Uiso,eq/Å2)
O1a (1/2 0 z)
O1b (0 1/2 z)
z
100 × (Uiso,eq/Å2)
z
100 × (Uiso,eq/Å2)
100 × (Uiso,eq/Å2)
z
O2 (0 0 0)
S (0 0 z)
1.05(1)
0.20469(2)
0.55(2)
1.48(4)
0.20448(9)
0.36(5)
4.2(7)
0.73(2)
0.20484(5)
0.48(3)
4.97(3)
0.78(1)
0.85(6)
0.2032(2)
0.39(5)
4.97c
100 × (Uiso,eq/Å2)
100 × (Uiso,eq/Å2)
Li fractionb
Li (0 1/2 0)
0.15c
0.98(3)
a
b
From ref 21. For tetragonal samples, x in LixY2Ti2O5S2 corresponds to double the fractional occupancy; for orthorhombic samples it is equal to the
c
fractional occupancy. Not refined.
Table 4. Results of Rietveld Refinements of the Structures of
Phases Constituting the Sample Li0.66(5)Y2Ti2O5S2 against Neutron
Powder Diffraction Data Collected on POLARIS at 298 K (sample
is two phase) and at 483 K (sample is single phase)
compound
Li0.66(5)Y Ti2O S
Li0.66(5)Y Ti2O S
2 5 2
2
5
2
instrument
T/K
POLARIS
298
POLARIS
483
space group
phase fraction
a/Å
I4/mmm
0.58(1)
3.83590(5)
a
22.4566(5)
330.43(1)
93
1.431
0.0153
0.0249
Immm
0.42(1)
3.8189(2)
3.8542(2)
22.468(1)
330.70(2)
I4/mmm
1.0
3.83378(4)
a
22.5319(3)
331.170(9)
95
1.039
0.0092
0.0664
b/Å
c/Å
V/Å3
variables
ø2
wRp
R(F2)
confirmed that the material Li0.66(5)Y2Ti2O5S2 was single phase
above 200 °C. The results of this refinement are shown in Tables
4 and 5, and the fit is shown in Figure 3. After the sample was
cooled to room temperature, the 200 and 204 reflections
broadened, indicating disproportionation of the sample to the
previously determined orthorhombic and tetragonal phases.
Measurements of the diffraction pattern after further heating
above 200 °C and further cooling to room temperature con-
firmed that the transformation was reversible (see inset to Figure
3).
Figure 3. Results of Rietveld refinements of the sample with composition
Li0.66(5)Y2Ti2O5S2 against POLARIS data (145° detector bank). The
measured (points), calculated (line), and difference (lower line) profiles are
shown, and tick marks indicate allowed reflections. The upper diffraction
pattern measured at 298 K was modeled using 42(1) % orthorhombic
“Li0.8Y2Ti2O5S2” (lowest set of tick marks), 58(1) % tetragonal “Li0.6Y2-
Ti2O5S2” (uppermost set of tick marks), and 1.7(1) mole % of a Y2Ti2O7
impurity (middle set of tick marks). The lower diffraction pattern, measured
at 483 K, was modeled using a single tetragonal phase (lower set of tick
marks) and the Y2Ti2O7 impurity (upper set of tick marks). Peaks arising
from the vanadium sample holder and furnace element were excluded. The
left-hand insets show magnifications of the low d-spacing regions. The right-
hand inset accompanying the high-temperature diffraction pattern shows,
from top to bottom, the narrowing of the manifold arising from the 200/
020 reflections on warming from 298 K (two-phase material) to 483 K
(one-phase material) and the broadening which accompanies dispropor-
tionation on cooling back to 298 K.
The structures of the intercalates are shown in Figure 4, and
the detail of the TiO5S coordination polyhedron as a function
of composition is shown in Figure 5. The relationship between
structure and electron count will be discussed for the yttrium
derivatives. The Ti-O, Ti-S, and Li-O bond lengths and the
O-Ti-O bond angles for the full range of LixY2Ti2O5S2
21
intercalates, and including Y2Ti2O5S2 as the x ) 0 member,
are shown in Figures 6-8. In Y2Ti2O5S2, Ti is in approximately
octahedral coordination by four “equatorial” O atoms (O1) at
1.9427(1) Å, one “apical” O atom (O2) at 1.7941(4) Å, and by
an S atom at the opposite apex 2.8741(6) Å distant. The Ti-S
distance is a long bonding distance in Y2Ti2O5S2, and Ti lies
out of the plane described by the four equatorial O atoms and
toward the apical O atom as shown in Figure 5. On intercalation
of lithium, electrons enter bands which are Ti-O π-antibonding;
all the Ti-O distances therefore increase as shown in Figure 6,
and furthermore, the Ti atom moves toward the plane of the
equatorial O atoms (Figure 5). Thus, the basal a lattice parameter
(or the mean of a and b in the orthorhombic cases) increases
intense and almost perfectly overlapping 116 reflections main-
tained its width on heating to 200 °C. Refinement of a single
tetragonal phase in I4/mmm against data measured at 210 °C
proceeded without incident, and agreement factors of ø2 ) 1.039,
wRp ) 0.0092, coupled with an excellent visual fit to the data
9
J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004 1985

A R T I C L E S
Hyett et al.
Table 5. Atomic Parameters for the Two Phases Used To Model the Neutron Powder Diffraction Pattern of Material with Composition
Li0.66(5)Y2Ti2O5S2 at 298 K and the Corresponding Parameters for the Single Phase Required at 483 Ka
atom/site
parameter
“Li0.6(1)Y Ti2O S ”
“Li0.8(1)Y Ti2O S ”
Li0.66(5)Y Ti2O S
2 5 2
2
5
2
2
5
2
T/K
phase fraction
298
0.58(1)
298
0.42(1)
483
1.0
space group
I4/mmm
Immm
I4/mmm
Y (0 0 z)
z
0.33459(6)
0.3337(1)
0.61(4)b
0.0824(2)
0.43(1)b
0.1012(1)
0.98(2)b
0.0965(1)
0.98(2)b
1.09(8)b
0.2030(2)
0.61(4)b
9.3(3)b
0.33413(2)
1.03(2)
0.08215(4)
0.58(3)
0.09869(2)
1.33(2)
100 × (Uiso,eq/Å2)
0.74(2)b
Ti (0 0 z)
z
0.0826(1)
0.43(1)b
100 × (Uiso,eq/Å2)
O1a (1/2 0 z)
O1b (0 1/2 z)
z
0.09888(5)
100 × (Uiso,eq/Å2)
z
0.98(2)b
100 × (Uiso,eq/Å2)
100 × (Uiso,eq/Å2)
z
O2 (0 0 0)
S (0 0 z)
1.14(4)b
0.2056(1)
0.47(4)b
9.3(3)b
2.12(3)
0.20455(4)
0.83(4)
19(1)
100 × (Uiso,eq/Å2)
100 × (Uiso,eq/Å2)
Li fraction
Li (1/2 0 0)
0.23(1)
1.04(4)
0.34(1)
a
b
The compositions of the two phase sample are estimated to be Li0.6(1)Y2Ti2O5S2 (tetragonal) and Li0.8(1)Y2Ti2O5S2 (orthorhombic). Anisotropic parameters
for corresponding atoms constrained to be equal as far as allowed by symmetry.
a rather narrower phase gap (0.6 < x < 0.8) than that suggested
by the refined lithium contents of the two phases (0.5 < x <
1).
Physical Property Measurements. The magnetic suscepti-
bilities of the materials measured on POLARIS are shown as
functions of temperature in Figure 9. The major contribution
to the susceptibility in each case is a temperature-independent
term suggesting that the materials behave as Pauli paramagnets,
and the electrons are largely delocalized. Small Curie contribu-
tions are consistent with around 1-3% of the reduced Ti3+
1
species behaving as isolated S )
/ moments. The only
2
exception to this is the least reduced compound, Li0.30(5)Y2-
Ti2O5S2, in which the Curie contribution is much more important
than in the other compounds and corresponds to 26% of the
added electrons behaving as isolated S ) 1/2 moments. Resistiv-
ity measurements performed by cold-pressing 5-mm diameter
pellets (1 mm thick) at 0.5 GPa produced values of 760, 50,
and 18 Ωcm for the samples Li0.30(5)Y2Ti2O5S2, Li0.99(5)Y2-
Ti2O5S2, and Li1.52(5)Y2Ti2O5S2, respectively. The resistivities
of the two lithium-richer compositions are fairly typical for
unsintered pellets of fine powders of metallic oxides, suggesting
that single-phase materials with x > 1 are metallic, while the
higher resistivity of Li0.30(5)Y2Ti2O5S2 suggests that materials
with low lithium contents are semiconducting. This is consistent
with the magnetic measurements, the colors of the compounds
(Table 1), and the changes in structural features with x as
described below. The increase in Pauli susceptibility with
increasing x in LixY2Ti2O5S2 (Figure 9) indicates a proportional
increase in the density of states at the Fermi level as the Ti-O
antibonding bands are filled, in line with the calculations
described below.
Figure 4. Structure of orthorhombic Li0.99(5)Y2Ti2O5S2 showing the
intercalation of Li into the vacant sites within the oxide layers. The four
Li-O bonding distances are indicated by dotted lines in the left-hand part
of the figure. The other parts of the figure show views, down the long axis,
of the LixO (0 < x < 2) sheets showing anisotropic displacement ellipsoids
(99% level) in Li0.99(5)Y2Ti2O5S2 (center) and tetragonal Li1.52(5)Y2Ti2O5S2
(right).
Figure 5. TiO5S coordination polyhedra in, from left to right, Y2Ti2O5S2,21
Li0.99(5)Y2Ti2O5S2 (orthorhombic), Li1.52(5)Y2Ti2O5S2, and Li1.85(5)Y2Ti2O5S2.
Displacement ellipsoids obtained from refinements at room temperature are
shown at the 99% level (isotropic ellipsoids were used for Li1.85(5)Y2-
Ti2O5S2). Bond lengths are given in angstroms.
approximately monotonically with electron count as shown in
Figure 2. The long Ti-S bond shortens dramatically to 2.512(5)
Å in Li1.85(5)Y2Ti2O5S2 (Figure 6); this and the response of the
YS rock salt layers to the increase in the basal dimensions lead
to an approximately monotonic decrease in the c lattice
parameter with electron count (Figure 2) despite the increase
in the Ti-O2 bond length. The cell volume increases ap-
proximately linearly with electron count as shown in Figure 2.
The changes in lattice parameters and bond lengths are ap-
proximately linear with x for the single-phase samples, and the
corresponding values for lattice parameters and bond lengths
derived from the refinements for the two phase sample
Li0.66(5)Y2Ti2O5S2, which we have investigated in detail, suggest
Band Structure Calculations and Interpretation of Struc-
tural Changes. To rationalize the tetragonal-to-orthorhombic-
to-tetragonal structural changes as a function of x in LixY2-
Ti2O5S2, extended Hu¨ckel calculations were performed using
the YaEHMOP34 software package with the parameters listed
in Table 6. The three-dimensional band structure and the
corresponding density of states (DOS) diagram for the material
with the highest experimentally attained Li content, Li1.85(5)Y2-
(34) Yet Another Extended Hu¨ckel Molecular Orbital Package (YAEHMOP)
Home Page; Cornell University: Ithaca, NY, 1997. http://yaehmop.source-
forge.net/.
9
1986 J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004

Electronically Driven Distortions in Li Intercalates
A R T I C L E S
Figure 7. Li-O1 (O), Li-O2 (b), and mean Li-O (0) distances as
functions of x for intercalates LixY2Ti2O5S2. The shaded area indicates the
two-phase region. The lines are guides to the eye.
Figure 8. O-Ti-O angles as functions of x for intercalates LixY2-
Ti2O5S2: O2-Ti-O1 (or O2-Ti-O1b) (b), O2-Ti-O1a (O), and O1-
Ti-O1 (or O1a-Ti-O1b) (0). The shaded area indicates the two-phase
region.
to the formal Ti4+ assignment and semiconducting behavior
observed for the undoped material.35 On topotactic intercalation
of Li, electrons fill the lowest available Ti 3d-based bands near
-9 eV. The bands are almost dispersionless along Γ f Z and
X f P, consistent with electronic isolation of the oxide slabs
along the crystallographic c axis perpendicular to the layers.
Thus, analysis of a single, two-dimensional Ti2O8/2OS2 slab can
be employed without losing any relevant features present in three
dimensions. To follow the changes associated with increasing
levels of intercalation in detail, the orbital contributions of the
bands near the Fermi level need to be examined. Molecular
orbital calculations on isolated, hypothetical TiO5S units with
the experimentally observed geometries show that the distribu-
tion of the Ti 3d orbitals closely resembles the familiar t2g-
below-eg sets derived from an octahedral crystal field. Since
the titanium coordination environment is distorted from Oh
symmetry by replacement of one apical oxygen by sulfur, the
degeneracy of the “t2g” orbitals, which are first to be populated
in the course of intercalation, is lifted. For all experimentally
Figure 6. Axial Ti-O2 bond lengths (top), equatorial Ti-O1 (Ti-O1a
(b) and Ti-O1b (O) for the orthorhombic phases) bond lengths (middle),
and the Ti-S bond lengths (bottom) as functions of x for intercalates LixY2-
Ti2O5S2. The shaded area indicates the two-phase region.
Ti2O5S2, are presented in Figure 10. The valence states of O
and S are fully occupied, corresponding to formally completed
octets, and there is considerable mixing (∼30%) of Ti states
below the Fermi energy as depicted in Figure 10, indicating
significant covalent character in the Ti-O and Ti-S bonds.
The bands at and slightly above the Fermi level are mainly Ti
3d in character with a non-negligible admixture of O 2p and S
3p states. These crystal orbitals belong to the antibonding
π*(Ti-O, Ti-S) manifold. The high energy states of Y and Li
remain unoccupied, corresponding to their full valence ioniza-
tion. Calculations on the unintercalated Y2Ti2O5S2 host yield
fully occupied S 3p states and empty Ti 3d bands corresponding
(35) Ishikawa, A.; Takata, T.; Kondo, J. N.; Hara, M.; Kobayashi, H.; Domen,
K. J. Am. Chem. Soc. 2002, 124, 13542.
9
J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004 1987

A R T I C L E S
Hyett et al.
oxygen neighbors in the ab plane compared with that in the
dxz,yz orbitals, and hence, the corresponding bandwidth of the
dxy-based crystal orbital is expected to be roughly twice as large
as that of the dxz,yz bands. This unequal dispersion of the bands
based on the “t2g” orbitals will change the orbital order suggested
by the molecular picture in Figure 11a and results in the dxy
states being lowest in energy at the Γ point for all geometries
of intercalation. The interpretation of the band structures
presented in Figure 11b follows from the above considerations.
The six bands shown in each diagram contain a majority of
their contributions from the six Ti 3d states derived from the
hypothetical Ti2O8/2OS2 molecular units. Following the argu-
ments above, the (dxy bands follow each other closely while
the (dxz,yz band centers are separated by ∼0.5 eV, and the
dispersion of the dxy bands is greater than that of the dxz,yz bands.
As a consequence, the bands to be first occupied are the (dxy
combinations, which can interact with the equatorial O1 2px,y
orbitals in a π* fashion (for the orthorhombic cases, these
equatorial oxygen atoms are designated O1a and O1b). How-
ever, due to translational symmetry requirements at Γ, no oxygen
orbital is allowed to mix with the dxy orbitals at this reciprocal
lattice point, and the oxygen orbital contribution to the band
remains quite small in going from Γ f X,Y. This is shown in
the DOS diagram presented in Figure 11c, where the orbital
contribution due purely to dxy states follows the total DOS very
closely at low levels of lithium insertion (small x). While the
Figure 9. Magnetic susceptibilities as functions of temperature for
Li0.66(5)Y2Ti2O5S2 (two-phase mixture) (b), Li0.99(5)Y2Ti2O5S2 (O), Li1.52(5)Y2-
Ti2O5S2 (9), and Li1.85(5)Y2Ti2O5S2 (0). The inset additionally shows the
larger Curie tail arising from Li0.30(5)Y2Ti2O5S2 ([).
Table 6. Orbital Parameters Used in Extended Hu¨ckel
Calculations
atom
orbital
H (eV)
ii
úi1
c1
úi2
c2
Ti
4s
4p
3d
2s
2p
3s
3p
5s
5p
4d
2s
2p
-8.97
-5.44
1.0750
1.0750
4.55
-10.81
-32.30
-14.80
-20.00
-11.00
-5.3370
-3.5150
-6.7990
-5.3420
-3.4990
0.4206
1.40
0.7839
dxy orbital occupation increases by 7% from 0.323 in Y2Ti2O5S2
O
S
2.2750
2.2750
2.1220
1.8270
1.2790
1.0790
2.5540
1.0750
1.0750
to 0.465 in Li0.30(5)Y2Ti2O5S2, the Ti-O equatorial COOP values
(Table 7) show very little weakening reflected in a very small
increase in the corresponding bond lengths (1.9427(1) Å in
Y2Ti2O5S2 and 1.9466(4) Å in Li0.30(5)Y2Ti2O5S2). The integrated
COOP curves in Figure 11d show that the Ti-O apical bonds
are strongest, while the Ti-O equatorial bonds are slightly
weaker, with the Ti-S apical bonds being weaker still. Low
levels of intercalation have little influence on these bonds, and
the inserted electrons will mostly occupy Ti 3dxy states (Table
8), diminishing the already almost negligible Ti-Ti δ-bonding
states listed at the bottom of Table 7. Due to the very small
admixture of O 2px,y at small values of x, the added electrons
are expected to be significantly localized in Ti 3dxy states, and
this may explain the high levels of localized magnetic moments
observed in the case of Li0.30(5)Y2Ti2O5S2 (see above).
Y
0.6020
1.0680
0.5780
Li
determined geometries in the LixY2Ti2O5S2 series, the dxy orbital
is slightly lower in energy than the dxz,yz orbital set, which
becomes nondegenerate in the orthorhombic structures with the
dyz orbitals being slightly more stabilized. This is illustrated for
three LixY2Ti2O5S2 compositions on the sides of the three panels
of Figure 11a. Joining two hypothetical TiO5S units at their O
apex, to reflect the double layer present in Y2Ti2O5S2 and its
derivatives, allows the d orbitals to mix further with their
appropriate symmetry counterpart, thus giving rise to a total of
six states, which are reflected in the centers of the three panels
of Figure 11a. The ( combinations of dxy orbitals are separated
by a small energy gap (∼0.1 eV), since their overlap is delta
type and thus no orbitals at their joint oxygen apex could further
participate in their mixing. The dxz,yz orbitals, however, mix with
each other in a π fashion which can also involve the px,y orbitals
of the apical oxygen (O2), which amounts to a larger (0.5 eV)
energy difference. The antibonding (in-phase, (+)) dxz,yz com-
binations will be more stabilized than the dxz,yx - dxz,yz bonding
pair, since the in-phase (+) combination precludes mixing with
the apical oxygen’s orbitals, while the out-of-phase (-)
combinations allow for the inclusion of O2 2px,y orbitals. The
molecular models help in interpreting the outcome of the two-
dimensional calculations, but an additional feature needs to be
accounted for in going from a molecular orbital to an extended
crystal orbital system: the dxy orbitals have twice as many
As x in LixY2Ti2O5S2 approaches 0.7, the Fermi level reaches
the base of the dxz,yz bands. Structures having slightly greater
levels of Li intercalated show the orthorhombic distortion of
the crystal structure. The largest difference in the a and b lattice
parameters (b - a ) 0.09 Å) occurs near the composition
Li0.99(5)Y2Ti2O5S2, calculations for which are presented in the
middle panels of Figure 11. The lowering of symmetry removes
the degeneracy of the dxz and dyz states, and two geometrical
parameters have great influence on the nature of these bands:
the different Ti-O1a and Ti-Ob bond lengths (Figure 6, Table
7) and the O2-Ti-O1a and O2-Ti-O1b angles (Figure 8),
which reflect the departure of the oxygen atoms from the plane
described by the Ti atoms. The Ti-O1b bond length is only
slightly longer (0.01 Å) than the Ti-O1a distance. Much more
noticeable (Figure 8) is the 5.8° difference between the O2-
Ti-O1a (101.74(2)°) and O2-Ti-O1b (95.90(2)°) angles: the
O1b atoms located along the b-axis (the longer axis) lie much
closer to the plane described by the Ti atoms than the O1a atoms
9
1988 J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004

Electronically Driven Distortions in Li Intercalates
A R T I C L E S
Figure 10. Band structure (left) and DOS (right) for Li1.85(5)Y2Ti2O5S2. The Fermi energy is indicated by the dashed horizontal line at -9 eV. The black
projected DOS insets show the distribution of oxygen-based orbitals, while the medium and light gray areas show S and Ti orbital contributions, respectively.
Dashed lines indicate integrated occupation of atomic states.
along the a-axis (the shorter axis), as shown in Figure 5. The
energetic stabilization of the dyz states over the dxz states depicted
in Figure 11c is a direct consequence of this change in geometry
between the tetragonal and orthorhombic symmetries. Trans-
lational symmetry requirements mean that at Γ, the dxz,yz orbitals
are unable to engage in any bonding with the O1a or O1b 2pz
orbitals, irrespective of the level of out-of-plane bending. The
O1a-2px/O1b-2py orbitals do, however, have nonzero overlap
with Ti dxz,yz orbitals because the O1a/O1b atoms lie out of the
plane of the Ti atoms. Near the Fermi energy, the mixing
between these orbitals will be antibonding. Because the O1-
2px/py-Ti-3dxz/dyz mixing vanishes when the oxygen and
titanium atoms lie in the same plane, the closer the oxygen atoms
are to the plane of Ti atoms, the less will be the destabilization
of the corresponding Ti 3d-based crystal orbital. The O1b atoms
approach the Ti plane more closely than the O1a atoms do;
thus, the dyz orbitals lie lower in energy than the dxz orbitals, as
shown in Figure 11, parts a-c. The orbital occupation of the
dyz states increases from 0.362 to 0.527 on going from x )
0.30(5) to x ) 0.99(5), while the orbital occupation of the
higher-energy dxz orbitals remains almost unchanged (0.362 in
Li0.30(5)Y2Ti2O5S2 and 0.372 in Li0.99(5)Y2Ti2O5S2). However,
the integrated COOP values show negligible contrast between
the Ti-O1a and Ti-O1b bonding (0.489 and 0.488, respec-
tively, at x ) 0.99(5)), and both bonds are only slightly weaker
than at lower levels of intercalation (0.497 at x ) 0.30(5)). The
retention of the Ti-O1b bond strength is possible since the
proportion of O1b-2py orbitals present below -9 eV in the dyz
crystal orbital is still very small. The orthorhombic distortion
allows the accommodation of additional electrons with only a
slight decrease in the Ti-O1b bond strength since the distortion
is such that the occupied 3dyz orbitals become less antibonding.
The Ti-O1a bonds remain strong and short as possible since
the distortion allows occupancy of the 3dxz antibonding orbitals
to be avoided.
For x g 1.5, the structure returns to tetragonal symmetry.
The results presented on the right-hand panels of Figure 11
correspond to the structure of Li1.85(5)Y2Ti2O5S2, where the 3dxz
and 3dyz states are again constrained to be degenerate by
symmetry. There is a further increase in the atomic occupation
of these orbitals accompanied by a much larger reduction in
their COOP values (from 0.489 in Li0.99(5)Y2Ti2O5S2 to 0.468
in Li1.85(5)Y2Ti2O5S2), signaling the onset of weakening in Ti-
O1 π-bonding. Our calculations cannot reliably predict the
minimum level of intercalation necessary to initiate the ortho-
rhombic-to-tetragonal transition (which must lie in the range
1.3 < x < 1.5), but the results show that there is a limited
amount of electron filling where the distortion to orthorhombic
symmetry allows the antibonding dxz bands to remain unoc-
cupied; at greater electron counts, the incentive for the above-
described distortion vanishes.
The decrease toward 90° of the O2-Ti-O1 angles with
increasing x (Figure 8), decreases the antibonding nature of the
3dxz and 3dyz orbitals. Any decrease in the bond strength of the
Ti-O framework is compensated by an increase in Ti-S
bonding. Indeed, the most substantial structural change during
the course of intercalation occurs in the Ti-S distance (Figure
6), which decreases from 2.8741(6) Å in Y2Ti2O5S221 to 2.512(5)
Å in Li1.85(5)Y2Ti2O5S2. The shortest distance is comparable to
the bond lengths observed in related titanium sulfides (LiTiS2
2.481 Å; Li0.33TiS2 2.509 Å)36 and indicates a full-fledged Ti-S
bond. The COOP curves in Figure 11d show the progression
of increased Ti-S bonding with increasing Li content. For x >
1, the Ti-S bond strength apparently exceeds that of any Ti-O
interaction, doubling in value between x ) 0 and x ) 1.85(5).
At x ) 1.85(5), the Ti-S bond has reached its maximum
potential, and further electrons entering the conduction bands
(36) (a) Patel, S. N.; Balchin, A. A. Z. Kristallogr. 1983, 164, 273. (b) Dahn,
J. R.; McKinnon, W. R.; Haering, R. R.; Buyers, W. J. L.; Powell, B. M.
Can. J. Phys. 1980, 58, 207.
9
J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004 1989

A R T I C L E S
Hyett et al.
as shown in Figure 4. The Li-O bond lengths are shown as a
function of x in Figure 7. Although the Lib-O2 distance,
corresponding to half of the b-axis length, is slightly longer
(1.95941(2) Å) than the corresponding distance for occupation
of the Lia site (1.913 Å), the decrease in the O2-Ti-O1b angle
(95.90(2)°) relative to the O2-Ti-O1a angle (101.74(2)°),
described above, makes the axial O1b-Lib contact much shorter
(2.0884(4) Å) than the O1a-Lia alternative (2.281 Å). Thus,
the electronically driven orthorhombic distortion is reinforced
by providing more favorable electrostatic Li-O cohesion.
Increasing the lithium content above 1.0 must entail partial
filling of the Lia site, although we have not investigated this in
detail.
Interpretation of Phase Gap Behavior. At the compositions
close to the change from tetragonal to orthorhombic symmetry
(x ≈ 0.7), there is a phase gap with coexistence of a tetragonal
phase and a Li-richer orthorhombic phase. This is good evidence
that the change from tetragonal to orthorhombic symmetry is
electronically driven rather than merely driven by the coordina-
tion preferences of Li+. Presumably the two phases identified
as representing the edges of the phase gap have similar chemical
potentials, which are both lower than that of a hypothetical single
tetragonal phase with the chemical composition of the sample.
At elevated temperatures (above 200 °C), a single tetragonal
phase does exist, and our variable temperature PND study shows
that this disproportionation is reversible. This transformation
requires that the lithium ion has a high mobility below 200 °C.
The two phases present at room temperature in the sample
Li0.66(5)Y2Ti2O5S2 have very similar unit cell volumes despite
their different lithium contents (Figure 2). Since their lattice
parameters were measured in a single experiment this must
reflect a true departure from the general trend in increasing cell
volume with electron count (Figure 2). The Ti-O bond lengths
in the two-phase region also depart from the general trends
(Figure 6); the closely similar mean Ti-O bond lengths in the
two-phase region of 1.933(2) Å and 1.934(2) Å for the tetragonal
and orthorhombic phases, respectively, should be compared with
the linear 3% increase in the mean Ti-O distance from 1.913(1)
Figure 11. (a) Molecular orbital picture for the Ti 3d-derived orbitals of
a fragment obtained by joining two TiO5S octahedra at a common oxide
vertex using geometries obtained by Rietveld refinement against neutron
data for LixY2Ti2O5S2 x ) 0.30(5) (left), tetragonal; 0.99(5) (center),
orthorhombic; and 1.85(5) (right), tetragonal. Extension to the two-
dimensional Ti2O5S2 slabs produces the Band structures (b), DOS (c) and
COOP curves (d) for the three compositions. In (c), the individual and total
Ti 3d contributions to the DOS are indicated as shaded regions; the dashed
lines incorporate oxygen states, and the outer solid line is the total DOS
incorporating the sulfur states. In (d), the COOP curves for Ti-O2 apical
(dark-shaded/solid line), Ti-O1 equatorial (medium-shaded/short-dashed
line), and Ti-S (light-shaded/long-dashed line) are shown with correspond-
ing integrated curves.
21
Å in Y2Ti2O5S2 to 1.974(1) Å in Li1.85(5)Y2Ti2O5S2. These
observations are consistent with a transition from localized to
itinerant electron behavior with increasing x, separated by the
two-phase region, with the Li-richer metallic phase having a
smaller cell volume and shorter mean Ti-O distance than
expected for an imaginary isoelectronic insulating phase.37
Coincidence of the insulator to metal transition with the
tetragonal to orthorhombic distortion is also consistent with the
electrical resistivity and magnetic susceptibility measurements
described above and with the colors of the materials. In the
analogous sodium intercalates21 in which Na+ is located in the
12-coordinate site, the cooperative distortion observed here does
not occur, rather local lifting of the C4V symmetry is manifested
as crystallographic disorder in the equatorial oxygen positions
(i.e., O1), as described elsewhere.21
would begin to occupy Ti-S π* antibonding states. Thus, the
Y2Ti2O5S2 host can accommodate close to one Li per Ti without
losing much in the overall cohesion by a gradual decrease in
the Ti-S distance, which has a very limited influence in the
Ti-O framework in the ab plane.
Thus, the degree of orthorhombic distortion also increases
with electron count to maintain the edge of the unfilled 3dxz
above the Fermi energy. In a solid-state system, the extent of
such a distortion is constrained by the other parts of the structure,
so at high electron counts, corresponding to high levels of Li
intercalation, partial filling of the 3dxz band is unavoidable. The
transition to orthorhombic symmetry is electronically driven by
lifting the degeneracy of the Ti 3dxz and 3dyz and occurs with
retention of the unit cell (aside from the slight increase in
volume). It should thus be understood as analogous to a
cooperative Jahn-Teller type distortion.
Comparison with Other Lithium Intercalates. The struc-
tures of the analogous lithium intercalates of WO3 and ReO3
have been established previously.38-40 In the former case,
(37) Goodenough, J. B. Struct. Bonding (Berlin) 2001, 98, 1.
(38) Wiseman, P. J.; Dickens, P. G. J. Solid State Chem. 1976, 17, 91.
(39) Cava, R. J.; Santoro, A.; Murphy, D. W.; Zahurak, S.; Roth, R. S. J. Solid
State Chem. 1982, 42, 251.
(40) Cava, R. J.; Santoro, A.; Murphy, D. W.; Zahurak, S.; Roth, R. S. J. Solid
State Chem. 1983, 50, 121.
In the orthorhombic compound Li0.99(5)Y2Ti2O5S2, measured
on POLARIS as well as those with similar compositions
measured on D2B, lithium occupies only one of the two
available four-coordinate planar sites in the perovskite blocks,
9
1990 J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004

Electronically Driven Distortions in Li Intercalates
A R T I C L E S
Table 7. Average Values of Crystal Orbital Overlap Populations (COOPs) and Bond Lengths (Å) for LixY2Ti2O5S2
bond
x ) 0
x ) 0.30
0.570
(1.819(2))
0.497
(1.9466(4))
0.497
x ) 0.99
0.546
(1.8861(5))
0.489
(1.9569(1))
0.488
(1.9698(1))
0.581
(2.6791(8))
0.225
(2.8781(2))
-0.001
(3.83189(3))
-0.010
(3.91881(3))
0.003
x ) 1.85
Ti-O2 (apical)
0.588
0.516
(1.950(3))
0.468
(1.9801(3))
0.468
(1.7941(4))a
0.500
Ti-O1a (equatorial)
Ti-O1b (equatorial)
Ti-S (apical)
(1.9427(1))a
0.500
0.407
0.457
(2.810(3))
0.250
(2.823(1))
0.007
(3.80002(2))
0.007
0.748
(2.512(5))
0.193
(2.920(1))
-0.016
(3.94260(8))
-0.016
(2.8741(6))a
0.258
Y-S
(2.8048(2))a
0.021
Ti-Ti (along a)
Ti-Ti (along b)
Ti-Ti (along c)
(3.76956(1))a
0.021
0.024
(3.588(1))a
0.018
(3.638(3))
-0.011
(3.888(4))
(3.768(1))
a
Results from ref 21.
Table 8. Atomic Orbital Occupations Derived from Extended
Hu¨ckel Calculations
in terms of geometry, coordination number, and Li-O separa-
tions with that observed in other Li-intercalated oxides. The
distortions displayed by the Li-poor and Li-rich intercalates of
WO3 and ReO3 are a response to the coordination preferences
of Li+ and are available on account of the fact that the WO6
and ReO6 octahedra share vertices only, allowing distortion by
relative tilting of the octahedra. Although the TiO5S polyhedra
in Ln2Ti2O5S2 share vertices, the layered nature of the structure
and the restraining presence of the YS “rock salt” layers
precludes such distortions.
Ti orbital
x ) 0
x ) 0.3
x ) 0.99
x ) 1.85
3dxy
3dxz
3dyz
0.323
0.356
0.356
0.459
0.374
0.465
0.362
0.362
0.469
0.375
0.649
0.372
0.520
0.490
0.377
0.819
0.592
0.592
0.511
0.376
2
3dz
2
2
3dx -y
intercalation up to Li0.36WO338,40 results in distortion of the WO3
framework and accommodation of lithium in a four-coordinate
square planar site. This site is in the center of the 12-coordinate
vacant cation site in WO3, and the tilting distortion of the WO6
octahedra is such that four of the oxide anions surrounding the
vacancy move toward the original 12-coordinate site and
coordinate to Li (Li-O distances: 2.192(2) Å × 4). Similar
behavior is observed in Li0.2ReO3.40 The 4-coordinate site
occupied in the LixY2Ti2O5S2 intercalates lies, in contrast,
between two 12-coordinate vacancies, and the mean Li-O bond
lengths (Figure 7) are rather shorter than those in the WO3 and
ReO3 intercalates but are similar to the sum of the Li+ and O2-
radii. In Li2FeV3O8,41 the Li intercalate of FeV3O8, Li occupies
an approximately square pyramidal site with a mean Li-O
distance of 2.04(3) Å. In the Li-rich intercalates of ReO3
(LiReO3 and Li2ReO3),39 there are very significant tilting
distortions of ReO6 octahedra which result in each 12-coordinate
vacant cation site transforming to a pair of face-shared
octahedral sites which are all filled in Li2ReO3. Li is displaced
from the centers of these octahedral sites (Li-O distances of
1.95(1) × 3 and 2.30(2) × 3). The coordination environment
observed in the LixLn2Ti2O5S2 intercalates is thus comparable
Conclusion
The LixLn2Ti2O5S2 compounds offer an unusual example of
a series in which the rigidity of the two-dimensional framework
and the relatively low steric demands of the lithium ion permit
the electronic effects of intercalation to dominate the details of
the structure. The system shows qualitative similarities to the
La1-xSrxTiO3 perovskite system26,37 in which a metal-to-insulator
transition occurs across the two-phase region 0.05 e x e 0.08
and is complementary to other two-dimensional reduced titanates
such as the NaTiO2 system27 in which an electronically driven
structural distortion is evident in a system containing layers of
edge-sharing TiO6 octahedra.
Acknowledgment. We thank the U.K. EPSRC (grant GR/
N18758) for funding and access to ISIS and ILL. S.J.C. thanks
the Royal Society and the Nuffield Foundation for further
financial support. We are grateful to Dr. Alan Hewat and Mr.
Peter Cross for assistance with measurements carried out at ILL
and to Dr. Ron Smith for assistance with measurements carried
out at ISIS.
(41) Cava, R. J.; Santoro, A.; Murphy, D. W.; Zahurak, S.; Roth, R. S. J. Solid
State Chem. 1983, 48, 309.
JA037763H
9
J. AM. CHEM. SOC. VOL. 126, NO. 7, 2004 1991